Public Disclosure Authorized

INU- 83
The World Bank

Policy, Planning and Research Staff

Infrastructure and Urban Development Department

Public Disclosure Authorized

Report INU 83

WELFARE ANALYSIS OF RENT CONTROL IN BRAZIL

THE CASE OF RIO DE JANEIRO
Public Disclosure Authorized

by

Ricardo Silveira L I
and

Stephen Malpezzi
Public Disclosure Authorized

June 1991

DISCUSSION PAPER

This is a document published informally by the World Bank. The views and interpretations herein are those of the author and should not be
attributed to the World Bank, to its affiliated organizations, or to any individual acting on their behalf.
Copyright 1991 o
The World Bank
1818 H Street, N.W.

All Rights Reserved

First Printing June 1991

This is a documentpublished informallyby the World Bank. In order that the informationcontained in
it can be presentedwith the least possibLedelay, the typescripthas not been prepared in accordancewith the
proceduresappropriateto formal printed texts, and the World Bank accepts no responsibilityfor errors.

The World Bank does not accept responsibilityfor the views expressedherein, which are those of the
author and should not be attributed to the World Bank or to its affiliatedorganizations. The findings,
interpretations,and conclusionsare the results of research supportedby the Bank; they do not necessarily
representofficial policy of the Bank. The designationsemployed,the presentationof material,and any maps
used in this document are solely for the convenienceof the reader and do not imply the expression of any
opinion whatsoeveron the part of the World Bank or its affiliatesconcerningthe legal status of any country,
territory,city, area, or of its authorities,or concerningthe delimitationsof its boundariesor national
affiliation.

Ricardo Silveirais an economistwith the Populationand HumanResourcesDivisionof the SouthernAfrica

Departmentof the World Bank. Stephen Malpezzi is on the faculty of the Departmentof Real Estate and Urban
Land Economicsof the Universityof Wisconsin,Madison. Before joining that departmenthe was an economist in
the Urban DevelopmentDivision of the World Bank.

The authors would like to thank RobertBuckley,MichaelCohen, EmmanuelJimenez,StephenMayo, Margaret

Thalwitz,and ProfessorDavid Dowall for cormments on an earlier draft of this paper. They are not responsible
for remainingerrors. Gwendolyn Ball and William Stephens provided computationalassistance.

This paper has been producedunder a World Bank researchproject on Rent Control in DevelopingCountries
(RPO 674-01), directed by Stephen Malpezzi. Other papers produced by that project include:

Stephen Malpezzi and C. Peter Rydell, Rent Controls: A Framework for Analysis. World Bank, Water
Supply and Urban DevelopmentDepartmentDiscussionPaper No. 102, 1986.

Stephen Malpezzi, Stephen K. Mayo, Ricardo Silveira and Carmela Quintos. Measuring the Costs and
Benefitsof Rent Control: Case Studv Design. World Bank, INU DiscussionPaper No. 24, 1988.

Stephen Malpezzi and Vinod Tewari. Costs and Benefits of Rent Regulationin Bangalore.India. World
Bank, INU DiscussionPaper, forthcoming.

Stephen Malpezzi, Graham Tipple and Kenneth Willis. Costs and Benefits of Rent Control in Kumasi,
Ghana. World Bank, INU DiscussionPaper No. 51, 1989.

Stephen Malpezzi and Gwendolyn Ball., Rent Control in DevelopingCountries:A Synthesis,World Bank,
PRE Working Paper, forthcoming.

In additionto the authorsof the present paper,many peoplecontributedto this project. Peter Rydell
coauthoredthe original design. Ricardo Silviera,Vinod Tewari, Graham Tipple and Kenneth Willis coauthored
the case studies. Waleed El-Ansary,Nachrowi, William Stephens and Carmella Quintos were among those who
providedcomputationaland other assistance. Informationon rent control laws and housingmarkets in various
countrieswas provided by many individuals,includingSamuel Afrane, Richard Arnott, Samuel Boapeah, Robert
Buckley, Bruno De Borger, Manfred Fischer, Frederik Hansen, Dennis Keating, Krishna Kumar, Saitel Kulaba,
ZewdinehLakew,StephenMayo, L.M. Menezes,Eduardo Neto,Bakar Nominuma,Yomi Oruwari, Ayse Pamuk, John Prince,
E.F.N. Ribeiro, Louis Rose, Peter Rydell, Bo Sandelin,N. Sridharan,Raymond Struyk, Carolyn Tager, Martha
Tamtakoe,Graham Tipple, Bengt Turner,Margery Austin Turner, Elia Werczberger,and Jun Zhang. Most certainly
they are not responsiblefor our shortcomingsin interpretingand making use of the informationthey so kindly
provided.

Managerial support from David DeFerranti,Per Ljung and Michael Cohen is gratefully acknowledged.
Detailed comwents on the project and on the case study papers were provided by many Bank staff, researchers,
and country policy makers, includingRichard Arnott, Esra Bennathan,Robert Buckley, Dennis De Tray, William
Dillinger,Gregory Ingram,Emnanuel Jimenez,Johannes Linn,Per Ljung, StephenMayo, William McGreevey,Rakesh
Mohan, MichaelMurray, EdgarOlsen, BertrandRenaud, Evan Rofner, Margery Turner, and Anthony M.J. Yezer, among
others. So many other people made useful commentsthat they cannot be acknowledgedindividually. Finally,
Stephen Mayo's contributionsto the formulationand executionof this project were legion.
 The World Bank

WELFARE ANALYSIS OF RENT CONTROL IN BRAZIL

THE CASE OF RIO DE JANEIRO

DISCUSSIONPAPER
 WELFARE ANALYSIS OF RENT CONTROL IN BRAZIL
THE CASE OF RIO DE JANEIRO

Table of Contents

Page No.

I. SUMMARY AND CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . 1

A. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 1
Objectives of the Study . . . . . . . . . . . . . . . . . . . 1

B. Key Questions . . . . . . . . . . . . . . . . . . . . . . . . . 2
Evolution of Housing and Rent Control Institutions . . . . . 2
Costs and Benefits of Rent Control . . . . . . . . . . . . . 3
DistributionalImpact of Rent Control . . . . . . . . . . . . 5
Rent Control, Decontrol and the Housing Supply . . . . . . . 6
Future Work ............ 7

II. THE EVOLUTION OF RENT LEGISLATIONIN BRAZIL . . . . . . . . . . . . 9

A. Early Legislation . . . . . . . . . . . . . . . . . . . . . . . 9
Pre-Rent Control Regulation of the Rental Market . . . . . . 9
First Rent Control Laws ... 9
Impact of the First Rent Control Laws . . . . . . . . . . . 10

B. Populist Regulations . . . . . . . . . . . . . . . . . . . . 11
Rigid Rent Control Under Law-Decree4598 and Extensions . . 11
New Contracts Exemption Under Law No. 1300 and Extensions . 11
Impact of Populist Rent Regulations . . . . . . . . . . . . 12

C. Gradual Decontrol and Indexationof Rent . . . . . . . . . . 14

Impact of Decontrol ....... ........... . . 14

D. Return of Rent Controls ....... . .. .. .. .. . . . 16

Impact of Re-TighteningRent Controls . . . . . . . . . . . 16
Adjusting to HyperinflationUnder Rent Control . . . . . . 19

III. COSTS AND BENEFITS OF RENT CONTROL: THEORY AND METHOD . . . . . 20

A. Analysis of Rent Control as a Tax on Housing . . . . . . . . 20

B. Rent Control as ExpenditureControl . . . . . . . . . . . . . 20

C. Rent Control. Mobility and Tenure Security . . . . . . . . . 21

D. The RelationshipBetween Controlledand UncontrolledMarkets 22

 - ii -

E. Measuring Costs and Benefits of Rent Control:

Cross-CountryModel .. 23
Renters. . . . . . . . . . . . . . . . . . . . . . . . . 25
Owners ......................... . 25
UncontrolledRenters .................. . 26
ControlledRenters ............ . 26

F. Measuring Costs and Benefits of Rent Control:

Olsen's Model . . . . . . . . . . . . . . . . . . . . . . . . 27
Hedonic Estimates ............ . 30

G. Empirical Problems . . . . . . . . . . . . . . . . . . . . . 30

H. Cost-BenefitStudies . . . . . . . . . . . . . . . . . . . . 33

I. Rent Control and the Housing SupplY . . . . . . . . . . . . . 36

Reduced Maintenance . . . . . . . . . . . . . . . . . . . . 37

J. Rent Decontrol .... . . . . . . . . . . . . . . . . . . . . 39

IV. COSTS AND BENEFITS OF RENT CONTROL: DATA AND EMPIRICAL RESULTS . 42

A. Household Survey Data . . . . . . . . . . . . . . . . . . . . 42

Housing Stock Characteristics. . . . . . . . . . . . . . . 42
Household Characteristics. . . . . . . . . . . . . . . . . 43

B. Estimating Costs and Benefits of Rent Control . . . . . . . . 45

Choice of Reference Group .... . . ..... . . . . . . 46
Controlled and UncontrolledHouseholdsCompared . . . . . . 47

C. Rent Paid in the ControlledSector: PcO . . . . . . . . . . . 49

D. Estimating PmOm With a Cross Country Model of Housing Demand 50

UncontrolledRents Compared to Predictionsfrom
the Cross-CountryModel .... . . ..... . . . . . . 50

E. Estimating PmOm With Rio Survey Data . . . . . . . . . . . . 51

F. Estimating PmOc With Hedonic Indexes . . . . . . . . . . . . 52

G. Cost-BenefitMeasures .... . . . ...... . . . . . . . 52

Costs and Benefits from the Cross-CountryModel . . . . . . 52
Costs and Benefits Constructedfrom the Hedonic
and Demand Equations .... . . ...... . . . . . . 55

V. POLICY ISSUES . . . . . . . . . . . . . . . . . . . . . . . . . . 57

A. Distribution of Costs and Benefits from Rent Control . . . . 57

Distributionby Landlord and Tenant . . . . . . . . . . . . 57
Distributionby Income . . . . . . . . . . . . . . . . . . 58
Distributionby Class of Tenure: Formal and
Informal Markets .... . . . . ..... . . . . . . . . 60
 - iii -

B. A Simple Model of Tenure Choice . . . . . . . . . . . . . . . 62

C. A Present Value Model of Housing Investment in

Rio de Janeiro ...................... . 64
A Present Value Model of Housing Investment in
Rio de Janeiro .................... . 65
Gains and Losses from Four Components and
Their Interaction .................. . 70
Effects of Rent Control on Landlord Profitability . . . . . 73
Changes in Risk and Investment Decisions . . . . . . . . . 74
Changes in Rates of Return and Changes in Supply . . . . . 75
Alternative Decontrol Options . . . . . . . . . . . . . . . 78

REFERENCES ............................. . 82

ANNEXA ............................. . 88

BOXES

Box 1.1: Legislative Framework of Rent Control in Brazil

Summary of Main Events .. 4
Box 2.1: Modeling Landlord Pricing Behavior Under Law No. 6698 . 17
Box 2.2: Full Circle to Free Rental Markets . . . . . . . . . . 19
Box 5.1: Marginal Changes, Major Effects . . . . . . . . . . . . 79
Box 5.2: Closed Doors.. . ...... 80

FIGURES

Figure 2.1: Licenses to Build and Area Licensed . . . . . . . . . . 12

Figure 2.2: Construction Costs . . . . . . . . . . . . . . . . . . 12
Figure 2.3: Licenses to Build, Area Licensed
and Construction Costs . . . . . . . . . . . . . . . . 16
Figure 3.1: Rent Control with Elastic Supply . . . . . . . . . . . 20
Figure 3.2 Rent Control as Expenditure Control . . . . . . . . . . 21
Figure 3.3: Rent-to-Income Ratios, Cross-Country Model . . . . . . 24
Figure 3.4: Rent Control and Consumer's Surplus . . . . . . . . . . 27
Figure 3.5: Rents and Length of Tenure Under Rio's
System of Controls . . . . . . . . . . . . . . . . . . 32
Figure 3.6: Stylized Decomposition of Rental Housing Stock . . . . 36
Figure 3.7: Alternative Decontrol Options, Los Angeles . . . . . . 41
Figure 4.1: Characteristics of Sample Housing Stock . . . . . . . . 43
Figure 4.2: Characteristics of Sample Households . . . . . . . . . 44
Figure 4.3: Rent-to-Income Ratio by Income Quartile . . . . . . . . 45
Figure 4.4: Rent-to-Income Ratio in Controlled and Uncontrolled
Households .. 48
Figure 4.5: Rent and Income Levels Compared . . . . . . . . . . . . 50
Figure 4.6: Rent-to-Income Ratio Results Compared . . . . . . . . . 51
Figure 4.7: The Distribution of Cost-Benefit Measures . . . . . . . 54
Figure 4.8: Costs and Benefits from Rent Control, Controlled
Households ...................... . 56
 - iv -

Figure 5.1: Tenant Benefit Results from the Olsen and the Cross-
Country Models . . . . . . . . . . . . . . . . . . . . 57
Figure 5.2: Tenant Benefits by Length of Tenure . . . . . . . . . . 58
Figure 5.3: Rent Distribution by Income and Time Distribution . . . 59
Figure 5.4: Rent-to-Income Ratio by Length of Stay . . . . . . . . 60
Figure 5.5: Tenant's Benefits by Income Quartile . . . . . . . . . 62
Figure 5.6: Distribution of Losses in the Formal and Informal Units 63
Figure 5.7: Urban Tenure and Percent Urban . . . . . . . . . . . . 63
Figure 5.8: Urban Tenure and Per Capita . . . . . . . . . . . . . . 64
Figure 5.9: Percent Renters in Brazil, Actual and Predicted . . 71
Figure 5.10: Real Rental Prices and Share of Rentals . . . . . . . . 78
TABLES

Table 2.1: Index of Rental Prices in Rio de Janeiro: 1912-30 10

Table 2.2: Index of Rental Prices in Rio de Janeiro: 1930-64 13
Table 2.3: Index of Real Rental Prices in Rio de Janeiro: 1966-74 15
Table 2.4: Index of Rental Offer Prices in Rio de Janeiro: 1975-86 18
Table 3.1: Cross-Country Demand Results . . . . . . . . . . . . . 25
Table 4.1: Water and Sewage Coverage in Rio and Latin America . .
Table
44
4.2: Housing Structure Samples Compared . . . . . . . . . . 48
Table 4.3: Household Samples Compared . . . . . . . . . . . . . . 49
Table 4.4: Controlled and Uncontrolled Rents Paid by Sector
(Cruzeiros) .49
Table 4.5: Demand Equations . . . . . . . . . . . . . . . . . . . 52
Table 4.6: Hedonic Index . . . . . . . . . . . . . . . . . . . . . 53
Table 4.7: Summary Cost-Benefit Measures
from Cross-Country Model . . . . . . . . . . . . . . . 54
Table 4.8: Costs and Benefits Constructed
from the Hedonic and Demand Equations . . . . . . . . .
55
Table 5.1: Household Samples Compared . . . . . . . . . . . . . . 61
Table 5.2: Simple Cross-Country Tenure Regression . . . . . . . . 63
Table 5.3: Key Model Inputs .67
Table 5.4: Cash Flow Model of Rental Investment . . . . . . . . . 68
Table 5.5: Demand Side of Rental Investment Model . . . . . . . . 69
Table 5.6: Summary Returns . . . . . . . . . . . . . . . . . . . .
73
Table 5.7: Estimated Long-Run Price Elasticities of
Supply of Housing Supply . . . . . . . . . . . . . . . 76
Table 5.8: Impact of Rent Control Regimes: 1912-85 ........ . 77
 WELFAREANALYSISOF RENT CONTROLIN BRAZIL
THE CASE OF RIO DE JANEIRO

I. SUMMARY
AND CONCLUSIONS

A. Introduction

1.1 Rent control is a familiar institution in many cities of the United

States and Europe. It is also more common in developing countries than is
generallyperceived. In Brazil,for example,formal systemsof rent controlhave
been in place, in all states, since the 1920s. This paper is a case study which
presents empirical estimates of the costs and benefits of rent control in one
Brazilian city, Rio de Janeiro ("Rio" for short).

1.2 While a considerableliteratureexists on Brazilianhousing markets in

general, rental markets have been neglected. To our knowledge no study has
estimated the effects of controls with the rigor which we attempt here. This
paper is of particular interest to those concerned with the effects of rent
control in Brazil; but it is also part of a larger research project on 'Rent
Controls in DevelopingCountries,"which will compare the effects of different
rent control regimesunder differentmarket conditionsin a number of countries.
That research has already demonstratedthat rent controls differ markedly in
their effects depending on specific provisions, enforcement and market
conditions. Thus far, three case studies examining the rental markets of Egypt,
Ghana and India have been completed. Forthcoming papers from the ongoing
research project will synthesizeresults and provide comparative information.

Objectivesof the Study

1.3 The objectives of this paper are to review the evolution of rent
control legislationin Brazil, to estimatethe costs and benefits of controlsin
the specificmarket of Rio, and to evaluate the policy implicationsof present
and alternative rent control systems. No less important is the effort to
highlight where additional research would shed light on these issues. In
particular,we hope the present paper will motivate additionaldata collection
and analysis on a number of Brazilian markets. The 1990 census should provide
an invaluablebenchmark against which to compare the resultswe put forward for
1980. It also offers a chance to look into the dynamics of rent control
legislation and welfare.

1.4 While this paper is the study of one market in particular, it will
indicatewhich findingsare likely to be robust and generalizableacross markets
and which findings are likely to be peculiar to this regime. In addition to
presenting results for a large market at a particular time, and attemptingto
indicate which results are applicable to other times and places and which are
not, this paper aims to serve as a model for future empirical research on rent
controls in Brazil.
 -2 -

B. Key Issues

1.5 More specifically,the paper seeks to illuminate the followingbroad

issues and related questions:

Evolution of Housing and Rent Control Institutions

What is rent control like in Brazil and how did rent control
legislationand housing institutionsevolve into the present
regime?

1.6 Concerted housing policy in Brazil dates back to the 1920s when the
government began to address urban health problems through building codes and
infrastructureregulations. During the early years of the republic,the private
rental sector was crucial in the provision of housing, and the interaction
between tenants and landlordswas completelyfree of governmentregulation. The
first attempt at regulatingthe rental market is found in the civil code of 1917.
However, it is not until 1921 that the first set of controlswas imposed on how
rental prices could be set. This original rent control law was very mild and
primarily concernedwith temporaryrestraintson evictionsand rent adjustments.
The law remained in place for six years, initiating a Brazilian practice of
continuousextensions of temporary controls.

1.7 A series of timid institutionalreforms in the housing sector was

initiated in the 1930s. Unfortunately,all low-incomehousing programs in this
populist era eventually faced financialcollapse and were only able to reach a
small segment of the intendedpopulation. The first governmenthousing finance
mechanisms (IAPs) were put into place during the Vargas administrationwith
little success. PresidentDutra introduceda complementaryprogram to the IAPs
in the mid 1940s--the so-called "Popular Housing Foundation"--butlarge loan
subsidies and the government's inability to raise funds ensured a poor
performance of the program. The most stringent rent control laws to date were
enacted during the populist period. The war environment of scarcity and
widespread rationing facilitated the creation of a rigid rent control law in
1942. Except for minor adjustments, rents remained frozen for nearly two
decades, forcing landlordsto adapt to the system by stretchingthe law in their
favor or by simply exiting the market. The share of rented dwellings in total
housing declined from 70 percent in the 1920s to 43 percent in the early 1960s.
The anti-landlordstance of the populistgovernments,coupled with their failure
to deliver a meaningfulhousing policy or, at least, to improve long-term credit
conditions, led to the stagnation of low-income housing supply. Since the
populistyears were times of rapid urban growth (fueledby fast industrialization
and the exodus from the rural Northeast),equilibriumin the housing market was
reached only through an unprecedentedboom in the informalhousing sector.

1.8 Housing institutionswere substantiallystrengthenedduring the early

years of the militaryregimesby the creationof the now-defunctNationalHousing
Bank (BNH), responsiblefor executingthe NationalHousing Plan and managing the
NationalHousing Finance System (SFH). Financingfor the SFH came primarily from
passbook savings,real estate bonds and workers'contributionsto a multipurpose
fund (FGTS). The impact of the new housing institutionson housing finance has
been substantial,with over four million houses financed through the system
 -3 -

during its first twenty years of existence. However, the system has been unable
to adequately meet the needs of the low-income population and less than
successful at coping with high inflation. As a result, the formal housing
deficit continued to grow, though certainly at a slower pace than during the
populist years. Controls on commercial and residential rental units were
considerablyrelaxed during most of the militaryperiod, and the market responded
by increasing the supply of rental units. As a result, real rental prices
stagnated and even declined in some submarkets - - despite a strong surge in
aggregate demand during the economic boom of 1968-80. In the late 1970s and
throughoutthe 1980s, the market experienceda resurgenceof controls. Recently,
they have become more severe and constraining. Current controls in Brazil are
not the strict sort found in some countries, where governments may set, or
attempt to set, rent levels, but rather, they are regulations on changes in
rents.1' But as inflation accelerated,making partial indexationmore binding,
and as financial investmentopportunitiesexpanded, some landlords reacted by
employing "creative" leasing arrangementsand others by withdrawing from the
market.

1.9 Rent controls currently in force in Brazil are less stringent, in

general, than those in other countries studied in the comparative research
project. In particular, Brazilian rent levels are not controlled directly by
legislation,rent increases are. Such rent indexationis generally less strict
a form of control than, say, the setting of "fair rents" for some Indian units
or the controls on rent levels per room in Ghana.21

Costs and Benefits of Rent Control

What are the static costs and benefits of controls from the
point of view of tenants and landlords in Rio? How do
changes in rents and housing consumptionaffect the welfare
of typical individuals?

1.10 A typical household under rent indexation (controlled)pays a rent

amount not too different from what it would pay if market conditionsprevailed.
The rent paid by the median householdis 88 percent of the estimatedmarket rent.
Depending on assumptionsabout the price elasticityof demand (Ep), this modest
discount translates into a median net loss of between Cr$95 to Cr$356, once
changes in housing consumption are taken into account. For a stylized
"representativetenant", there is a modest positive benefit of Cr$374.

1.11 As expected,rent indexationimposesa statisticallymeasurable,though

not extraordinary,cost to the landlord of a controlledunit. The median cost
of the subsidy is estimated to be about Cr$175 per month or 6% of the actual
rent. In the case of the landlord to the stylized "respresentativetenant",
however, the loss is over 13 percent of the actual rent.

1/See Malpezzi and Ball (1990) for a classificationof countrieswhich set "fair
rents" versus those that regulate increases.

_/Of course it is possible to construct examples where partial indexation of

rents in highly inflationaryenvironmentsis more restrictivethan rent freezes
or rent level setting in other environments.
 Box 1.1: LEGISLATIVE FRAMEWORKOF RENT CONTROL IN BRAZIL SUMMARYOF MAIN EVENTS

PERIOD LEGISLATION DESCRIPTION

PRE 1917 NONE NO RENT CONTROL: PRIVATE AGREEMENTS CONFORMING WITH

TRADE PRACTICES.

1917-21 CIVIL CODE NO RENT CONTROL: FREEDOM OF CONTRACT PERIOD AND OF

RENTAL PRICE.

1921-27 LAW No.4403 OF WEAKRENT CONTROL:

12122/1921 FREE NEGOTIATION OF INITIAL
RENT; CONTROLS ON CONTRACT
EXTENSIONS; RENT ADJUSTMENTS
WITH TWO-YEAR PRIOR NOTICES;
SPECIFIES CONDITIONS FOR
EVICTION.

1927-42 RETURN TO CIVIL NO RENT CONTROL:

CODE ON 12/31/1927 FREEDOM OF TERM OF CONTRACT
AND OF RENTAL PRICE.

1942-50 LAW-DECREE 4598 OF STRONG RENT CONTROL:

8/20/1942 FREEZES RENTS AT LEVEL OF
1213111941, ESTABLISHES
CONTRACTUAL EXTENSIONS FOR
INDEFINITE PERIODS; SPECIFIES
CONDITIONS FOR EVICTION.
EXTENDED UNTIL MINOR ADJUSTMENTS TO RENTS AND
1950 BY A SERIES OF TO CONDITIONS FOR RECOVERY OF
LAWS PROPERTY.

1950-64 LAW No.1300 OF MODERATERENT CONTROL:

12/28/1950 INTRODUCES FREE NEGOTIATION OF
INITIAL RENT: ALLOWS INDEXATION
OF RENTAL CONTRACTS

1964-1967 LAW No.4494 OF MODERATERENT CONTROL:

11125/1964 ADJUSTS PREVIOUSLY FROZEN
RENTS; LINKS RENT ADJUSTMENTS
TO MINIMUM WAGE.

1967-77 LAW-DECREE 5334 OF WEAKRENT CONTROL:

4/7/1967 AND RETURNS ALL NEW LEASES TO
LAW No.5334 OF CIVIL MODE.
10/21/1967

1977-79 LAW-DECREE 1534 OF WEAK/MODERATERENT CONTROL:

4/13/1977 EXTENDS CONTRACTS FOR UP TO
24 MONTHS; SUSPENDS 'EVICTIONS
WITHOUT CAUSE' (DENUNCIA VAZIA);
RENTS ADJUSTED BY VARIATIONS IN INDEXED TREASURY NOTES.

1979-86 LAW No.6649 OF MODERATE/WEAK RENT CONTROL:

511611979 ABOLISHES DENUNCIA VAZIA ALLOWS
JUDICIAL REVIEW OF RENTS EVERY
FIVE YEARS (BRINGING RENTS TO
MARKET VALUES).

1986-PRESENT LAW-DECREE 2284 STRONGIMODERATERENT CONTROL:

OF 3/10/1986, AND FREEZES ALL RENTS FOR A PERIOD
REVISED/REGULATED OF ONE YEAR; FREEZE SUSPENDED
BY A SERIES OF ON 3/1/1987 AND FUTURE
LAW-DECREES ADJUSTMENTS BASED ON TREASURY
NOTES.
1.12 The general conclusionfrom the static cost-benefitanalysis is that
while society is made worse-offby controls,with many tenants and most landlords
of controlled units experiencing some degree of loss, these losses are not
remarkable. Under the more favorable assumption (i.e., Ep--l), society's
aggregate losses amount to an average of Cr$146 (US$2.75) per household per
month, and Cr$285 (US$3.38)under Ep--0.5.

1.13 While losses are still significantrelative to rents paid, they are
small relative to income. The cost of the subsidy from landlords is in the order
of 1 percent of median tenant income, while tenant losses are 0.5 percent of
income. Although measurable, it is clear that static welfare costs are much
lower in Rio in 1980 than in other rent control regimes surveyed by the rent
control project.

DistributionalImnact of Rent Control

What are the static distributionalimplicationsof controls?

Who are the losers and winners?

1.14 Because the income distributionof Rio is very unequal, the impact of
rent indexationon relativeincomes is a crucial issue always present in the rent
control debate. Since we do not know the distribution of the income of
landlords,we concentrateour attention instead on the impact of control on the
tenants'welfare and income distribution. As it turns out, the net losses to
tenants are positively related to household income, suggesting that rent
indexationmay be somewhat progressive.

1.15 Despite the above result indicatingthat the "average" or "typical"

tenant is not greatly affected by controls, such measures of central tendency
mask the fact that controls do have significanteffects on many tenants. For
example,while our best estimate of the median sample benefit is very low (Cr$-
95, a small welfare loss statisticallyindistinguishablefrom zero) within the
sample of controlled renters, a quarter of the sample receives benefits over
Cr$745, and a quarter of the sample experiencesa net welfare loss greater than
Cr$850. Furthermore, there are several distinguishable patterns in the
distributionof net benefits. Among controlledunits, householdsin the highest
income quartile experiencelosses of Cr$1,136while those in the lower quartile
actually gain Cr$154. These results from the Olsen model are very similar to
cross-countrymodel estimatesof a Cr$43 gain in the lower quartile and Cr$1,227
loss in the upper quartile. The cross-countrymodel also estimateslosses in the
upper quartile of uncontrolledunits to be forty times higher (Cr$2,416)than
those in the lower quartile (Cr$60).

1.16 More important to the welfare of poor tenants than the impact of rent
indexation on income distribution is its welfare impact relative to income
levels. As it turns out, rent indexationis also particularlyhostile to tenant
households in the highest quartile. Among controlledhouseholds, the welfare
benefits of the poor representapproximately2 percent of income relative to an
equal loss of 2 percent for householdsin the highest quartile. For uncontrolled
 -6 -

units, the cross-countrymodel estimates that losses to the poor amount to less
than 1 percent of income while the rich lose over 5 percent.

1.17 Finally,tenantsexperiencelosses in both formal and informalmarkets.

However, while low-incomehouseholds generally benefit from rent indexation,
those living in informalhousing do not. Median tenant losses in the formal
sector are less than one-third of informal sector losses. Part of the welfare
losses in the informalsector is due to the underconsumptionof housing and part
to the overpricingof housing services. Whereas rent indexationenables tenants
in the formal sector to pay rents which are 9 percent below market, those in the
informalsector pay 8.5 percent more than what the unit is worth. It is little
wonder that slumlords registera welfare gain from rent indexationwhile formal
housing landlordsregistera loss. As a share of income, informalsector tenants
are slightly more penalized too. Whereas tenants' losses in the formal sector
represent2 percentof median income,losses in the informalsector are 3 percent
of income.

Rent Control. Decontrol and the Housing SuDRlv

What are the dynamic costs of rent control? Are there

measurable losses to tenants from reduced mobility and to
landlords from inability to convert units to their highest
and best use? What are the effects of rent control on the
profitabilityof rental housing? What are the implications
for housing supply? Many alternativesfor change present
themselves. What can we infer about the effects of
different changes on profitability and supply? On
affordability?What other constraintsmust be addressedfor
the housing market to respond to changes with increased
output rather than increasedprices?

1.18 Brazil's current system of rent indexationdoes not depress the rate
of return to rental housing by as much as other countries who have more
restrictivesystems. However,the system does increaserisk and, hence, discount
rates, as well as the thresholdrate requiredfor investment. Historically,some
correlationbetween strength of regime and housing supply is apparent, but it
requires further work to establish causality,if any exists.

1.19 Simple simulations demonstrate that, under Brazil's current regime,

regulations increasing tenure security beyond prevailing lease periods and/or
restrictionson conversionto alternateuses can reduce profitabilityof rental
housing more than rent indexationalone.

1.20 One of the lessons to be gained from the observation of short-run

movements in rental prices, before and after changes in the rent control
legislation, is that the phasing out of rent controls should be gradual. The
main justificationfor a gradual approach is that it minimizes the impact on
householdbudgets. Any sudden increasein rents (such as what took place in 1950
immediatelyafter Law No.1300 was enacted) is not likely to contributemuch to
the supply of rental housing in the short run, but would have a devastating
impact on low-incomehouseholds. Such a change could simply lead to another
imposition of controls. On the other hand, gradual changes (such as the ones
that finally eliminated nearly all rent controls between 1964-77) have taken
place without real rent increases.

1.21 Furtherjustificationfor the gradualapproachresides in the fact that

Brazilian institutionsare unprepared to deal with the displacementthat would
result from sudden decontrol. There is no welfare system to provide temporary
housing or supplementaryincome to those negatively affectedby decontrol,and
there is no residual space in the courts to deal with the increase in case load
that might be created by decontrol.

Future Work

What additionalwork is required to analyze controls?

1.22 Most theoretical research to date has focused on the analysis of a

simple price control without much reference to common methods of adjustmentof
the controlledprice. Two exceptions(Arnottand Rydell et al.) suggestthat the
effects of different stylized laws can vary drastically. The little empirical
research which has been done is inconclusivein regards to the effects of rent
control on investment. Research on benefits to tenants suggests that when rent
control yields benefits, they are very small when compared to costs. There is
no evidencethat rent controloperatesas an effectiveredistributivedevice, and
rent control seems to decrease household mobility.

1.23 The impact of rent control on governmentrevenues is another under-

researchedarea.a' Little is noted in the literatureabout whether rent control
can depress income tax collections in countries, such as Brazil, that collect
taxes on rental income. Also neglected is the impact of rent control on the
finances of local governments. In Brazil, property taxes provide one-third of
municipal taxes and 8 percent of revenue. In most other developing countries,
the share is even higher, averaging17 percent of recurrentreceipts in a sample
of nineteen countries.A' Rent control affects property taxes in several ways;
the nature and magnitude of the effect depends on whether the property is taxed
based on its capital value or on the stream of rents it generates.

1.24 Some countries, such as Brazil, base property taxes on the assessed
value of the property. Thus, rent controls lower taxes by decreasingthe value
of the underlying asset. The extent of the reduction depends not only on the
reduction in current and future rents, but on regulationsaffecting the security
of tenants and the conversionof property to other uses. From the point of view
of investors, the decreased rent will be partly offset by decreased property
taxes and maintenancecosts, but aggravatedby increaseddepreciationfollowing

3. Malpezzi and Tewari (1990) demonstrate that rent control supresses

collections in India, and that the system used there breaks the links between
taxes and property values, and income.

i/ For a general discussion of the property tax, see Dillinger (1988).

the reducedmaintenance. All of these effects should be netted out in examining
the effect of controls on taxes.

1.25 Obviously the effects of controls depend on the nature and impact of
controls on rents. In particular, it is important to establish whether the
control,of whatever type, results in a decreaseroughly proportionalto market
value, market rent or income (i.e., to some measure of ability to pay taxes); or
whether the decreases are either random or mainly correlatedwith factors like
the age of the structure,which are themselvespoor measures of ability to pay.
If controls merely shift the basis for assessmentby some constant amount, tax
revenue can be recovered without doing violence to equity or efficiency by
appropriateadjustmentsin the tax rate. If controlsshift rents stochastically,
then taxes themselvesare stochastic.

1.26 In general, the review of previous research demonstrates other

importantpracticalgaps in our knowledgeof rent control. Little is known about
the extent of rent control generally and even less about the prevalence of
specific provisions of rent control laws, how they are enforced, what related
laws exist, and what market conditions are like in controlled markets. Few
econometric studies have been done anywhere on the costs and benefits of rent
control laws, and only two in developingcountries. No econometricstudieshave
been done in Brazil, despite the long history of controls there. This paper is
designed to fill some of these gaps. Yet, at the end of this study we are still
left substantivelyignorantof many aspectsof Brazilianrent controland housing
markets. The main areas of fertile future researchare in the modeling of rent
control's impact on governmentrevenues,demand and supply relationsover the mid
and long term, and land markets. Expandingthe present study to other cities and
to 1990 would be a first step towards increasingour knowledge of the Brazilian
rental market at low cost.
 - 9 -

II. THE EVOLUTION OF RENT LEGISLATIONIN BRAZIL

2.1 The purpose of this chapter is to review the legislativeframeworkof

rent control in Brazil from the early years of the republic to the present.
Since it is not possible to address rent controlswithout a generalunderstanding
of the broader issue of housing, the chapter was structured to include a brief
survey of housing institutionsin Annex A.

A. Early Legislation

Pre-Rent Control Regulationof the Rental Market

2.2 Before rent control laws were enacted,the interactionbetween tenants

and landlords was regulated by a civil code which emphasized the concepts of
contractualfreedom and absoluteownershipof property. Spelledout in the civil
code, we find the principlesof free determinationof rent (Article1188) and of
duration of contracts (Article1200). Rental properties could be recoveredby
the owner at the conclusionof the lease (Articles1129 and 1194) or after a one-
month notification to tenants (Article1209).

2.3 Rental regulation by civil code lasted only four years, from the
creation of the code in 1917 until the first specificrent legislationin 1921.

First Rent Control Laws

2.4 The first rent control Law (No. 4403 of December 1921) was enacted to
reduce the inflationarypressureswhich originatedduring World War I. The law
received widespread acceptance since the rental market was perceived as being
highly concentratedin the hands of a few landlordsand, thus, not subject to the
benefitsof perfect competition. The first set of controlslasted four years and
was relativelymild since free negotiationon initial rental price was preserved
(Article 10). The actual "control" came in the form of automatic extension
clauses on contracts,which were activatedif the landlordfailed to request the
unit at least six months prior to contractualexpiration(Articles4 and 5). At
the same time, rents could only be adjusted after two years of previous
notificationto the tenant (Article10). This rent legislationalso sought to
provide significant protection to tenants in other important areas. The
situationsunder which a tenant couldbe evicted,for instance,were specifically
detailed and basically restrictedto failure to make payments (Article6, para.
1), damage to the property (Article6, para. 2), and requests by the owner to
recover the property for personal use (Article11).

2.5 Parts of the rent control law were extended by successive laws
numbered: 4624 (Article1) of 1922, 4793 (Article18) of 1924, 4975 (Article1)
of 1925, and 5177 (Article1) of 1927, and Decree 4840 (Article1) of 1924. It
was not until 1928 that the regulation of rental agreements returned to the
jurisdictionof the civil code, where it would remain for the next fifteenyears.
 - 10 -

2.6 The BNH-CENPHAdocumentation(BNH, 1971) stresses that World War I was

the catalytic event leading to the first rent control legislation. however,
anecdotal evidence on the workings of political movements at the turn of the
century indicatesthat internalpressures,which had been building up before the
war, might have played at least as critical a role in the establishmentof the
first set of rent controls. Already in 1907, tenants in Rio had organized
themselvesto strike against rising rental prices. Between 1912-14, tenants in
Sao Paulo also became politically organized. The movement spread to several
districts of the city and culminatedwith a large demonstrationin the center of
town and the subsequentdeportationof immigrantanarchist leaders suspectedof
being involvedin the movement (Rolnik,1983). Landlordsalso became politically
organized and finally succeeded in revoking the law (Ribeiro, 1985).

Impact of the First Rent Control Laws

2.7 Evidence from official

estimates of rental prices suggests Table 2.1: INDEX OF RENTAL PRICES
that the controls imposed by Law No. IN RIO DE JAEIRO: 1912-30
4403 and subsequent laws were largely (1929 - 100)
ineffective in protecting the average
renter from increases in rents. In Rental Prices
fact, real rents increased faster Year_______l_Real
during the controlled period than they 1912 32.8 87.5
did in the other years (see Table 19143 32.8 85.8
2.1). Between 1912-29, rents in the 1915 34.4 82.8
city of Rio rose at about 7 percent 1916 34.4 77.2
p.a., or about 1 or
p.a., percent p.a. in real 1917
~~~~~ 36.1
39.3 73.5
~~~~~~~~~1918
71.5
terms. When controlswere introduced, 1919 42.6 74.9
real rents increased rapidly (at 4.2 1920 49.2 78.6
percent p.a. between 1921-27). 1921 49.2 76.3
percent
p.a. 1~~~~~~~~~ 57.4 ~~~922
81.5
1923 65.6 84.6
2.8 We have no reliable 1924 82.0 90.4
information regarding the degree to 1925 90.2 92.9
1926 100.0 100.4
which rent control laws were enforced 1927 100.0 97.8
during this period, and little data 1928 100.0 99.3
from which to infer the reaction of 1929 100.0 100.0
landlords to this government 1930_90_2_99_1
intervention in the market. We soe: IBGE, 1987.
suspect, however, that the passage of
this unprecedented law led many
landlords to perceive that a large risk component was being introducedto the
returns on their properties and to exit the rental housing market. Thus, the
share of rental housing in total housing declined by about one-third between
1920-40,despite the scarcityof long-termcredit to facilitatethis substantial
tenure transition.
 - 11 -

B. PopulistRegulations

Rigid Rent Control under Law-Decree4598 and Extensions

2.9 Sharp increases in nominal and real rents, coupled with a decline in
real wages, led the governmentto intervenein the rental market at the outbreak
of World War II.51

2.10 The war environment of scarcity and widespread rationing facilitated

the creation of a rigid rent control law (DL 4598) in 1942. Rental prices on
residentialhousing were to be frozen for a period of two years at values charged
on the last day of December 1941 (Article 1), with the tenants now being
responsible to pay all taxes incurred on the unit, including property rates
(Article 2).

2.11 All eviction actions were suspended,and evictionrequestswere to be

accepted only if provisionsof contractualagreementswere broken by the tenant,
or if the landlordrequestedthe unit for either personaluse or "urgent reforms"
(Article4).

2.12 In 1943, the rent legislationwas extended to commercial units (DL

5169), while the control on residential units was slightly relaxed to allow
landlords to request their propertiesfor the use of direct relatives. In 1944,
one month before its expiration, the rent control law was extended for another
year with the enactment of DL 6739. Rental prices for residentialunits were to
be kept frozen, although commercialunits were allowed a 10 percent adjustment
(Article 1, para. 1) -- slightly below inflation. As an incentive to new
investment in rental housing, Article 12 of DL 6739 established that rent on
units constructedafter 26 July 1944 was to be determinedby market forces.

2.13 After some minor adjustmentsin the rent control legislation(DL 7339,
and DL 7762 of 1945), a more detailed law, DL 9699, was enacted. Except for a
drastic revision which eliminated the provision that exempted post-July 1944
constructionsfrom rent control, the new law preserved the basic spirit of its
predecessors:control of rental prices (Article4), restrictionson the duration
of the contract (Article20), and limits to the reasons for evictions (Article
18). Prices in old residentialcontracts were raised by 15 percent (over their
value in December 1941!) and commercialcontracts were adjusted by 25 percent.
Contracts were regulated by DL 9699 until December 1950 when a new law was
enacted by Congress.

New Contracts ExemptionUnder Law No. 1300 and Extensions

2.14 The new law (No. 1300) did not affect old contracts,which continued
to be regulated by the provisions of DL 9699, but allowed for rental prices of
new contracts to be freely negotiated (Article 3) and indexed. Although
stipulatedto remain in force for a period of two years, the new law lasted, in
fact, over thirteen years through a series of "extensionlaws" numbered: 1462

J/ Our evidencehere is mostly anecdotalsince there are no rental prices series

for the period 1940-47.
 - 12 -

(October,1951), 1708 (October1952), 2328 (November1954), 2620 (October1955),

2699 (December1955), 3085 (December1956), 3336 (December1957), 3494 (December
1958), 3844 (December 1960), 3912 (July 1961), 4160 (December1962), 4240 (June
1963), 4292 (December1963), 4346 (June 1964), and 4416 (September1964). Only
one of these laws (No. 4240, Article 4) authorizeda rental price increasebased
on the age of the contract.

2.15 We have no specific statisticson rental housing constructionduring

the intervals marked by major changes in rent legislation. However, the fact
that about half of the housing market consistedof rentals and that requestsfor
building licenses and overall area licenses (residential and business
construction)increased substantiallyin the early 1950s suggests that Law No.
1300 might have had a positive impact on overall housing construction.

2.16 Between 1935-63, the index on constructionstarts (licensesto build)

registered an increase of 2.5 percent p.a. (in Rio), with the area under
construction expanding even faster, at 5.3 percent p.a., as high-density,
multistory buildings began to dominate the landscape. Between 1949-52,
concurrentwith the enactment of Law No. 1300, the area of housing starts grew
at an unprecedented33 percent p.a., followedby a sharp decline of 8.4 percent
p.a. in the next four years.
Figure 2.1: LICENSES TO BtUILD AND AREA LICENSED Figure 2.2: CONSTRUCTION COSTS

tl d. $ Iui|si

lo1ssIf Is 1 14S.4?I44S 1a £1 £ so 14 is to £7 It I 644

s S ES It 14441 4 41 41 41 If B1St £4 5£ LI
SI I1 IS *1 s SI
e I$ 64

Year Year

e misbtof Liagee, -4- Artea Ultd

Li l Ceastlslt. Clet -+- Nemlas Camel,.Cell

ImRact of Populist Rent Regulations

2.17 Rent control laws in the period 1942-64 were initially designed to
remedy the short run imbalances brought about by the war. However, as the
legislationwas ritually extended, it became harder to repeal it, since tenants
began to consider the continuousoccupationof their rental units as a right (to
shelter) often voiced by populist leaders in government.
 - 13 -

2.18 Landlords adapted to the rent control law by totallyneglectingtheir

propertiesand then performingmajor reconstructions(inwhich case tenantscould
be evicted). Real estate entrepreneurs of those times also report on the
widespread use of "creative sale agreements" as a substitute for rental
contracts. This practice involved selling a rental unit to the prospective
tenant for a price substantiallyabove market, without a down payment and with
the landlord carrying the loan and amortizingit in such a way that the monthly
payments of interestand principal would correspondto a perceivedmarket rent.
The buyer was supposed to make a balloon payment at the end of the year for the
balance of the loan. However,since the buyer was invariablyunable to meet this
last payment, the property then reverted to the original owner.

Table 2.2: INDEX OF RENTAL PRICES IN RIO DE JANEIRO: 1930-64

(IBRE: 1939 - 100 for period 1930-39)
(Xingston: 1948 - 100 for period 1948-60)
(IBRE - 100 for period 1953-64)

Nominal Real
Rental Prlces Rental Prices
Year IBRE Kingston IBRE Kingston

1930 84.6 121.9

1931 76.9 211.1
1932 70.8 105.4
1933 70.8 106.3
1934 76.9 107.3
1935 76.9 101.6
1936 92.3 106.3
1937 95.4 102.1
1938 97.5 100.1
1939 100.0 100.0
1940-1947 - - - -
1948 100.0 100.0
1949 106.0 100.0
1950 110.0 93.4
1951 224.0 171.6
1952 247.0 156.6
1953 100.0 261.0 100.0 141.8
1954 114.9 276.0 91.0 118.7
1955 130.9 310.0 87.1 112.0
1956 155.8 406.0 85.2 120.5
1957 181.3 548.0 88.1 144.6
1958 216.8 658.0 89.2 147.0
1959 261.9 804.0 70.8 118.1
1960 307.4 950.0 67.1 112.7
1961 359.1 54.8
1962 421.6 41.5
1963 783.2 42.7
1964 1034.7 30.2

Source: IBGE, 1987.

RevLsta Brasil.ira do Economia, 1960.
 - 14 -

2.19 Estimates of rental values shown on Table 2.2 indicate that nominal
rents in Rio (Guanabara) increased by less than 1 percent p.a. in the 1930s,
representingabout a 20 percent reductionin real value over the decade. We have
no data on rents for the period 1940-47,but between 1948-53,real rental prices
increasedby an average of 7.2 percent p.a. (Kingston,1960) and then declined
until 1964 at 10.3 percent p.a. according to IBRE's statistics. The immediate
impact of Law No. 1300, which freed rents on new leases, was appreciable. Real
rents increasedby 84 percentwithin one year after the law became effective,but
then declined almost continuouslyuntil the end of the decade. By 1960, rents
had already lost nearly one-third of their purchasingpower relative to 1951.

2.20 Despite all the maneuvers by landlords to circumvent the rent

legislation,and despite the provisionsof Law No. 1300 (Article3) which relaxed
the controlson rental prices of new contracts,it is still apparent that, in the
long run, the controls managed to discourage investments in rental housing.
During the populist period, the share of rented dwellings in total housing
declined from 70 percent in the 1920s, to 49 percent in 1940, to about 47 percent
in 1950, and finally 43 percent in 1960. Thus, the availabilityof the only type
of formal housing affordableto the droves of migrantsenteringthe larger cities
in the Southeasternstates was effectivelyreduced,a situationwhich encouraged
an unprecedentedboom in the informalhousing sector.

C. Gradual Decontrol and Indexationof Rent

2.21 In November 1964, Law No. 4494 was enacted to remove the freeze on
residential and commercial rents and to provide a stable environment in the
rental housing market. The law was intendedto achieve an orderly approximation
of market rental prices, an approximationconstrainedonly by a level of social
conflict deemed bearable by government. New leases would be freely negotiated
and then indexedto variationsin the minimum wage, minus a factor reflectingthe
depreciation of the property (assumed to occur at between 2.3 and 3 percent
p.a.). Old leases would be based on the rent originallyagreed to by the parties
concerned,and then updated (bymonetarycorrection,minus a depreciationfactor)
over a period of 10 years, in order to prevent sudden rent increases.

2.22 Rent controlwas progressivelyundermineduntil October 1967,when all

properties, except those leased before 7 April 1967 and constructedprior to 30
November 1965, were priced according to free market rules. In November1965, Law
No. 4864, Article 28, removed the regulation of commercial property and
residential units in new buildings from the scope of Law No. 4494 and
incorporatedit into the civil code. Two years later, DL 322 (substitutedlater
by Law No. 5334) transferredall new leases to the jurisdiction of the civil
code.

Impact of Decontrol

2.23 Between 1967-75, an increasingnumber of propertiesbecame regulated

by the civil code. Interestingly enough, the return to near free-market
conditions did not imply a rise in real rents. A study of the rental market in
Rio by CENPHA/BNHin 1974 indicates that, although rents rose more rapidly in
controlled premises than in units under the auspices of Laws No. 4864 and No.
 - 15 -

5334 (comprising73 percent of those surveyed),real rents actually declinedan

average of about 20 percent over the period 1966-74. Changes in real rents
ranged from -28.4 percent for one-bedroomapartmentsin middle-incomeareas to
0.6 percent for two-bedroomapartmentsin upper-incomeareas; while rents in low-
income areas fell by an average of 24 percent (see Table 2.3).

Table 2.3: INDEX OF REAL RENTAL PRICES IN RIO DE JANEIRO:

1966-74
(1966 = 100)

figh-Income Areas Low-Income Areas

Year One Bedroom Two Bedroom One Bedroom Two Bedroom

1966 100.0 100.0 100.0 100.0

1967 90.2 98.1 87.9 91.4
1968 90.6 100.6 83.2 88.7
1969 90.2 99.7 87.9 90.1
1970 83.9 90.0 81.3 80.1
1971 74.6 80.3 79.4 74.2
1972 71.9 79.1 73.8 72.2
1973 79.5 94.4 77.6 76.8
1974 79.9 100.6 75.7 76.2

Source: BNE-CENPHA, 1974.

2.24 That real rents did not increase while rent controls were being
removed is partly explained by the fact that rents for most units probably
reflectedmarket values before 1967 anyway. Since approximately40 percent of
all leases in Rio were less than two years old (accordingto a 1974 survey), a
substantialnumber of premisesmust have had recent leases in 1967 and thus were
subject to market rents (by Law No. 1300) sometime in the near past.
Furthermore,the bulk of rental agreementswas reached outside the parametersof
the rent control legislation. According to the 1974 study by CENPHA/BNH, as'
quoted in a United Nations' review of rent control in developingcountries:

"..ninety-seven percent of all tenants in Rio de Janeiro

during the years 1964-74 paid rents that were agreed upon
with landlordswithout regard to the rent legislation.111

This is, of course, an exaggeratedstatement:even though the readjustmentsof

rents might usuallybe implementedabove or below the rate prescribedby law, the
index specified in the rent control legislationis always the startingpoint of
any negotiationbetween the rental parties.

§/ United Nations, Review of Rent Control in DeveloRing Countries (Departmert

of Economics and Social Affairs, ST/ESA/85,1979), p.4 7 .
 - 16 -

D. Return of Rent Controls

2.25 Between 1975-80, the rental market slowly fell back into a controlled
regime. In 1977, DL 1534 established that expiring rental contracts must be
extended by a minimum of six months and a maximum of twenty-fourmonths on the
basis of two months p.a. of residence in the premises. The law also required
that during the extension period rents be adjusted by the variations in the
nominal value of "IndexedTreasury Notes', (ORTNs).

2.26 A more drastic revisionof the rent control legislationwas introduced

in 1979 by Law No. 6649, and later by Law No. 6698, which revoked all previous
laws on residential rentals and extinguished the denuncia vazia (allowing
landlords now to request the return of their rental property only with "due
cause', as specifiedby Law No. 4468 of 1964). The new Law No. 6698, Article 49,
para. 2, maintained the 1977 provision whereby rents could not be adjusted by
more than the variation in the ORTN. The greatest relief to landlords came in
the form of a provision (Article 49, paragraphs 4 and 5) that allowed the
proprietors of residentialunits to petition for a judicial review of the rent
to bring it into conformitywith full market value. This could be done five
years after the date of the last written agreement or from the date of the
signature of the original contract, whichever occurred later. This provision
remains in force to this date, but its usage is conspicuouslylimited due to the
slownesswith which the process transitsthough the court system. Nevertheless,
the fact that landlords can periodicallyadjust rents back to market values has
interestingtheoreticalimplicationswhich we explore in Box 2.1.

Impact of Re-TighteningRent Controls

2.27 Over the period 1980-85, most Figure 2.3: LICENSES TO BUILD, AM
changes in the rent control legislation LICENSED MM CONSTRUCTION COSTS
were related to the correction index and us'
the manner through which the index could 4--
be utilized for rent adjustments. Law
No. 7069 of 1982 in its Article 1, for ---
a@|- X . .. ..
instance, specifies that the increase in ... .......
rents could not be above 80 percent of
the variation in the 'National Consumer '"
Price Index' (INPC). Between 1982-86,
the increase in rents continued to be tSS
restricted to 80 percent of the INPC by Year
successive laws extending Law No. 7069: Uts -+Au,t l^ti414S1.. C4.tUtI.
DL 2069 of 1983 (Article 1) and Law No.
7335 of 1985 (Article 1).

2.28 Table 2.5 shows the evolutionof nominal and real (offer)rents in Rio
over the period 1975-85. As we have already observed, in the rental time series
of 1966-74, the highest rates of growth in rental prices are registeredfor the
higher-priced units. One- and two-bedroomapartmentsin higher-incomeareas of
Rio experienced nominal increases of 85 and 90 percent p.a., respectively,
whereas in the low-income areas (suburbs), the increases were of 79 and 85
percent, respectively. Over the same period, prices were rising at an average
118 percent p.a. Real rents increasedbetween 1975-77, and then began to fall,
 - 17 -

BOX 2.1: MODELLING LANDLORDPRICING BEHAVIOR UNDER LAW NO.6698

If financial markets functioned properly, we should expect landlordsto set initial rents that reflect
the forecasted inflationof the next five-year period (capitalizing future period losses). The main
difference between the operation of such a market and a free market would be a net welfare loss to socLety
consistLng of a portion of the risk factor (substantial in a country with high variance in inflation
rates). RestrLetions in the types of legal rental contractsavailableto economic agents force them to
take risk and time discount positions that are unnatural or less preferred,and for which they are
unable to recover fully through adjustmentsin the paths of price and quantity of housing contracted.
Moreover,the inabilityof most tenantsto borrow in the formal market effectively constrain the Initial
rents that landlords can set so that even this seemingly laxed rent control legislation (under Law
No.6698) is expected to interfere significantly with the functioning of the rental housing market.

We have worked out an example of landlord pricing behavior under this type of rent control:
ASSUMPTIONS
I. General
1. inflationat tO and t-l (PPO, PP-1) - 100 percent;
2. expected inflation: E(PP) - aPPO + bPP-1 a + b - 1;
3. average duration of a lease - one year;
4. equilibrium real rate of return (rent component of total return to housing
investment.no change in real housing prices assumed) = 15 percent p.a.;
5. average price of a housing unit at tO - $8,000;
6. negligibleoperatingcosts.
7. perfectly functioning financial markets
8. no risk premium

I1. No Rent Control Resime (NRC}

real rent = $1,200 p.a.

Rent Control Rexime (RC)

Adjustment at end of year:
- RCa) 90 percent of inflation;
- RCb) 50 percent of inflation;
- RCc) no adjustment (price freeze).

PATH OF YERLY MTS OVERFM YEARS

Cs)

FIRST-YEAR
REGIME YEAR 1 YEAR 2 YEAR 3 YEAR 4 YEAR 5 MARK-UP
(percent)

(NRC) 1,200 2,400 4,800 9,600 19,200 -

(RCa) 1,399 2,658 5,050 9,596 18,232 17
(RCb) 2,654 3,981 5,972 8,957 13,436 121
(RCc) 7,221 7,221 7,221 7,221 7,221 502

Present Value of NRC path at 15 percent discount rate = $24,205.

RCa: let RO + (1.9RO11.15)+ + [(1.9)ex4]R0(1.15)ex4- $24,205

RD - $1,399

RCb: let R0 + (1.5R011.15) + + [(1.5)ex4]R01(1.15)ex4 - $24,205

RO - $2,654

RCc: let RO + ... + R0/(1.15)ex4 - $24,205

RO - $7,221

For the NRC case, the landlord, at each year, sets rent such that average real return is US$1,200
(15 percent of US$8,000). For the RC cases, rent is set such that the present value of the stream equals
that which vas derived for the NRC case (we did not includea risk premium in our calculations).
 - 18 -

accelerating the rate of decline during the recession years. In the ten-year
span, the real value of rentals in high-income areas showed average annual
declinesof 7 and 4 percent for one and two bedrooms, respectively,while in the
low-income areas, real rents declined by 10 and 7 percent in one and two
bedrooms, respectively.

2.29 In 1986, all rents were frozen under the "CruzadoEconomicPlan." The
Plan did not change the rent regulations stipulated in Law No. 6649.
Nevertheless,the structure of rent adjustmentswas altered, insofar as rents
were included in the generalizedprice freeze mandated by the Plan. Laws No.
2283 and 2284 of 1986 establishedcontractual rules for rental adjustments in
residentialand commercialunits: nominal values of residentialrental contracts
were correctedon the basis of the last adjustmentto the average of the previous
twelve months and commercial contracts on the peak value of the same period.
Then, they were kept constant for a period of one year until 28 February 1987;
and all eviction actions were suspended until the next revision in the
legislation,scheduledfor March 1987. The value of new contractswas determined
by the market and also frozen until the end of February. Rents were adjusted in
March and a new freeze for a ninety-dayperiod was institutedby DL 2335 in June.

Table 2.4: INDEX OF RENTAL OFFER PRICES IN RIO DE JANEIRO, 1975-86

(1975 = 100)

High-Income Area Low-Income Areas

One-bedroom Two-bedroom One-Bedroom Two-bedroom
Year Nominal Real Nominal Real Nominal Real Nominal Real

1975 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

1976 186.7 132.0 190.9 135.5 187.5 132.5 322.2 228.9
1977 260.0 128.7 277.3 137.3 287.5 287.5 400.0 197.8
1978 346.7 124.0 359.1 128.2 412.5 178.9 500.0
1979 473.3 110.0 522.7 121.4 550.0 127.5 688.9 160.0
1980 - - - - - - - -
1981 1,486.7 82.0 1,359.5 75.0 1,750.0 96.3 2,155.6 118.9
1982 2,333.3 66.0 2,363.6 66.8 2,900.0 82.5 3,889.9 110.0
1983 6,669.3 74.0 6,381.8 70.9 5,050.0 56.3 7,500.0 83.3
1984 1,406.7 50.7 13,990.9 48.6 12,275.0 42.5 19,866.7 68.9
1985 46,493.3 49.3 61,545.5 65.5 32,937.5 35.0 47,777.8 51.1
1986* 272,133.3 119.3 240,809.1 105.9 90,500.0 40.0 224,222.2 98.9

SOURCE: BNH - CEMPEA, Various years.

*until AugU3t.

2.30 As a result of the unanticipatedinitial rent freeze and of the lack

of trust by landlords on the commitment of the government to allow rents to
adjust after the one-year freeze or of its ability to maintain inflation at low
levels, rents on new leases rose sharply. The increasewas much higher than what
was perceivedto be requiredto compensatefor the forecastedinflationuntil the
end of the freeze. Within the first six months of the freeze, new rents
increased by 175 percent for one-bedroom, low-income apartments, and by
291 percent for the more expensive two-bedroomunits in high-income areas.
 - 19 -

Adiusting to Hyperinflationunder Rent Control

2.31 In every period in the

history of Brazilian rent control, a Box 2.2: PULL CIRCLE TO
series of informal arrangements has FREE RENTALMARKETS
developed between the landlord trying Regina Gargano, a metallurgical worker, scavenged
to defend his income--and, indeed, the the city of Sao Paulo ln search of an apartment
value of his asset--and the desperate for six months without luck. At the end, she had
to settle for acme space in a partially occupied
tenant trying to find a place to live. house. And she only got the place because she was
The stronger the enforcement of the wiLILng to slgn a special borrowing agreement. To
law, or the more sudden and extreme close the deal Regina was forced to incur great
financial risk by signing twelve promissory notes
the real or potential loss was to the corresponding to twelve months of rent. *The
landlord, the more informal was the worst part of it is that I can be put on the
arrangement. streets at any mont' complains Regina.
Soure: Veja, March 7, 1990, pp.78-80.
2.32 At the end of the New
Republicperiod, inflationhad reached
over 1,000 percent p.a. and despite
adjustmentsin the law that allowed for rent realignmentsevery four months, the
real value of rents fell quickly. A frequent arrangement to cope with
residential rent control under rapid inflation involved a tenant entering into
a commercial contract (which can be adjusted monthly) in order to rent a
residentialunit. As uncertaintyabout commercialcontracts rose and as monthly
inflationneared triple digits, preferencesbegan to shift towards less explicit
'understandings"between parties. One such deal has the appearanceof a simple
and friendly temporary property loan. It consists of the tenant signing a
document stating that he is living on the premises for free and that he will
promptly vacate it when requested to do so.

2.33 We have come full circle to the civil code regulation of rental
contracts,before the early days of tenant protectionwhen all the law tried to
provide was some stability to tenants by stipulating conditions under which
evictioncould take place. Despite its good intentions,the present rent control
law places tenants under more uncertainty than ever.
 - 20 -

III. COSTS AND BENEFITS OF RENT CONTROL: THEORY AND METHOD

3.1 This chapter outlinesthe economictheory and analysis of rent control

to provide a perspectiveand prelude to the empirical results of the effects of
rent control in Rio, which are reported in the next chapter.

A. Analysis of Rent Control as a Tax on Housing

3.2 Simple rent control,which ignoresdynamicprice adjustmentmechanisms,

is usually viewed as a tax on the return to housing capital. Such a simple model
of rent control as price control,where the price per unit of housing charged by
landlords is reduced by administrativefiat, is depicted in Figure 3.1. Rent
control is represented as a move from price (rent) PO to Pl. If, rather than
being reduced, rents are frozen at existing levels or are allowed to rise at
administratively-set, low rates; then rapid price inflation,as has occurred in
Brazil, leads to a similar divergencebetween equilibrium (P0) and controlled
prices.

3.3 In the short run the housing FiRure 3.1: RENT CONTROLWITH
stock is fixed at QO, i.e., the supply ELASTIC SUPPLY
of housing S is perfectly inelastic; a
but at P1 there now exists excess
demand (Q1-QO). Previously the
available units only went to buyers
who valued them at P0 or above. But
now price has been reduced to P1, and
demand has risen to Ql. Demand rl U
exceeds supply.

3.4 The divergence between P0

and P1 provides a strong incentivefor D
the development of anticipated
inflation premia (AIP), advance Qo 0I
payment and key money systems.

3.5 In the longer run, the supply schedulehas more elasticity (S1), and
so if neither AIP, advance payments nor key money have become effective
equilibratingmechanisms (e.g., because low incomes and inadequate capital
markets make it difficult for many renters to finance key money or advance
payments); then landlords simply decrease the quantity of housing services
supplied to Q2. Some units are demolishedearly, others are converted to owner
occupationor transferredto alternativeland-uses,such as commercialpremises,
and some new starts are forgone. In the very long run, with an elastic supply,
the simple competitivemodel implies an unhoused population.

B. Rent Control as ExpenditureControl

3.6 An alternative view of rent control is to model it not as a price

control but as an expenditure control, following Olsen (1969) and Frankena
(1975).
 - 21 -

3.7 In this model, rent control initiallylowers real rents to P1 (Figure

3.2) from P0. Rent is fixed at PlQO. While supply is inelasticin the short run
at QO; in the intermediaterun, landlordshave some latitudeto vary the quantity
of housing services availablein the market as representedby the supply curve.

3.8 Real world rent control regimes fix rents, not the price per unit of
housing services. Specifically,for rental expenditurefixed at PlQO, landlords
are constrainedby the rectangularhyperbola E, the locus of all combinationsof
housing services and prices yielding expenditurenumerically equal to PlQO. To
prevent reductionsin unit housing servicesand quality, continual reassessment
of controlled rents in light of the landlord's maintenance policy would be
required to deter the impliedprice per unit of housing service from rising above
P1. Downward filtering could be prevented if rent control was reinforcedwith
the introductionof heavy fines in the event of deteriorationof the landlord's
housing stock, but this is a judgment fraught with evidentiaryproblems. Such
a strict rent control regime, based upon unit housing services, has not been
adopted in real world housing markets.

3.9 Indeed, the final price per Fisure 3.2: RENT CONTROL AS M(PENDITURE CONTROL
unit of housing services can exceed ,,;,
the original uncontrolled price.
Thus, landlords may reduce the unit
supply of services to Ql during the PI
intermediateperiod, but charge P2Q1.
Again, under this model there is _ v
excess demand for housing. However, PI
under the expenditure control model,
the demand curve is not relevant for
determining the new equilibrium. The _ .,.
demand curve indicateshow much buyers
want, given that they can buy all that E
they can pay for at a fixed price per Qt t O
unit of housing service. Clearly,
this condition is not met under the
expendituremodel of rent control.

3.10 The existence of an alternativehousing market (e.g., owner-occupied

or government-ownedsector) would further complicatethis analysis, in that the
existence of such sectors may limit prices to P0 since, if prices rise further,
householdswill switch sectors.

C. Rent Control. Mobility and Tenure Security

3.11 Two studies,by Clark-Heskin(1982) and Boersch-Supan(1983),examine

the relationshipbetween rent control, tenure discounts and mobility. It has
been well documentedthat even in the absenceof rent control,long-termtenants
pay lower prices for rental housing, ceteris paribus [Follain and Malpezzi
(1979),Malpezzi et al. (1981)]. Landlordsfind it difficult to raise rents for
sitting tenants relative to increases for new tenants. They have an incentive
to keep desirabletenants in place through discounts,and tenants themselvesare
 - 22 -

less likely to move when they receive a good deal. Clark and Heskin (1982)
examined length of tenure discounts and mobility rates, with and without
controls, in southern California in the late 1970s. They desegregated their
resultsby geographiclocation,race and income,among other factors. They found
the following: length of tenure discounts increased under rent control; tenant
mobility decreased,except for the youngestgroup of renters;and the differences
were slightly larger for lower-incomehouseholds.

3.12 Boersch-Supan (1983) provides a detailed microeconomic model of a

housing market which implies that rent control of a particular type--a freeze in
real rents which is lifted whenever a new household moves in--will result in
larger length of tenure discountsand lower mobility rates. Initially,he found
empirical support from Follain and Malpezzi's (1980) estimates of the length of
tenure discountfor renters in thirty-nineU.S. metropolitanareas: the average
annual discountwas .95percent for uncontrolledcities,and slightlyhigher--1.1
percent--in controlledcities. The average length of stay was half a year longer
in the rent controlledmarkets--4.7years versus 5.2 years. But when in later
work (Boersch-Supan1984) he used Malpezzi,Ozanne and Thibodeau'slarger sample
of fifty-ninemetropolitanareas, the statisticaldifferentialdisappeared.

D. The RelationshiRBetween Controlledand UncontrolledMarkets

3.13 Severalpapers have addressedthe potential effects of a price control

on a related, though nominallyuncontrolled,market. Needleman (1965)was among
the first to note the possibility of a price control covering part of a market
price in a related market, though he presentedno formal model. Gould and Henry
(1967) demonstratedthat price controlscan increaseor decrease the price of a
substitute. However, their model cannot be directly applied to the housing
market, because it assumes households consume some of each of the two
substitutes. The most thorough treatment of the possible effects of rent
controlson related,uncontrolledmarkets is the recent paper by Fallis and Smith
(1984).

3.14 Fallis and Smith actually develop two related models: one for rent
control regimes which exempt new units from price controls and one for regimes
with vacancy decontrol provisions. Their short-run models predict that under
most conditionsexcess demand spills over into the uncontrolledmarket; and, in
the short run, drives up the uncontrolled price. In the long run, they
implicitly assume an elastic supply function that implies a reduction in the
quantity of housing services from the controlledsector and an expansion in the
uncontrolledsector, narrowing the wedge between prices.

3.15 They also present an empirical test of the model using data from Los
Angeles (1969-1978). FollowingRosen and Smith (1983), they assume that there
 - 23 -

is a straightforwardrelationshipamong rental rates, R, operatingexpenses, E,

and the vacancy rate, V, estimated as:

Rt - -6.25 + .078 Et + 34.09 (l/Vt)+ 26.49 (l/Vt_)

(3.30) (1.64) (4.12) (3.10)

where dots indicate time derivativesand t-statisticsare in parentheses. Rent

control was introduced in Los Angeles at the end of this period (1978). The
estimates are used to forecast what rents would have been in the absence of
controls,and the forecastcomparedwith rents in the controlledand uncontrolled
sector. After two years, controlledrents had risen by 10 percent Ies than the
forecast, and uncontrolledrents by 22 percent more, confirmingthe hypothesis
that rent control increasesprices in the uncontrolledsector in the short run.

E. Measuring Costs and Benefits of Rent Control:

Cross-CountryModel

3.16 As noted earlier, we will present two alternative approaches to

measuring the static costs and benefits of controls. The first and simplest
static model estimates the difference between actual housing expendituresand
what we would expect them to be in the absence of controls,using a cross-country
model calibratedwith uncontrolledmarkets. The simplemethod does not decompose
expenditure into prices and quantities, but it does have the advantage of
yielding information about the effect of rent regulation on the uncontrolled
market.

3.17 The first, simple approachcan be expressedwithout any jargon with the
following two questions:

(a) What do householdspay for rental housing under controls?

(b) What would they pay in the absence of controls?

The answer to the first questionis directlyobservablefrom a survey. Obtaining

an answer to the second is the tricky part. Malpezzi and Mayo (1985, 1987a,
1987b) have developed a cross-countrymodel of housing demand which can be used
to estimate rents paid by typical households in a city, at a given level of
developmentin the absence of controls.

3.18 These estimates can be used to predict market rents in cities where no
uncontrolledsector exists for comparison;to test for bias and to adjust rents
in the uncontrolledsector if such rents have been affectedby controls, as in
Fallis and Smith's model; and as an independentcheck on other methods.
 - 24 -

3.19 The general model for housing expenditures:

ln R - a + E. (ln y) + bH + cH 2 + U

where:
R is rent;
y is income;
H is household size;
E. is the estimated income elasticity of demand; a, b, and c are
regression coefficients;and
u is an estimated disturbance.

3.20 The model was stratifiedfor Fiture 3.3: RENT TO INCOME RATIOS,
renters and owners. For renters,rent CROSS COURY MODEL
was defined as net rent, exclusive of
separate utility payments. For Roilis .I (Rnotou
owners, rent was defined variously, 001
and in order of availability,as owner
imputations of net rent; hedonic I-n-_s,,,,
estimates of net rent based on
applying renter-based hedonic price 4i.t -- ----- -------
equations to owners' housing
characteristics; or imputed rents
based on applying a fixed amortization
ratio (from 1 percent to 1 1/2 percent _
per month depending on the country) to- _ 4 ,
owners' estimates of housing value. a too
1$o
I o
ti ie100 sos
41ao 400 501 110
While other functional forms were litems Isliti U.S. Delicti
tried, and other demographic variables Malozuiad Mee:ya#11[15
were included in alternative
estimating equations, results from the simple log-linearmodel were found to
provide adequate fits and robust findingsregarding major demand parameters.

3.21 In general, the results were remarkably consistentwith the results

from developedcountries (see Mayo, 1981). The regressionfits were typical for
this type of equation: typical R-squared statisticsare in the 0.1 to 0.3 range
(minimum is 0.06, maximum, 0.57). Fits were similar for owners and renters.

3.22 The median of all renters income elasticitieswas 0.49; developing

country elasticities ranged from 0.31 (Pusan, Korea) to 0.88 (Davao, the
Philippines). Most clustered between 0.4 and 0.6, with estimated U.S.
elasticitieslower than developing country estimates. The median of all point
estimates of owner income elasticitieswas 0.46, with extremes of 0.17 in Cairo
and 1.11 in Santa Ana, El Salvador. The majority of point estimateslie between
0.4 and 0.6. In nine of fourteencases where comparisonwas possible,estimated
developingcountry,owner-incomeelasticitieswere greaterthan those of renters.
Comparing expenditure equations across countries revealed practically no
systematicvariation of income elasticitieswith country or city income level or
populationsize, but considerablevariationin dollar-adjustedintercepts,which
were positivelyrelated to averagecity income. Rent-to-incomeratios therefore
declined systematicallywith income within cities, but increased with income
across cities.
 - 25 -

3.23 These relationshipsare shown graphicallyin Figure 3.3 for renters in

four representative cities. Relationships for owners are similar, although
average rent-to-incomeratios are invariablyhigher at every income level for
owners within given housing markets.

3.24 The relationshipsportrayed in Figure 3.3 are very similar to the

consumption patterns within and across countries documented by Kuznets (see
Kuznets,1961 and otherworks cited therein). Qualitatively,housing consumption
is remarkably smaller at various income levels than are between-country
differences at different average income levels. Malpezzi and Mayo explored
alternativetheoreticalexplanationsfor these results and then tested a series
of long-run, cross-countryhousing expenditure models. The simplest cross-
country model parallels the log-linear, within-country model, but with the
addition of a price term, the relative price of housing which was constructed
using data from Kravis, Heston and Summers (1982).

3.25 Defining R as rent, y as household income and PH as the relativeprice

of housing, the followingmodels for renters and owners in developingcountries
were estimated:

Renters:

ln R - - 5.39 + 1.60 ln y + 0.15 ln pH

(0.18) (0.15)

R2 = 0.90
d.f. - 13

Owners:

ln R - 3.57 + 1.38 ln y + 0.65 ln PH

(0.35) (0.50)

R2 - 0.76
d.f. - 11

where rent and income are city means converted to 1981 U.S. dollarsZ',and pH
price index,with the U.S. relativeprice normalized
is the Kravis-Heston-Summers
at one. Standard errors are in parentheses;R2 is the multiple correlation
coefficient;and d.f. are the numbers of degrees of freedom.

3.26 The implicationsof thesemodels,which were confirmedwith alternative

specifications,are straightforward. In the very long run, housing consumption
is income elastic. Price elasticitiesare smaller in absolutevalue than income
elasticities,althoughconfidenceintervalsare quite wide for the former. Long-

1/ Note that in a log-linearexpenditureequation the coefficientof price is

equal to one plus the price elasticity; thus the price elasticity is the
estimated coefficientminus one, or -0.85 and -0.35 for owners.
 - 26 -

run income elasticities are estimated to be slightly higher for renters than
owners. This means that as cities' economies develop over the very long run,
owner and renter consumption patterns increase at a similar pace, ceteris
paribus. However, because renter price elasticitiesare estimated to be higher
than owner elasticities,the net effect of both incomes and prices rising with
economic developmentis that the owners' consumption increases faster than the
renters' consumptionover most of the range of the data.

3.27 However, the sample used in the cross-countrystudies included both

controlledand uncontrolledmarkets. The impact of rent controlwas tested, and
no precise or robust effect was found in their sample. The cross-countryprice
term, which was (unsurprisingly)lower for the controlledmarkets, seemed to be
picking up most of rent control'smeasuredeffects. But the sample was too small
to be particularly confident about this result, and it is singularly
inappropriateas a maintainedhypothesis for cross-countryestimates of demand
used to evaluate costs and benefits of controls. Therefore,Malpezzi et al.
(1988) developed new estimates using only uncontrolledmarkets.

3.28 The cross-countrymodel was re-estimated segmenting the samples by

controlled and uncontrolled markets.Y1 The model yielded the following
estimates for uncontrolledmarkets:

UncontrolledRenters:

ln R - 4.017 + 1.355 ln y
(1.733) (0.299)

R2 .71
d.f. = 7
3.29 For controlled markets and the original data the model yields the
following:

Controlled Renters:

ln R - 5.934 + 1.709 ln y
(1.412) (0.286)

R2 .85
d.f. 5

3.30 As expected, the point estimate of the elasticity from this

uncontrolledsample is greater than one, although the limiteddegrees of freedom
reduce the precision of the estimates from the original cross-cointrymodel.

3.31 This simple model does not answer some interesting questions. For
example,what effect does rent control have on the rents for particularkinds of

i/ The Kravis-Heston-Summers price term was dropped from the new, smallermodel
for reasons described in Malpezzi et al. (1988) pp. 47-57.
 - 27 -

units? What are the consequentcosts to owners of these units? How much do
households value the reduction in rent from controls? What is the "transfer
efficiency"of controls (i.e., tenant benefitsnet of landlordcosts)? The next
few pages present a more sophisticatedmodel that can answer some of these
questions.

F. Measuring Costs and Benefits of Rent Control: Olsen's Model

3.32 The second method we use is similar to that used by Edgar Olsen (1972)
in his econometric analysis of rent control in New York.2 1 It is assumed that
there is an uncontrolledhousing market as well as a rent controlledmarket. The
quantity of housing services provided by a unit reflects all of the
characteristicsassociatedwith the unit: size, amenities,appearance,location
and physical features. Thus, the rent of any unit reflects all of the
characteristicsassociatedwith housing. Differencesin rent in an uncontrolled
market would thus reflect differences in services associatedwith the good.

3.33 The costs and benefits of rent controlcan be assessedby comparingthe

controlled situation with the uncontrolledsituation. One way of implementing
this with/withoutperspectiveis to estimatehow much controlledunits would rent
for in the absence of controls, and to considerthe differencebetween that rent
and the observed controlled rent as the cost imposed on the landlord and,
conversely, the transfer to the tenant.

3.34 These transfers lead to Figure 3.4: RENT CONROL AD

changes in producers', but more CONSUMER'S SURLUS
importantly, consumers' surpluses,
resulting from the existence of Pilot,PerUmI .. lm.
controls, as can be seen in
Figure 3.4.
A
3.35 With an uncontrolled rent
per unit of housing service, Pm,
households would consume Qm units of
housing service, and pay a rent PmQm.
The immediate effect of rent control
is to reduce rent to PcQm. Thus the "
consumer spends (PmQm - PcQm) more on| _~
non-housing goods. Om .t

3.36 At price Pc the consumer

would demand Qd units of housing
services. However, under real world rent control regimes, landlords have no
incentive to increase the flow of housing services to Qd; and indeed, as
landlords filter housing downwards, tenants are likely to end up consuming Qc
housing units. Householdswill find it more difficult to obtain and move to a
suitable unit, and will systematicallyconsume "off their demand curve." The

.2/ For conveniencewe refer to this as the Olsen model because his 1972 paper
was (to our knowledge) the first published study to analyze rent controls with
such a model.
 - 28 -

tenant receives an implicit subsidy of (Pm-Pc)Qc,whose cost is borne by the

landlord. However, notice that the tenant has also given up the consumer's
surplus equal to ABC. His net gain is the differenceof these two areas.

3.37 With rent control,expenditureon the units is reduced to PcQm. In the

short run, price control has no effect on supply. But it has a profound effect
on the allocationof supply between demanders. Previously,the availableunits
only went to renters who valued them at Pm or above. But now that price has been
reduced to Pc, demand has risen to Qd. Demand exceeds supply and the total
market value of rental units will be less than the value before price control.

3.38 At the end, housing is filteredto Qc and there is a further efficiency

cost since supply will be altered; i.e., there is an additionalloss of producer
surplus (BCD). Thus, the triangleACD is a minimum estimateof the medium-term
welfare cost of price (rent) control to society.

3.39 This geometricexpositionillustratesthe basic method quite well, but

an algebraicgeneralizationis better suited for actuallyestimatingthe size of
welfare gains and losses using a sample. It can be shown that if the price
elasticityof demand is constant, the benefit of a program which changes prices
and quantitiescan be written as:

[1lIBenefit - (
(Q,
)bb(-)[--.
ll
b+1 )|Qcb
b+]
Qmb|
+ P.Qm - CQc

where:

Benefit = cash equivalentvalue, a measure of change in consumer's

surplus

Qm = predicted housing consumptionin the absence of rent

controls

Qc - housing consumptionunder rent controls

PmQm- estimated rent in the absence of controls, also

denoted Rm

PcQc - observed controlledrent, also denoted Rc

b - price elasticity of demand.

3.40 In the special case where the price elasticityof demand, b, is equal
to -1, the expressionb/(b+l) is undefined. But it can be shown that in this
special case the benefit can be expressedusing natural logarithmsas:

(2] + PmQm - PcQc

Benefit - PmQm [log (PmQc) - log(PmQm)]
 - 29 -

3.41 These two related equations will be the centerpiece of the empirical
analysis in the next chapter. The benefit may be thought of as composed of two
parts. The first is comprised of the two terms to the right of the brackets in
equations [1] and [2]. This is simply the additional spending on non-housing
goods brought about by paying a rent, Rc (-PcQc),rather than Rm (-PmQm). This
simple differencebetween market and controlledrents, Rm-Rc, is often used as
an approximationof tenant benefits from the impositionof controls. But this
simple benefit measure does not take into account how householdsvalue changes
in housing consumptionin additionto changes in disposableincome. The second,
comprising the terms in parentheses and brackets in the two equations,depends
on the difference in housing consumption,with and without rent controls. But
whereas in the simple benefit measure (Rm-Rc) an extra dollar of non-housing is
counted as being worth exactly one dollar to the tenant, in the benefit measures
[1] and [2] extra housing is discountedbased on the tenant'srelativepreference
for housing vis-a-vis other goods.

3.42 The measures in [1] and [2] do not include all possible costs and
benefits to tenants. For example, rent control may increase transactioncosts
for tenants,including search costs (Clark, 1982), and increasewaiting time for
housing units (the cost of which to tenants may be considerable--seeWillis,
1984). All of these will reduce the benefitsto tenants,but the full system may
also increasethe bundle of propertyrights, such as securityof tenure,enjoyed
by tenants, thus increasingtheir benefits in this area. The above measures [1]
and [2] are then better approximationsof benefits than Rm-Rc, but they are still
approximations.

3.43 The cost imposedon landlordsis straightforwardly

approximatedby PmQc
- PcQc, or the difference between controlled and market rents for the unit
inhabitedby the tenant. This measure of cost to landlords is static and does
not include losses from prior accelerateddepreciationof the unit. However,
this could be regarded as a saving in maintenancecosts, which would generate
benefits elsewhere,perhaps equal to the opportunitycost forgone. The cost to
landlordswould also include losses from the uncompensatedtransferof property
rights to renters. Thus, the true costs to landlordsmay thereforeexceed the
(PmQc - PcQc) estimates.

3.44 Estimatingthese costs and benefitsrequiresfour pieces of information

for each consumer:

(a) the rent currentlypaid for the current controlledunit, PcQc;

(b) the rent that the current unit would rent for in the absence of
controls, PmQc;

(c) the rent that the household would pay if they were at their
equilibriumdemand at market prices, PmQm; and

(d) the price elasticityof demand for housing, b.

 - 30 -

Hedonic Estimates

3.45 Hedonic models have been widely used in environmentaleconomics and

housing market analysis for a variety of purposes. The purpose here, however,
is to use hedonic price models to estimate PmQc (market price for controlled
quantity)by comparing rents for differentkinds of dwellings in the controlled
and uncontrolledsector. Hedonic equationsare one way that rents for different
dwellings can be compared,or rents for identicaldwellingsin differentmarkets
can be predicted.

3.46 The hedonic model has the form:

R - f (S, L, C)

where: R X contract rent

S - structuralcharacteristicsof dwellings
L - neighborhoodcharacteristics,including
location within the market
C contract conditions or characteristics
which affect the price, such as utilities
included in the rent, length of tenure
1 lQ
in dwelling, etc.

3.47 The independent variables (S, L, C) represent the individual

characteristicsof the dwelling, and the regression coefficientsare estimates
of the implicitprices of those characteristics. The results provide estimated
prices for housing characteristics.It is then possibleto comparetwo dwellings
by using these prices as weights. For example, the estimated price for a
variable measuring the number of rooms indicates the change in value or rent
associatedwith the additionor deletionof one room. It indicates,in a dollar-
and-cents way, how much more "house" is provided by a dwelling with an extra
room.

3.48 A hedonic model can be used to estimate the implicit prices of

measurable housing characteristics in the uncontrolled market sector. The
coefficients of this model can then be used to estimate market rents for the
controlled sector units.

G. Empirical Problems

3.49 A few potential empiricalproblemsdeservediscussion:problemscaused

by misspecification,especially of the hedonic index and length of tenure in
hedonic and demand models; and definitionaland measurementproblems.

3.50 Possiblemisspecificationof the hedonic index. General discussionof

functional form, principles of variable construction, and so on, have been

JO/ Note that household characteristicswhich do not affect the price per unit
of services--suchas income--do not enter the hedonic regression. Household
characteristicswhich do affect the price per unit should enter.
 - 31 -

treated in detail in earlier papers. 111 To conserve space, we will focus

briefly on possible problems from omitted variable bias and, in particular, on
the effect of omission of the age of the structurefrom the hedonic regression.
Specificationproblems involvingthe length of tenurewill be treated separately
below.

3.51 Theory provides only a general guide to the choice of variables for
inclusion in the hedonic index. There are potentially hundreds of dwelling
characteristicswhich can affect rents or market values. Specification is
therefore always somewhat ad hoc and data-driven. Some variables which are
sometimesincludedin hedonic models are missing from the Rio data. For example,
we do not know the age of structures. Does this bias our results? The key
result from studies of this problem using datasets with more complete
specifications can be stated succinctly: omitted variable bias can indeed
seriouslybias the estimates of individualcoefficients,but given a reasonable
specificationwhich includes measures of dwelling size (such as rooms) and
quality (such as sanitaryfacilities),the bias in predicted rent for a dwelling
is negligible. 121 Analyzing individualcoefficientsand estimatingthe demand
for individualcharacteristicscan be difficultin practice for this reason. But
hedonic indexes reliably predict rents for individual units or for types of
units; thus, uses like ours, and the constructionof place-to-placeindexesusing
hedonics estimated in different markets, are fine.

3.52 It remains true that we would prefer to be able to estimate more

completemodels, especially given the key role that age of the structure plays
in the rent control system. But the fit of the equations (describedin Chapter
IV) is very good, and analysisof residualsand other tests demonstratethat this
specificationis adequate.

3.53 Possible misspecification of the demand relation. The general

specificationof the housing expenditureequation is described in Malpezzi and
Mayo (1985, 1987a). Those papers explain several shortcomings of the
specification, including the omission of a price term and the reliance on
household income rather than some measure of long-run permanent income. Those
papers described several tests of the specificationin markets with better data
and found that the biases due to these misspecificationswere remarkablysmall.

3.54 Length of tenure in hedonic and demand models. One issue which
deservesmore detaileddiscussionis the role of the length of tenure in hedonic
and demand models. First we will describe the role that this variable plays
generally in such models and then discuss some possible problems of
interpretationin a controlledrental market.

3.55 Length of tenure discountshas been documentedin many rental housing

markets, for example, Malpezzi et al. (1981, pp. 78-80). Supply costs may be
lower for landlordsrenting to tenantswho are a known quantityand often are at
least perceived as being more stable than prospectivenew tenants. Second, it

/.,Malpezzi et al. (1981),Mayo (1981),Malpezzi et al. (1987).

2/ See Butler (1982) and Ozanne and Malpezzi (1986).

 - 32 -

is easier for landlords to raise rents as new tenants move in, even in markets
without rent controls. Such raises are often customary to recoup the costs
associatedwith the search for a new tenant. Third, tenants have an incentive
to remain longer than usual in dwellingswhich rent for less than their highest
market value.

3.56 In a controlled market such as Rio's, where new units are not
controlled and fair rents are based at least partly on historical cost, rents
will decline even faster with the length of tenure.

3.57 Also, the uncontrolledsample is strictlycomposed of recent tenants.

Analysis of length of tenure discounts of uncontrolledmarkets shows that the
discount is typicallyvery large in the first few years, then tapers off. If we
estimated a length of tenure-priceprofile using recent movers and tried to
extrapolateto long-termrenters in the controlledsector, estimatedrents would
be biased downwards.

3.58 The problem, and possible Fiaure 3.5: RENS MD LENGTH OF TENE UMER
solutions,can be illustratedwith the RIO'S SYSTEM OF CONROLS
help of Figure 3.5. The free market Pile Pei Vuil II HousIog Serlvle,
discount profile starts steep and t
tapers off. Consider several alter-
natives:

(a) Estimate the hedonic ',

with the uncontrolled '
sample, including the
length of tenure as a '.'..".
right-hand side vari-
able. Use this coef- ,- ,- - ---

ficient to estimate the *.

rents of the controlled ' 21 "
units, most of which L .ia I Toaa to
have longtime tenants.
This procedure wiLl
underestimate market
rents for controlled
units.

(b) Estimate the hedonic as above, but cancel the discount by

computing the predicted rent without any length of tenure
adjustment. This will overestimate market rents for controlled
units.

(c) Exclude length of tenure from the regression completely. This

will also overestimate market rents for controlled units.

(d) Estimate with length of tenure,but rather than use the estimated
effect, as in (a), use an exogenous estimate of the discount for
predicting rent for controlled units. If reliable exogenous
estimates exist, this is the best procedure. If the exogenous
estimate is correct, the result is unbiased.
 - 33 -

3.59 We have adopted the fourth procedure. The exogenousestimatesare from

fifty-nineU.S. housing markets (Ibid.). It would be preferable to be able to
estimate this directlywith Brazilian data, but it is not possible. The choice
of procedure is dictatedby the fact that it will result in smaller biases than
the other three alternatives. The exogenousestimates presented in Malpezzi et
al. were used to derive the followingdiscount function:

log RA 2
- Xb - .03L + .0008L

if the length of tenure is less than or equal to eighteen years, and

log R - Xb - .30

if the length of tenure is greater than eighteenyears. This quadratic function

yields a discount that declines to 30 percent after eighteen years and is then
held constant (in percentage terms). 131

H. Cost-BenefitStudies

3.60 Perhaps the first careful study of the costs and benefits of rent
control is Olsen's (1973) paper. Using data from New York City in 1958, he used
estimatesfrom a hedonic index of uncontrolledunits to predict the uncontrolled
rentals of controlledunits. In an analogousfashion,he used the data from the
uncontrolledportion of the market to estimate the free market Engel curve for
housing services. The average controlledrent for an apartmentwas US$999 p.a.;
for purposes of comparison, the average income was US$6,229. The average
uncontrolled rent predicted by the hedonic results for those same units was
US$1,405, implyinga subsidyof US$406. The average free market expenditurefor
the controlledhouseholds was US$1,470, indicatingthat they consumed slightly
less housing than they would have in the free market. The average household in
the controlledmarket consumed about four and a half percent less housing than
they would have in the free market.

3.61 Olsen computed the economic benefit of rent control to each tenant
under the assumptionof a unitaryprice elasticity,i.e., using equation [2] from
above. Olsen's estimate of the average net benefit is US$213, little more tharn
half the gross subsidy implied by rent control.

3.62 The benefits are found to be slightly negatively related to income,

larger for larger households and larger for householdsheaded by older people.
The annual benefit is estimated to decreaseby about one cent for every dollar
of additionalincome,US$9 p.a. of head's age and US$69 per additionalhousehold
member. Olsen notes that these results may understate the regressivity of
benefits because lower-incomepeople are more likely to rent in the controlled
market and, hence, appear in the regression sample. Benefits do not vary
significantlyby race or sex of head of household. Rent control in New York City

12/ The interpretationof length of tenurevariable for owners in an uncontrolled

market is quite different. As Malpezziet al. explain,the variable correctsfor
appraisalbiases rather than price discounts in free markets.
 - 34 -

in 1968 appears to redistribute income, but very weakly, and in no way

proportional to its cost. Olsen showed that there is a slight tendency for
lower-incomehouseholds in New York City in 1968 to receive slightly larger
benefits.

3.63 A recent paper by Peter Linneman (1987) updated Olsen's study. In

1969, 1971 and 1974 New York made importantchanges in its system of controls,
with two key effects: (i) most post-1947 units were brought into the rent
control system; and (ii) a distinction was drawn between "controlled" and
"stabilized"units. Controlledunits (roughly,pre-1947 units with tenants who
moved in prior to 1947) have rents set by the Rent Control Division while
stabilizedunits (newer units or pre-1947units with new tenants)have rents set
by negotiations between landlord and tenant subject to approval of a board
comprising landlord, tenant and government representatives(see Linneman for
details). In 1981, about 19 percent of privately rented units were in the
controlledsector, 62 percent were stabilized,and 19 percent were uncontrolled.
Private rental housing was about 72 percent of stock.

3.64 Linneman examines (i) the length of tenure in each sector (controlled,
stabilized and uncontrolled) and (ii) rents paid in each sector compared to
predicted rents from hedonic regressionsfrom the stabilized sector. He finds
that, after controllingfor age and other householdcharacteristics,(i) tenancy
duration in the stabilizedsector, and (ii) the controlled sector, has a much
longer tenancy duration (twelveyears) than the other two sectors. The hedonic
confirms large differencesbetween the controlledand stabilizedsector rents,
after controllingfor quality:controlledunits rented for US$951 p.a. less (on
average) than the hedonic predicted they would rent for if stabilized. And, in
fact, rents are actually lower in the uncontrolledsector than predicted if in
the stabilizedsector (US$200less p.a., on average). A detailedanalysisof the
distributionof the benefits of rent controldemonstratethat the New York system
does, on balance, redistributeincome from high- to low-incomehouseholds,but
the effect is weak and very poorly targeted.

3.65 Daniel Pena and Javier Ruiz-Castillo (1984) carried out a similar
household level cost-benefitanalysis for Madrid. Madrid also has, in effect,
a two-tiered system. Roughly, units occupied before 1964 have their rents
controlled by the government. Only small increases in their rents have been
permitted. Units occupied after 1964 are under a slightlymore liberal system:
leases must be renewed, but at a rent agreed upon by the landlord and tenant,
subject to a government ceiling which is more generous than the increases
permitted in the strictly controlledsector.

3.66 The authors treat the post-1964 sample as approximatelyuncontrolled,

a limitationimposed by the data. They find an average monthly rent of Ptas 945
in the strictly controlled sector, while the average predicted rent for these
units (using moderately controlledhedonic prices) is Ptas 4694. The average
income in the strictly controlled sector is about 75 percent of the average
income in the moderately controlled sector, suggesting some redistributive
effect. However, extensivemultivariatetests suggestthat the subsidy is poorly
targeted:personalcharacteristics,includingincome,explain only 30 percent of
the variances in benefits. The size of the benefit is positivelycorrelatedwith
income. Further, households with 'Lowersocioeconomicor educational status,
 - 35 -

unemployed household heads and female household heads receive systematically

lower benefits.

3.67 Malpezzi (1986) presents estimates of the costs and benefits of rent
control in Cairo, Egypt. Controlled units in Cairo rent for much less than
estimates of their market rent in the absence of controls. However, this paper
shows that when side payments are taken into account, including key money,
utilities, maintenance and repair and upgrading by tenants, the discount is
greatlyreduced for the typical (median)household. When these are excluded,the
median estimate of the price per unit of housing services is about 38 percent
from the estimated long-run equilibrium free market price. When they are
included, the ratio increases to 70 percent of the market price. But it must
also be emphasizedthat there is a wide distributionaround this median. Quite
a few Cairo households do receive large discounts,just as some pay very high
prices for housing services. These differencesappear to be largely unrelated
to tenant characteristicsof ability to pay, raising questions of horizontal
equity. Otherwise equal households receive quite different housing "deals."
Most Cairo renters are well off their demand curve--much farther off than can be
explained by the stochastic nature of the estimated demand relation.
Corresponding to this departure from equilibrium, many households have
significantwelfare losses from under- and overconsumptionof housing services.
Underconsumptiondominates,but about a third of the renters consumemore housing
than predicted by their demand relation. This conclusion holds up even if
householdsvery far from their demand relationare analyzedseparatelyfrom those
within a 95 percent confidence intervalof their equilibriumdemand.

3.68 Struyk (1988) presents evidence on the distributionof benefits from

rent control in urban Jordan. Using Olsen's method, Struyk finds that average
benefits are equal to 27 percent of mean rents in Amman and 7 percent in smaller
towns. The distribution of benefits is only weakly related to income; lower
income householdsdo receive slightlylarger benefits, but the biggest benefits
accrue to householdswhich have been in their units the longest, regardlessof
income.

3.69 Malpezzi, Tipple and Willis (1989) analyzed the costs and benefits of
controls in Kumasi, Ghana. Ninety percent of Kumasi's population rent or live
as tenants in family houses. Based on 1986 data, typical controlledrents were
less than 2 percent of total consumption. A simple cross-countrymodel predicted
that the median rent-to-income level would be about .08 in the absence of
controls. Malpezzi, Tipple and Willis found that renters pay a fractionof the
estimated market rents for their units. The actual rent paidW4 is roughly
half the estimatedmarket. Furthermore,while the controlledrents PcQc hardly
vary, the estimated market rents PmQc vary with size and type of unit. Market
demand PmQm varies even more. The median cost of the subsidy implied by these
rent reductionsis estimated to be about 274 cedis per month in the tenementand
301 in indigenoussector. But household would spend even more on housing in the

14/ For the great majority of units, rents were fixed at 300 Cedis per room.
Such fixed rents are in some sense a more strict regime than Indian systems,
almost all of which permit some variation by type of unit. In 1986 US$1-C90
(approximate).
 - 36 -

absence of controls. Median estimatedmarket demand PmQm is over 1,000 cedis in

both sectors. ComparingPmQc and PmQm, it appears that while units rent for less
because of controls,householdswould spend even more at market prices; that is,
consumption of housing serviceshas been greatly reduced under controls.

3.70 Rent control imposesa landlordcost (PmQc-PcQc),which exceeds the net

benefit to tenants in both sectors. The "transferefficiency"(ratio of benefits
to costs) is therefore low. Under the most "favorable"assumption in terms of
controls' efficiency, the efficiency is 40-50 percent. Tenants receive net
benefits which are less than half the static cost to landlords. If the price
elasticity is on the order of -0.5, net benefits to most tenants is negative;
both landlords and (most) tenants are made worse off by controls. While costs
and benefitsare large relative to rents paid, they are small relativeto income.
The cost of the subsidy is usually on the order of 2-3 percent of consumption.
Net tenant benefits are, at best, negligiblecompared to total consumption.

3.71 The bottom line, then, is that rent control reduces the rents
householdspay, but the benefit of this rent reductionis more or less offset by
the welfare loss from underconsumptionof housing. We estimate that existing
units of typical quality would have rented for about twice the current rent in
1986,but that householdswould typicallyspend more than three times the current
rent--implyinghigher housing consumption,if supply was elastic.

I. Rent Control and the Housing Supply

3.72 Many theoretical treatments suggest that rent controls reduce the
supply of rental housing.151 Despite the keen interest in supply responses,
most of the work on supply to date has been theoretical, and/or based on
simulationmodels. Disappointinglylittle empiricalwork has been done on the
supply side of the housing market to date.

3.73 Figure 3.6 shows in a very Figure 3.6: STYLIZED DECOMPOSITION OF

stylized way how the supply side of RETAL HOUSING STOCK
the rental housing market can be
decomposed. Abstracting for the
moment from the existence of a related
market (an uncontrolledportion of the
rental market, and/or an owner
occupied market), the imposition of
controls and related tenure security
regulationscan (i) cause suppliersto Iy Rug
forego supplying rental housing; (ii) \r_
cause them to hold some of the stock l..
vacant; (iii) cause the remaining
stock to be used less efficiently. UeeuIle .i C.lIill*E
Foregone supply, (i), can be furt]her
subdivided into foregone starts, reduced maintenance and repairs (i.e.,

.j/ See the literaturecited in this chapterpassim, and Maclennan (1978),Olsen

(1969),Vitaliano (1985).
 - 37 -

accelerated depreciationof the existing stock) and premature demolitionof the

existing stock.

3.74 Most discussionsof the supply side effects of rent controls address
their presumed negative effects on rental starts;t1 but it is quite difficult
to unambiguouslyattribute declines in starts to rent controls (just as it is
possible that rent controlscould retard starts ceribusparibus while we observed
total starts rising). The extreme cyclical nature of the housing market! 71
makes imputing such effects difficult,as does the difficultyof acquiringgood
data on the number of rental starts separatelyfrom owner-occupiedstarts. Even
worse (from the point of view of analysts),investmentdecisionsare affectedby
expectations about controls, not by simply, or even mainly, the contemporary
policy environment. Finally, in many countries (most notably but not only in
developedcountries),the supply of rentalhousing is greatly affectedby tenure
conversions and good data on these are sorely lacking.

3.75 The same point can be made about premature demolitions,conversions

from owner-occupiedto rental, and vice versa. Data for direct tests are hard
to come by. There has been, however, some work on the effects on maintenance.

Reduced Maintenance

3.76 We have just seen that comparativestatic models which allow for some
price elasticityof supply of housing servicesimply that rationallandlordswill
permit their dwellings to deteriorate over time. If the rent control regime
fixes rent, it can be modeled as a tax or a tariff in the immediate market
period, but a reduction in the quantity of housing services supplied over a
longer period results in a fixed expenditure,i.e., lower quantity,with a return
to market price.

3.77 Dynamic models of profit-maximizinglandlords have provided some

insight into this process, and a little empirical support. One of the first
studies in this vein was by Moorehouse (1972),who examined the optimal choice
of maintenance inputs by a value-maximizing landlord. Using a three input
production function (capital, current inputs which must be employed at some
exogenous fixed rate, and variable maintenance inputs) and assuming geometric
physical depreciationof the unit, Moorehouseinvestigatesa rent control regime
which freezes nominal rents during an inflationaryperiod. In other words, the
time path of real rents falls by the rate of general price inflation and rises
by the rate of depreciationnot offset by maintenance. His model predicts that
in a competitive market with general inflation, dynamic equilibrium requires
concomitant increases in rents; but in a rent controlledmarket, the burden of
adjustment falls on maintenance.

1j/ Cf. Mayo et al. (1982) who suggest that large key money payments and the
absence of alternative investments in Cairo could fuel investments in rental
housing (which are then undermaintained).

1/ The ratio of housing starts peak-to-troughapproachesthree in the United

States, for example.
 - 38 -

3.78 Related studies by Dildine and Massey (1974), Arnault (1975), and
Kiefer (1980) reach the same conclusion: rent control reduces maintenance of
housing units ceterisparibus. Kiefer additionallystudies the optimal economic
life of a structure and finds that rent control leads to premature abandonment,
compared to the uncontrolled case. In a related empirical study of an
uncontrolledmarket, Bender (1979) finds that declininghousingprices in general
are associated with higher rates of demolition. This unsurprising finding
indirectlysupports Kiefer's conclusionsthat rent control reduces the optimal
life of a dwelling. It can do this because (i) the present value of rents is
lowered relative to alternative uses; and (ii) lower maintenance can directly
shorten the life of a structure.

3.79 Rydell and Neels (1982) estimate directly the elasticity of housing
services with respect to maintenance using data from the Housing Assistance
supply experiment. In any period, housing services are the sum of housing
services last period, plus some (nonlinear) function of the last period's
maintenanceand repair inputs, miniusgross depreciation:

Qt+l Qt + a4tL -bQ,

Using iterative techniques, they find that the best fit is obtained when a gross
depreciation rate, b, of 8 percent is assumed (L, the elasticity of housing
output with respect to maintenance, is estimated to be .17). In other words,
without maintenance dwellingswill depreciateby 8 percent p.a. This places a
boundary on how fast landlordscan decreasethe quantityof housing services as
a response to the impositionof controls.

3.80 There are conditionsunder which the effect of controls on maintenance

is ambiguous. Malpezzi (1986) and Olsen (1989) demonstratetheoreticallythat
rent control regimes which reset rents based on the condition of the unit could
increase maintenance (since the marginal return to a unit of maintenance
expenditures could exceed its cost if the entire unit was revalued past some
threshold). While such laws exist in some markets, it is not clear enforcement
is sufficientlydiscriminatingfor this effect to be observed in practice, and
no empirical test has been conducted.

3.81 In summary,directlytestingthe supply side effectsof rent control--

foregone starts, vacancy rates, conversions, demolition, and maintenance--is
inherentlydifficult,especiallywithout major improvementsin available data.
While it is difficultto directly test many supply-sidehypotheseswithout major
improvementsin data, indirectevidencecan be gleaned by analyzing the effects
of controls and related regulationson Rrofitability.

3.82 Malpezzi,Tipple and Willis (1990)analyzedthe effect of Ghanaianrent

controls on the profitabilityof rental investment in Kumasi. They studied a
number of cases, but their central representativecase suggested that controls
reduced the rate of return on rental housing from about 6 percent p.a. to about -
1 percent. However, they noted that in the recent past Ghana's economy was so
disrupted that -1 percent was not an unattractive return to some investors;
during the seventiesreturnsto Cedi-denominated financialassets were around -40
percent. Merely preserving capital--or losing it slowly--is attractiveunder
such conditions. Malpezzi, Tipple and Willis point out the paradox that as
 - 39 -

Ghana's economy recovers, the constraintcontrolsplaced on rate of return will

begin to "bite," and controls may restrain investment more under "normal"
economic conditions.

J. Rent Decontrol

3.83 Two earlier studiesemphasizethe analyticaland empiricalimplications

of alternativereal world price adjustmentmechanisms: Richard Arnott's (1981)
study of Ontario's rent control regime and the study of Los Angeles by Peter
Rydell and his associates (1983).

3.84 Arnott presents a complete analysis of an actual law and its

enforcement. Ontario has had rent control since 1975. The law applies to the
existing stock (pre-1975), but rents are not frozen. Landlords choose an
automatic six percent annual increase, or can apply for exceptionsbased on one
of three criteria: (i) on the basis of cost increasesexceeding the guideline;
(ii) to attain a modest rate of return, currently two percent; (iii) because of
a documented increase in the quality of the unit, presumablythrough increased
maintenanceand repaid expenditures.Most landlordstake the automaticincrease,
and the third option for exceptionshad not been used at all as of the study
date.

3.85 Arnott's analysisis primarily theoretical,some of it an extensionof

the analysis cited earlier. He reiteratesseveral importantarguments from the
literature: (i) controls unambiguously hurt landlords but their true
distributional implications are unknown; (ii) spillover effects into the
uncontrolledmarket are also ambiguous (see also Fallis and Smith, 1984, and
Marks, 1984); (iii) effects on new constructionof laws that currentlyexempt new
units depend on expectations of landlords, so new constructionmay still be
adversely affected. An interestingnew result from his analysis is that rent
control could actually increasebook profits in early years, which would be more
than offset by later losses. Arnott also provides one of the few cogent
discussions of alternative decontrol schemes.

3.86 Decontrol is an importanttopic not often analyzed in the literature.

The familiar comparativestatic models provide little guidanceon how best to get
there from here; by their static nature they assume blanket decontrolworks best
in all circumstances. There are a number of options which could be considered
for removing or relaxing controls. Arnott presentsa clear taxonomyof options:

Blanket lifting: all rent controls are completely removed as of a

certain date. This is the simplest method, but is very difficult
politically and may lead to short run dislocations.

Decontrol new construction: an obvious option which is being

undertaken in India and a number of other markets. But new
construction can still be inhibited unless government credibly
guaranteesunits will not come under controls later. Rents could also
be immediatelydecontrolledfor units which meet certain standards,
either now or after upgrading (e.g., for units which provide
acceptablewater supply and sanitation). Standards would have to be
 - 40 -

carefully chosen, however, to meet requirements without imposing

unnecessarycosts.

Floatingup and out: controlsare graduallyrelaxed,e.g., rent rises

are some multiple of CPI]or wage index changes,until controls are no
longer binding on most units. Then controls can be abolished. This
method can permit a smoother adjustment if potential landlords view
the gradual program as credible.

Vacancy decontrol: Units are decontrolled as they become vacant.

This method has been tried in some North American markets, but may
keep mobilitydown, with possibleadverse effects on housing and labor
markets.

Vacancy rate decontrol: particular markets are decontrolledas the

vacancy rate rises above some threshold. But while controls (and
other problems) remain, vacancy rates will probably remain extremely
low. How can vacancy rates increasewhile controls remain?

Rent level decontrol: decontrol by market segment. Rents could be

decontrolledfrom the top down or the bottom up. The current system,
with very low rent units exempted from the provisions of the Act,
embodies this to a limited extent. But such a system can provide
perverse incentivesto raise rents above long-run equilibrium levels
in order to escape controls or can lead to higher increases for the
lowest-incomeunits than would obtain in broader decontrol.

Contracting out: landlord and tenant negotiate a payment to the

tenant in return for his giving up the right to controls.

3.87 Of course, these options are not all mutually exclusive. In many
respects, floating up and out has some a priori appeal, because the market may
take time to respond,particularlygiven the current problems in input markets,
etc. Blanket lifting carries the danger of a sharp short-runrise in rents which
would be reduced over time. The present value model from above can be used to
study scenarios derived from the above.

3.88 Rydell et al. (1981) present an analysisof Los Angeles' rent control
regime. 18J

3.89 The authors use a combinationof empirical and simulationmethods to

measure the effects of the laws on rents, depreciation and conversion. In

it/ Los Angeles' rent control has;been in force since 1978. After an initial
freeze, rents for units occupiedby sitting tenantscould be increasedto recover
repaid costs, or in recent years by some fixed percentage (7 percent) which has
been lower than the general rate of inflation. Units are decontrolledas they
are vacated, however, and several classes of rental accommodation are
uncontrolled:newly constructedunlits,luxury housing (definedby pre-control
rents) and single family structures. Substantial rehabilitation (US$10,000-
US$17,000, depending on size of urnits)also exempts units from controls.
 - 41 -

Fikure 3.7: ALTERNATIVE DECONTROL

addition to analysis of the effects OPTIONS, LOS ANGELES
from recent years, they forecast the p i a,,., ..i.. e.
effects of six proposals for decontrol ##.,, O81 ,.,,
or tighter control (see Figure 3.7): ", .
-t...........---------------- ''---'--- ' t ' ... . .....................
..
............................
.. .

(a) End rent control ,. . . ,.-.

immediately (labeledno . ; / XJ.
law).

(b) Phase out current law --.--

over two years, meaning ,.........

two years of rent ,/.. *. .... $. 8

increases up to N 8 N
88 8 8 8 88 N
88 8 a
88 8

75 percent of the
general price level,
followed by complete
decontrol (labeled
phase-out law).

(c) Extend the current law (7.6 percent increasefor sitting tenants,
no cap for units when they turn over).

(d) Lower the increasefor sittingtenants to 5.6 percent,with no cap

on turnovers.

(e) Keep the 7.6 percent increase for sitting tenants, but cap the
increase at turnover at 10 percent.

(f) Lower sitting increasesto 5.6 percent, cap turnover increaseat

10 percent.

3.90 Examinationof Figure 3.7 highlights the wide range of outcomes from
alternative policies. All simulations assume a 10 percent increase in the
general price level and that, in the absence of controls, the relativeprice of
housing would remain fixed. Blanket decontrolresults in rapid price adjustment
and a slower adjustmentback to the long-runequilibriumhousing stock. A two-
year phase-inperiod, unsurprisingly,has similareffects with a lag. Note that
large real rent reductionof the tightestlaw (5.6 percent allowableincreasein
rents for sitting tenants, a 10 percent cap at turnover) is accompaniedby a
large decrease in the quantityof housing servicesavailable. Note that in this
case prices are being reduced at a rate which is declining over time, while the
quantity reduction is accelerating.

3.91 Estimating the time path of rents under alternativeregimes places a

lot of demands on any general equilibriummodel, and no such model has yet been
built for developingcountry markets. On the other hand, qualitativeinferen9es
about the time path of rents can be made in the absence of such a market clearing
model. Malpezzi, Tipple and Willis (1990) presented a simple analysis of
profitabilityand supply, and used it to make simple inferencesabout decontrol
by making what amounted to educated guesses about the time path of rents under
alternative changes in controls. The implicationsof these estimates can be
studied iteratively,and the sensitivityof profitabilityand affordabilityto
alternative plausible time paths of rent can be studied.
 - 42 -

IV. COSTS AND BENEFITS OF RENT CONTROL: DATA AND EMPIRICAL RESULTS

A. Household Survey Data

4.1 The primary data for the cost-benefitanalysis is derived from the
Brazilian census of 1980. The original data set contained approximately
3 percent of Brazil'spopulation--3.5million individualfiles and nearly 800,000
household files.12 1 Our subsample is comprised of 1,293 household files of the
metropolitanregion of Rio and representssome 0.03 percent of the total number
of households. Data on individualmembers of the household include variables
such as age, sex, race, education,marital status,religion,occupation,income,
length of stay, and national and regional origin. Data on housing
characteristicsinclude variables describingthe size of the unit, the physical
quality of walls, floor, ceiling and roof, as well as the type of water, sewage,
and electrical services available (refer to Annex B for a sample field
questionnaire.)

Housing Stock Characteristics

4.2 Physical Characteristics. The housing stock of Rio is predominantly

durable and formal. In our sample, over one-third of the units have at least
five rooms, and three quartershave more than three rooms. Few householdshave
to cope with rudimentaryconstruction. Even informalunits (about 8 percent of
the total sample)2 1 are often built with standard constructionmaterials such
1
as brick walls, wooden floors and tile roofs.

4.3 Access to Utilities and Services. Figure 4.1 shows the urban services
characteristicof our sample. While the percentageof units with sewage house
connections is identical to the independent estimate of WHO for Rio, water
coverage is slightly lower in the sample. If we expand our definition of
adequate services to include water drawn from wells and rudimentary sewage
systems,then nearly 90 percent of the populationwould be included. Practically
all households receive electricity.

4.4 Access to vital urban services such as water, sewage and garbage
collection in Rio is relatively good. House connections of water reach 79
percent of households (86 percent by WHO estimates),a number very similar to the
average Brazilian standards and way above the average for Latin America
(68 percent). As for sewage systemswith house connection,they cover 67 percent

L9/ See IBGE, Censo Demografico,Metodologiae Manual do Usuario, 1980, for a

complete description of the census implementation and analysis of sample
constructionand precision.

.Q/ Characterizedby units with inadequatewater and sewage systems. The number
of units in these conditionsunderestimatesthe phenomenonof "illegal"housing.
Overall, over 14 percent of all householdsin Rio lived in irregular conditions
in 1980.
 - 43 -
FiLure 4.1: CHARACTERISTICS
OF SPMLE HOUSINGSTOCK

Number of Roomi Rool Type

8.4 Cueotoll
rooms 42%
41% 1* 2

1. ~~~~~~~Wood
Io mI or Zing
I% 1%

rooms 32% TiIu.01111

O SSisad
45%

Waler Sewage

Wewgl#r No.. or
3% rudimoafory
No pipes 17%
12% SpltI
S. flak~~~~~~~~~~~~~Ta
1%~~~~~~~~~~~~~~~~~6
P~~~~it he e to

link
sirtie id aid
71% {si1et liak
.1%

of Rio's households, which is nearly twice as much as in the national and

regional averages. Adequate garbage collectionand disposal is provided to the
whole metropolis, including virtually all informalareas.

Household Characteristics

4.5 The great majority of renter householdsin Rio are small when compared
with other cities in developingcountries. More than half of the householdshave
four or less persons, and less than one-thirdhave more than six. This reflects

21/ COMLURBE, Rio's garbage collection and disposal public enterprise, has
developed mini tractor-likevehicles to be able to drive through the narrow,
winding and high grade areas where informal settlementstake place.
 - 44-

the absence of the extended Latin

family in the big city and indicates Table 4.1: WATER MAD SEWAGE
that, at least up to this point, rent COVERAGE IN RIO MID
control had not resulted in severe LATIN AMRICA
crowding. And literacy levels are at House Connection
least 10 percentage points above the City/Country Water Sevage
national average. In all, 91 percent System System
of the heads of households had at
least one year of schooling at the Rio de Janeiro 85.6 66.8
time of the census. Unexpectedly,in Brazil 86.3 36.2
a city with incomes nearly one-third Mexico 48.2 43.9
higher than the national average only Argentina
Colombia
60.9
75.1
32.4
55.1
10 percent of the households had gone Peru 60.8 49.6
beyond primary education. As for Chile 93.1 74.5
racial composition,blacks comprise 9 Ecuador
Guatemala
75.5
61.3
63.3
48.4
percent of the sample and whites 62 Bolivia 45.2 29.0
percent; and then there is a vast Haiti 28.6 0
segment of society which is El Salvador
Honduras
62.5
29.4
50.0
17.7
ambiguously classified as "brown" or Paraguay 46.2 30.8
"mixed race." Costa Rlca 98.0 36.7
Panama 95.4 72.7
Guyana 91.3 16.8
Belize 73.3 7.0
Fisure 4.2: CHARACTERISTICS OF SAMPLE Suriname 99.2 2.5
HOUSEHOLDS
Source: WHO, The International Drinkins Water
SuDDlY and Sanitation Review, 1987.

~~~~~~~~~~~~~~~~~~~~~..,.1-.,......
.-..... - ,

Nm,................ 4.6 Despite the existence of rent

,,,,,,,,,,,,m,
,. ~ ,I.I,,. controls, there is evidence of
significant mobility among renters. The
fact that 45 percent of households have
----------------.. .... . .., -,. . _ signed a new lease within the past 12
............ .. months, and that 71 percent of all units
....... :i t
A beenihaveinhabited by the present tenant
....
,. fi-
. 6 144k It#*&, ' for at most three years, indicates that
="-Ib*I." =" the long-term impact of controls on rental
""".'*"
"""'"" '. . housing will tend to be limited. This
high tenant turnover will also play an
important part in landlord rent-setting
behavior. A low average tenancy period
reduces the landlord'sinflation risk and
thus produces a smaller first period rent
than would be likely under a long tenancy.

4.7 Renters in Rio had a mean monthly income level of Cr$28,000 (US$528)
per household while rents averagedCr$4,475 (US$85). Figure 4.3 shows the steep
descent in the rent-to-incomeratio as income increases. To put it differently,
high-income tenants spend about half of the proportion of income spent by the
 - 45 -

poor on rent. This is the result of a low variance in rents (by income quartile)
combined with a highly skewed distributionof income.
Figure 4.3: RENT TO INCOME
RATIO BY INCOME QUARTILE
B. EstimatingCosts and Benefits ____________.__.__

of Rent Control

4.8 Two separate approacheswill be

taken for estimating costs and benefits. .. *. ... ..........
The first and simpler approach will be
based on the difference between actual ' ... -...
controlledrents and rents predictedby /Ii
the cross-country model. The second '"
approachwill be based on the Olsen cost-
benefit model. The parameters used to a a cl 4

derive the first set of estimates were aiim OatiI*

taken from Malpezzi et al. (1988) and were
presented in the previous chapter. The
parameters for the second set will be derived here from the household survey
data. Each will be compared to the actual current rent paid, PcQc.

4.9 Recall that the cross-countrymodel requires two basic pieces of

information:

(a) the rent currentlypaid for the current controlledunit, PcQc; and

(b) the rent that the household would pay if they were at their
equilibrium demand at market prices, PmQm.

The empirical estimationof the Olsen model discussed in the previous chapter
requires four pieces of information:

(a) the rent currentlypaid for the current controlledunit, PcQc;

(b) the rent that the current unit would rent for in the absence of
controls, PmQc;

(c) the rent that the household would pay if they were at their
equilibriumdemand at market prices, PmQm; and

(d) the price elasticity of demand for housing, b.

Both methods require PcQc and PmQm. For both we use the household survey data
for PcQc, but we use the cross-countrymodel to estimate PmQm in the first
instance and demand estimation,using a sample of "uncontrolled"(i.e., market
priced) Rio renters, to estimate PmQm in the second.

4.10 The rest of this chapter is organized as follows. First we will

describehow we estimate or otherwise arrive at the componentsjust described:
PcQc from the household survey; PmQm from a cross-countrymodel of housing
demand; PmQc using survey data and the method of hedonic indices;the alternative
measure of PmQm using the survey and demand estimation;and the price elasticity
 - 46 -

of demand. Finally, static cost-benefit estimates are presented for the

controlled units.

Choice of Reference Group

4.11 Perhaps the single most difficult empirical problem is choosing a

reference group. It must be reasonable to assume that they are enough like the
controlled group to be comparable--or can be made so statistically. It must be
reasonable to assume that rents are not so distorted in the reference group by
the presence of controls that they are unreliable guides to rents in the absence
of controls or that a good adjustment can be made for that distortion. Here we
note the following:

(a) Regression analysis is, in fact, a statistical method which

enables analysis of "treatment" and "control" groups which are not
identical.

(b) One possible problem is that households in the reference and

control groups are systematically different in their demand for
housing. Malpezzi (1986)has found such selectivity bias does not
make much difference in Cairo, but Caudill et al. (1987) found it
did make some difference in Vancouver. We can use a simple
correction from Olsen (1980) to test for selectivity bias in our
results.

(c) As noted earlier, rent controls can, under some circumstances,

affect rents in the uncontrolled sector. But the cross-country
model of housing demand can be used to test for and, if necessary,
correct for such a problem.

4.12 There are several variables in the survey which could be used to
estimate the market rent of the unit in the absence of controls:

(a) rents paid on units iin the informal sector;

(b) rents paid for units which are renting for greater-than-controlled
rents (are demonstrably uncontrolled.);

(c) estimate of amortized replacement value based on unit size and

standard construction costs; and

(d) new leases (i.e., those with less than one year duration).

None of these is without problems. Households under option (a) are not likely
to be representative of the uncontrolled sample; option (b) on the other hand,
introduces selectivity bias since the sample is chosen on the basis of the
dependent variable; and option (c) ignores the effect of land prices and
location. After examining the data we chose alternative (d). Under the Rent
 - 47 -

Control Legislationprevalentin 1980, all new leases were (and still are) exempt
from rent control (i.e., landlords are free to set rents at the time a new
contract is drafted).

4.13 We define our uncontrolledsubsampleas households that had signed a

new lease in the past twelve months. Despite the high inflation environment,
rents in our controlledgroup turned out to be rather constantacross time. Mean
rents of households leased within 0-6 months turned out not to be significantly
different from those leased within 0-12 months. In fact, regression results
using a 0-6 months or a 0-12 months group turned out to be almost identical in
signs and absolute numbers. Also, adaptations of the uncontrolled sample to
account for the legal provision that landlordsmay file a judicial request to
update rents to market levels every five years did not influence the results
substantially. This is in part due to informal contract settlements reached
between landlords and tenants before the fifth anniversary of the lease.

4.14 However, theremight be a tendencyfor the sample of uncontrolledunits

to be overepresentedby units that are newly built and therefore command higher
prices. And while units that have recentlybeen (re)introducedto the market are
outside rent control, common sense, as well as the model of Fallis and Smith
(1984), suggests that the presence of controls could still affect the rents of
these uncontrolled units as well. This could lead to upward bias in the
imputation of uncontrolled rents, even though we attempt to control for
characteristicsof the unit and the household.L3J

4.15 In the next few paragraphsthe issue of comparabilityof controlledand

uncontrolled samples will be addressed further. Then we will use the cross-
country model to examine the likely net bias of this procedure and to suggest a
correctionfor the bias.

Controlledand UncontrolledHouseholds Compared

4.16 Table 4.2 presents some key statistics. The evidence suggests that
there are no significantdifferencesbetween housing structurein the controlled
and uncontrolledsamples. The only exceptionbeing found in sewage systemswhich
are better in the units under rent control. Even then, it is largely a matter
of degree rather than kind: some 70 percent of controlled units have sewage
connectionsin contrastto 60 percent of the uncontrolled. The comparabilityof
samples allows us to concentrate on rent control as the dominant factor for
higher rents in the uncontrolledgroup, without any furtherneed for adjustments
to account for physical structuralbias.211

22/ Briefly, this can be thought of as the problem of predicting "out of

sample." Even though, in an idealizedmodel, correctly specified,there should
be no problem with such predictions,when working with real data and approximate
specifications(Leamer, 1978), predictingout of sample has some risks.

23/ However,we must still correct for the influenceof controlson uncontrolled
rents.
 - 48 -

Table 4.2: HOUSING STRUCTURE SAMPLES COMPARED

Characteristics Controlled Uncontrolled All Renters

Number of Observations 721 572 1,293

Rudimentary/l or no Sewage System 204 225 429
Rudimentary/b Water Connection 90 96 186
Rudimentary/c Walls 25 17 42
Rudimentary/d Floors 8 7 15
Rudimentary/e Roofs 26 8 34
No Electricity 15 14 29
Mean Number of Rooms (Bedrooms) 4.7 (1.8) 4.4 (1.6) 4.6 (1.7)
Mean Number Persons per Room
(Bedrooms) 1.1 (2.8) 0.9 (2.5) 0.9 (2.4)

No street link.
/b No house connection.
/c Wooden or other rustic material.
/d Dirt or other rustic material.
/d Zinc or other rustic material.

4.17 As fo r household Figure 4.4: RENT TO INCOME RATIO

characteristics, the uncontrolled IN CONTROLLED AND UNCONTROLLED
sample is composed of -
younger ~~~~~~~~~HOUSEHOLDS
households, with slightly fewer sxs
children and about the same level of
formaleducation as the controlled -- .
sample. The age of a representative
head of household in Rio is relatively -0
high--averaging forty-one years. And %. - _
this number conceals a sharp contrast
between controlled and uncontrolled _Medlin
samples. While the typical ,.
uncontrolled household is headed by a 01
thirty-seven year old, the average age --
in the controlled sample is over l I
forty-four. Moreover, salaries in the Cmlr,I,si, U,o,l all.ld A AUemleisi

unconstrained households are

substantially lower (mean and median
salaries are 25 and 13 percent less, respectively) and they pay a great deal more
rent for housing of very similar standards. Whereas the median uncontrolled
households must consume about one-fifth of their income on rent, controlled ones
spend a little over one-tenth. These statistics suggest that some vertical
inequities might be introduced by rent control. In fact, Chapter V expands on
the distributional issue and shows that rent control actually turns out to have
a progressive impact on income distribution.
 - 49 -

Table 4.3: HOUSEHOWLD SAMPLES COMPARED

Unit Characteristics Controlled Uncontrolled All Renters

Number of Observations 721 572 1,293

Mean Size of Household (Median) 5.0 (5.7) 4.0 (5.2) 4.0 (5.5)
Mean Number of Children (Median) 2.8 (2.0) 2.3 (2.0) 2.6 (2.0)
Mean Age (Median)tI 44 (43) 37 (34) 41 (39)
Mean Years of Schooling (Median)/I 4.2 (4.0) 4.2 (5.0) 4.2 (4.0)
Mean Length of Tenure (Median) 90 (60) 7.5 (8.0) 53 (24)
Mean Rent (Median) 4,325 (2,800) 4,665 (3,400) 4,475 (3,000)
Mean Income (Median) 31,377 (19,000) 23,755 (16,440) 28,005 (18,000)
Mean Rent/Income (Median) 0.18 (0.14) 0.24 (0.20) 0.21 (0.17)
/a
Head of household.
/b Head of Household.
/g Statistics are for Reent/Income variable. Note that sum or difference of medians (or of other
statistics) is not generally the median of the sum or the difference.

C. Rent Paid in the ControlledSector: PcOc

4.18 The first piece of information, the rent currently paid in the
controlled unit, PcQc, is directly available from the sample survey. The
distributionof rents is presented in Table 4.4 and correspondingFigure 4.5.

Table 4.4: CONTROLLED AND UNCONTROLLED RENTS PAID BY SECTOR (CRUZEIROS)

Rent (PcOc) Rent/Income

Q3 Median Q1 N Q3 Median Ql

All Renters 5,700 3,000 1,560 1283 0.27 0.17 0.10

Controlled Renters 5,310 2,800 1,500 717 0.23 0.14 0.08
Uncontrolled Renters 6,000 3,400 2,000 566 0.30 0.20 0.14

4.19 The resultsshow similarvariancein rent and incomebetween controlled

and uncontrolledhouseholds. Relativelylow variance in rents within each group
is, in turn, a reflection of the widespread use of standard construction
materials and the nearly universal availabilityof water, electricityand sewage
systems (see Figure 4.1). Within the uncontrolled sample, rent in the third
quartile is 77 percent (US$49)higher than the median, which is in turn only 70
percent (US$26) higher than in the first quartile. On the controlled units,
rents in the third quartileare 90 percent (US$47)higher than the median, which
is also nearly 90 percent (US$25)higher than in the first quartile.
 - 50 -
Fiure 4 .5S: RDI AND INCOME LEVLS CCHPARED

Ct$1Themmis C Thbge,edg
(TIeu,.mdp)

I' ... ,, v _ _- __ _ ........

1 ........

c.Il.ll.* u....l,*l| ^1l It..lff e..l,.ll* u...l..ll.s ^1l it......

CemtIaII.d U,..sIi,Itei ~~~All

Retains111816101 Allet
Reelt Level Compted Itme Level Cempesed

D. Estimating PmOm With a Cross-Country Model of Housing Demand

4.20 Using an average sample income of Cr$28,000 and an exchange rate of 53

cruzeiros to the dollar, the cross-country model predicts a rent-to-income ratio
of 0.16 for Rio in 1980 (see Chapter III for a discussion of the cross-country
model and applications).

4.21 Of course, these estimates from the uncontrolled variant of the cross-
country model are of the long-run equilibrium average rent-to-income ratio. In
order to predict the uncontrolled rent (PmQm) of particular households in a
controlled sample--or of representative households at an income level above or
below the average--it is necessary to combine an estimate of the average rent-to-
income ratio derived from the model, with an assumed within-market income
elasticity. Malpezzi and Mayo (1985) found most within-market elasticities for
renters ranged between 0.4 and 0.6. If anything, these estimates may have a
slight downward bias.24 1

4.22 There is no question that these estimates could be improved with a

still better cross-country model. While these estimates are reasonable and we
are confident of their utility for the rent control study, improving further the
precision of these estimates can have a high payoff.25'

Uncontrolled Rents Compared to PredLctions from the Cross-Country Model

4.23 Table 4.4 showed typical rents to be about 14 percent of total income
for controlled units and 20 percent for those that are uncontrolled. Using the
average income for Rio in 1980 cruzeiros, we predict that the rent-to-income
level would be 0.16 in the absence of controls. However, the median income in

24/ Analysis of the individual within-city elasticities from Malpezzi and Mayo
was unable to discern any relationship between the elasticities and income or
between the elasticities and the presence of controls.

25/ Particularly since there are many other uses of these estimates, such as
evaluating shelter projects and other government housing policies. See Mayo and
Gross (1986) and The Urban Edge (1984).
 - 51 -

the uncontrolled sector is Cr$16,440,compared to about Cr$18,000 for the city

as a whole. Using an elasticityof 0.6, we estimate the typical uncontrolled
rent-to-incomeratio to be 0.19.

4.24 Figure 4.6 presentsa simple Figure 4.6: ReT TO INCOE RATIO
comparison of these results. The RESULTS COMPARED
bottom line is derived from the actual 8al l"@m l
controlled data whereas the highest
line is from the uncontrolled Rioo
sample (unadjusted). The middle line @. _T - -- .
is from the uncontrolled demand .i...... .^===._- -
estimates of the cross-countrymodel. .=
In Rio, a typicalcontrolledhousehold
in the third-income quartile, with a 1...
monthly income of some Cr$25,000, I -Qia

would have an actual rent-to-income IncomeQuartils

ratio of 0.13. The median rent-to-
income ratio for an uncontrolled unit t +ueE
- - ..Ftti * liI'
in the same income quartile is 0.17, *f@toatoll
while the prediction from the cross-
country model is 0.19.

4.25 Given data on householdconsumption,the average rent-to-incomeratio,

and an assumed elasticity of 0.6, estimatingPmQm for each household with the
cross-country model is straightforward. In addition, when we estimate the
hedonic and demand relationsbelow with the uncontrolledsample, the sample rent
26 1
will be adjusted to reduce the bias.

E. EstimatingPmOm With Rio Survey Data

4.26 As noted above, the rent that households would pay if they were at
their equilibriumdemand at market prices (PmQm) can also be estimated from the
household survey. Table 4.5 presents the estimated coefficients from the
expenditureregression,estimatedusing the sample of uncontrolledrenters. The
dependent variable is (log) gross rent, adjusted as described above (i.e.,
multipliedby 0.9).

4.27 The demand results are reasonable,with an R-squaredof 0.50. And the
estimated income elasticityis 0.27, a slightlylower estimatethan most from the
cross-country results. Except for "Household Size," all coefficients are
significantat least at 1 percent level and their signs conform to expectations.

26/ See Malpezzi (1986) pp. 129-139 and Fallis and Smith (1984) for more
detailed justification.
 - 52 -

Table 4.5: DEMANDEQUATIONS

Dependent Variable: Log of Gross Rent (Adjusted)

Degrees of Freedom: 492

R square: 0.50 F: 60.59

R square(Adj): 0.49 PROB>F: 0.0001

Parameter Standard T
Estimate Error Statistics Prob>|T|

Intercept 4.590 0.259 17.66 0.0001

Log of Income 0.259 0.026. 9.80 0.0001
Age of Head of Household 0.011 0.002 4.59 0.0001
Household Size -0.001 0.009 -0.08 0.9355
Education Level (Head) 0.186 0.017 11.82 0.0001
Gender (Head-Male) -0.231 0.073 -3.15 0.0017
Race (Skin Color "Yellow") -1.484 0.604 -2.46 0.0144
Race (Skin Color "Black") -0.520 0.105 -5.94 0.0001
Race (Skin Color "Brown") -0.208 0.061 -3.41 0.0007

F. EstimatingPMOc With Hedonic Indexes

4.28 Hedonic regressionmodels can be used to estimate the second piece of

information required--PmQc,the rent that would be commandedby the controlled
units in the absence of controls.

4.29 Table 4.6 presentsthe hedonic index for 555 uncontrolledrenters. The
dependentvariable is the natural logarithmof adjustedrent, and the independent
variables are the number of bedrooms and other rooms occupiedby the household,
and a set of dummy variables indicatingthe quality of the unit's wall, floor,
sewage and electrical services. The coefficientsof linear variables can be
interpretedas (approximately)the percentagechange in rent, given a unit change
in the variable in question. The coefficient of a log variable can be
interpretedas the percentagechange in gross rent given a percentage change in
the variable. For a dummy variable, the coefficient is approximatelythe
percentagechange in rent comparedto some omitted category(e.g., not having the
use of a particularmode of a service).

4.30 In general, the regressionresults are quite good and the overall fit
of the equation compares very favorably with such models estimated in other
countries. Nearly 80 percent of the variance in the log of rent is explainedby
the model and coefficientsare generally of the expected sign.

G. Cost-BenefitMeasures

Costs and Benefits from the Cross-CountryModel

4.31 This procedure is very simple: we compute the predicted rent from the
cross-country demand model and compare it to the actual rent paid. In the
 - 53 -

Table 4.6: HEDONIC INDEX

Dependent Variable: Log of Adjusted Gross Rent

Degrees of Freedom: 518

• square: 0.79 F:73.01

• squared (Adj): 0.77 Prob>F: 0.0001

Variable Parameter Standard T

Estimate Error Statistics Prob>|T|

Intercept 7.9975 0.0876 91.293 0.0001

House (not apartment) -0.4880 0.0523 -9.330 0.0001
Wooden Walls 0.3303 0.3474 0.951 0.3422
Rustic Walls 0.9010 0.1330 0.684 0.4944
Other Walls 0.0980 0.2040 0.480 0.6311
Ceramic Floors 0.1502 0.0998 1.505 0.1329
Cement Floors -0.3099 0.2303 -1.345 0.1792
Dirt Floors -0.4501 0.0501 -8.989 0.0001
Oher Floors -0.6817 0.3260 -2.091 0.0370
Concrete Roof 0.0929 0.0457 2.032 0.0427
Cement Roof 0.5410 0.3435 1.575 0.1159
Zinc Roof 0.0087 0.0617 0.141 0.8878
Wooden Roof 0.3048 0.2976 1.024 0.3062
Other Roof -0.3671 0.2143 -1.713 0.0873
No Water System -0.6713 0.1267 -5.297 0.0001
WaterInternal Pipes, Well -0.2342 0.0676 -3.466 0.0006
WaterNo int. P., other -0.2186 0.2094 -1.044 0.2972
WaterNo Internal Pipes, Well -0.1670 0.0914 -1.827 0.0682
WaterNo Internal Pipes, Link -0.5129 0.0745 -6.888 0.0001
No Sewage System -0.3340 0.1454 -2.297 0.0220
SewageSeptic Tank -0.1443 0.0512 -2.821 0.0050
Rudimentary Sewage System -0.0680 0.0591 -1.151 0.2505
Other Sewage -0.1487 0.0960 -1.548 0.1221
No Electric Light -0.2823 0.1326 -2.128 0.0338
Unmetered Electric Light -0.0355 0.0483 -0.736 0.4621
Number of Bedrooms 0.1829 0.0269 6.802 0.0001
Number of Other Rooms 0.1361 0.0143 9.504 0.0001

notation above, we are comparingPmQm to PcQc. In other words, this producesno

estimate of the household's exact valuation of the benefits and is, at best, an
approximate measure of the cost of controls PmQc-PcQc. Although the cross-
country model is limited in analytical powers, it provides a glimpse of the
impact of controls on the uncontrolledsector. Thus, it is a useful complement
to the Olsen model, which only produceswelfare estimates for controlledunits.

4.32 The implicit assumption is that the cross-countrydemand results can

be used to predict long-runcompetitiveequilibriumrents in Rio. While we have
confidence in these cross-countrydemand models, it is clear that in any event,
the difference between estimates of PmQm and actual PcQc is large enough to
dominate any likely imprecisionof the estimates. Table 4.7 and corresponding
Figure 4.7 present summariesof the results. Note that rows of the table do not
add up exactly,because the sum or differenceof medians (or of other statistics)
is not generally the median of the sum or the difference. Actual rents (all
renters) are overwhelminglyin the range of Cr$1,560 to Cr$5,700,while most of
 - 54 -

the estimates from the cross-

country model range from Cr$2,135 Table 4.7: Str RY COST-BENEFIT MEASURES
to Cr$4,412. FRCM CROSS-COUNTRYMODEL

4.33 The model estimates that

tenants in controlledunits derives Actual
Rent
PredIcted
Rent
Difference
Pred.
benefits in the order of Cr$12 per -Actual
month in a median household, though
substantially more (Cr$204) for a Mean 4,475 3,590 937
representative consumer. Such Q3 5,700 4,412 -877
benefits are offset by average Median 3,000 3,057 -91
losses of Cr$325 in the Qi
3
1,560
1283
2,135
1278
-1907
1271
uncontrolled units to yield overall Representative
tenant losses of Cr$91. Tenant Consuer 3,000 3,057 57
losses in the uncontrolled units
will come about as landlords
attempt to recover part of their
losses in the controlledunits through markups in the uncontrolledones (i.e.,
they will impose markups in the first year of lease to compensate for later
periods' losses).

Figure 4.7: THE DISTRIBUTION OF

COSTIBENEMIT MSURES

Welltfs ChaRgel, I. Cdl

1500@

*150e . ,,,,....................................
................

Clrailloll Uleslrxill4 All Realarr

4.34 These results could be described as a 'smoking gun" signaling the

probableextent and directionof welfare change resultingfrom rent control. Or
rather, the absence of smoke indicates that the overall (controlled and
uncontrolled)static change in tenants'welfare is not likely to be substantial.
about the efficiencyof the transfer
However, the model cannot answer quxestions
of purchasing power to tenants, i.e., are the benefits received by tenants in
 - 55 -

line with the costs? For that, we turn to the Olsen model and the hedonic and
demand equations.

Costs and Benefits Constructedfrom the Hedonic and Demand Eguations

Table 4.8: COST-BENEFITMEASURES FROM SURVEY DEMAND AND HEDONIC MODELS

Market Cost of Rent Not

Current Rent for Estimated Control Welfare
Controlled Current Rent vith Subsidy to Benefit to Tenant Change
Rent Unit no Controls landlords 1 2 (Ben. -Cost)
PcQc PmQc PmQ

Mean 4,325 3,825 3,503 -447 -850 -1214 -390

Q3 5,310 5,141 4,450 979 745 569 -31
Median 2,800 3,176 3,060 175 -95 -356 -146
Q1 1,500 1,899 2,081 -953 -1,381 -1,762 -445
N 717 705 666 701 648 648 648
Representative
Consumer 2,800 3,176 3,060 376 374 372 -2

Note: Benefit 1 is benefit to tenants under unitary elasticty.

Benefit 2 is benefit to tenants under elasticityof -0.5.

4.35 Table 4.8 and correspondingFigure 4.8 present estimatesof the various
welfare measures and their componentsusing a variant of the Olsen's model which
was described in Chapter III.271

4.36 Each welfare measure can be calculated separately for each sample
observation(the approachused to derive the sample statistics)or calculatedfor
a representativeconsumer--constructed using, say, medians of the components.
The two approaches do not yield the same results. Again, note that rows of the
table do not add up because the sum or the difference of medians (or of other
statistics) is not generally the median of the sum or the difference.

4.37 The representativeconsumer rows are calculatedfor a representative

renter using median values of the components,PcQc, PmQc, and PmQm. Computing
costs and benefits separately for each observation allows us to study their
distribution. But since the true demand relationis unknown,and every household
is off their estimated demand curve, there is no information in the sample to

27/ The fourth piece of informationfor the Olsen model, the price elasticity
of demand, could not be directly estimated from the Rio housing data. Malpezzi
and Mayo (1985) surveyed estimates of this key parameter in a number of
developing countries and found that most estimates lie between -0.5 and -1.0.
Hence, these two values were chosen as upper and lower bounds for the cost-
benefit measures.
 - 56 Figure 4.8: COSTS AND BENEFITS FROM
RENT CONTROL, CONTROLLED HOUSEHOLDS
sort out how much of the difEference
between costs and benefits is due to the
stochastic nature of the demand relation, ,- -
and how much is due to rent control
(Olsen and Agrawal, 1982; Malpezzi, 1986;
and Gyourko and Linneman, 1986). V- , V

4.38 The estimates of PmQm are

generally higher than those taken
directly from the cross-country model, c.,, i lelt

but are of the same order of magnitude. c.st B*n.fit Measures

This is not surprising, since we also C,,M ,,s,i, ,
used the cross-country estimaltes to .i,Eicia.IiItIIj Ilp)..1
"calibrate" our estimates of uncontrolled
rents due to biases discussed previously. While the exact results here would
change given different cross-country estimates, the qualitative results presented
would not.

4.39 A typical controlled household pays rent which is not too different
from what it would pay if market conditions prevailed. The rent paid by the
median household is 90 percent of the estimated market rent. This meager
discount translates into a net loss of Cr$95 to the average renter (Cr$356
assuming the lower-demand elasticity of -0.5) once changes in housing consumption
are taken into account. For a representative tenant, there is a positive benefit
of Cr$374. However, this benefit is still exceeded by the cost of Cr$376 to a
representative landlord.

4.40 As expected, rent control imposes a significant, though not excessive,

static cost to the landlord of a controlled unit. The median cost of the subsidy
is estimated to be about Gr$175 pesrmonth or 6 percent of the actual rent. In
the case of a representative landlord, however, the loss is over 13 percent of
the actual rent.

4.41 The bottom line for tenrantsin controlled units is that rent control
reduces rents, but the benefit of this rent reduction is offset by the welfare
loss of consuming housing services outside the demand curve. Three points should
be made here. First, tenants ofternperceive that controls reduce rents even more
than they actually do.381 Second, these estimates do not account for
"persistence" or habit--tenants are now used to low rents, and change will be
resisted. Many tenants (in the higher-rent brackets) probably see no strong link
between low-rents and low-quality housing. Third, we estimate that existing
units of standard quality would rent for about 13 percent above current rents,
and that households would typically be willing to spend 9 percent more than
current rents, but 4 percent less than what the same unit would rent in the
market. The latter is a long run, comparative static result. How to get from
here to there is the topic of the next chapter.

28/ In what is, to our knowledge, the only direct test of these perceptions, a
study prepared for the city of Los Angeles by Hamilton et al. (1984) found that
tenants there believed controls reduced their rent by an average of 33 percent.
In fact, the average actual rent reduction in LA (which has a fairly loose rent
control system) was estimated at about 2 percent in 1981.
 - 57 -

V. POLICY ISSUES

A. Distributionof Costs and Benefits from Rent Control

Distributionby Landlord and Tenant

5.1 Table 4.8 shows that landlords of controlled units incur losses in
welfare as a result of rent controls. They lose Cr$175 per month in an average
unit and Cr$376 in a representativeunit. Tenants, also, lose an average of
Cr$95 (a Cr$374 gain for a representativeunit). Although we do not know the
income or wealth distributionof landlords,it is likely that the relativechange
in welfare brought about by the losses of tenants and landlords was
progressive."29 Evidently, transfers resulting from the impact of rent control
on controlled units are not substantialand neither is the expected change in
welfare distribution.

5.2 However, monetary losses of Fiture 5.1: TENAT BENEFIT RESULTS FROM THE
landlords are likely to be much OLSEN AD THE CROSS-COUNRY MODELS
smaller than we have estimated so far, c5l5,c.hlIll,d
B.eli
and tenants' losses much greater. In
general, landlords should be able to ""ss
recover most expected losses in the ---
controlled units through markups in _ _ _
the uncontrolled units (i.e., in the .,,.
early period of leases). As an ,, _ .
indication of relative tenant losses
in the controlled and uncontrolled ... ;
units, we compare the results of the-",,xc§L ..,s
Naive XC Medel Ceeu.me,luslipl..
cross-countrymodel with those of the al II 141
Olsen model. In the cross-country c I .1141 -llt
model, mean losses are 55 percent *,4, i .rc. sic
higher for uncontrolled (viz.
controlled) units, and median
uncontrolled households experience losses in contrast to gains in controlled
households. Losses in uncontrolledunits (cross-countrymodel) are also large
relative to those estimated from the Olsen model for controlledunits. Mean
losses are nearly one-third higher and median losses three-timeshigher.

5.3 Figure 5.2 shows the long-run cycle of rent markups and tenant losses
as estimated from the cross-country model. A typical tenant suffers a
considerableloss in the first year of rent and then recovers some of it between
the second and fourth years when renegotiationof the lease begins. Either
through informal renegotiationor through court-mandatedrevalorization,rents
receive another real increase to make up for another period of real declines
ahead. But the average unit that stays rented over a long period of time sees
its real rents continuouslyeroding as part of the bargaining and renegotiation
process (more common than the judicialrevalorization,which is costly, slow and

29/ The wealth distributionof Rio is found to be highly skewed, with a Gini
coefficient of 0.85 in 1980 (Silveira,1985).
 Figure 5.2: TENANTBENEFITS BY - 58 -
LENGTHOF TENURE

I.n
Cr1 i Tonsil,
its often biased against landlords). On
3110 the other hand, the average unit does
2 ..................................... not stay rented to the same tenant for
........ . ............... a long time. The average tenant moves
every two years or so and never fully
... ........ -.---..
--- "recovers" the first-period markup.
. ..........
. ... ........... Thus the average tenant incurs losses
*,... . ... in the rent control game in Rio.
VB;6,11.
,ssstbsl
It oii
CI
oi"@
14"1'
~ 0 441 as ($44
_I4)
It|
fil
gFigure
5.2 shows that these losses are
LENGTHOFTENURE' even higher if we use estimates from
the Olsen model. The cycle of markups
is still reflected in the results,but
:l.u...,:,..l
1iis, Le'all,lls:",, under the Olsen model, it is a cycle
%t *41,411
ist It Cic.tled of relative losses: (i) high in the
Cu.t4

median of the first quartile of length

of tenure (twenty-fourmonths), (ii) low in the second quartile (forty-eight
months), (iii)high again in the third quartile (eighty-fourmonths), and finally
(iv) registering a gain in the fourth quartile (180 months).

5.4 Since tenants' losses might actually be large, why are they so
supportive of rent control legislation? Part of the answer lies in their
unawarenessof PmQu (the market value of uncontrolledhousing services),PmQc and
PmQm. The only rental values seen by society are PcQc and PuQu (actual
controlledand uncontrolledrents). However,ignoranceof market prices is only
one--perhapssmall--elementof the story. One reason for resistance to change
in the law is that, at any point in time, a number of tenants (i.e., those
between markups) will have already paid the first-periodmarkup and now stand to
gain (i.e., recover earlier losses) from the permanence of controls.

5.5 For most tenants, the main benefit of rent control is rent stability.
To many, gains from the rent stability afforded by rent control far outweigh
monetary losses. However, rent stabilityis a service involuntarilyprovidedby
landlordswho assume the often substantialinflationrisk. And this subsidyhas
varied wildly in the past decade as rents have been allowed to increaseat rates
that ranged from zero (e.g., price freezes in 1986/87) to 90 percent of
inflation,and as inflation varied from zero to 30 percent per month.

5.6 The involuntaryinflationrisk imposed on landlordshelps explain the

continuing exit of investors fromithe rental market. Other factors, such as
costs, of housing production, relative housing values appreciation, and
alternative investment opportunities, are considered in Section C of this
chapter.

Distributionby Income

5.7 Because the income distributionof Rio is very unequal,301the impact

of rent control on relative incomes is a crucial issue which is always present
in the rent control debate. Since we do not know the distributionof income of

IQ/ For tenants in our sample, the Gini coefficientwas calculatedat 0.49--
indicatinghigh inequality,though lower than for the nation as a whole.
 - 59 Et&ue_3.3: Rent to Income Ratio
by Length of Stay
landlords, we concentrate our _________________

attention on the impact of control on

tenants' welfare and income
distribution. As it turns out, net ; . .
losses of tenants are positively . . . =.
related to household income, ,,,.+ .
suggesting that rent control may be * +
somewhat progressive. *

5.8 Within controlled units, I 2 Is

households in the highest-income z I6 1.5 1 4.2* 6. l.1 I. 0.5 I.2I
4 . t 1.1810.1i 1.1
01 _ l.l | 0.13 0.11 0.2 .11
quartile experience losses of 1.2 *
- 6.08 026 I0.21
.1 6.2.1 .31 6.1
Cr$1,136, while those in the lower al - 6.06 6.2 6.66 6.6 6.0I Ol 6.66 6.66 8.61
quartile actually gain Cr$154. These ", ''
results from the Olsen model are very
similar to cross-country model
estimates of a Cr$43 gain in the lower quartileand Cr$1,227 loss in the upper
quartile. The cross-countrymodel also estimates losses in the upper quartile
of uncontrolledunits to be 40 times higher (Cr$2,416)than those in the lower
quartile (Cr$60).

5.9 Greater losses for richer tenants in the uncontrolledunits comes as

no surprise. This concurs with other empirical evidence that high-income
householdsare better able to bear the full increasein first-yearrents required
to bring the present value of rents to what it would be without rent control (see
Box 2.1). They are better able to absorb rent increasesbecause they have larger
uncommitted income and better access to financialwealth and borrowing than it
is the case for lower-incomehouseholds. Figure 5.5 shows tenants' losses-by-
income quartile in controlledand uncontrolledunits.

5.10 Likewise,we can explain the apparentoverconsumptionof housingby the

upper quartile of controlled units (i.e., PmQm < PcQc) by a lagged--though
weakened--effectof the progressivelylarge first-yearmarkup. For the poor, a
small gain is realizedbecause the value of housing servicesreceived is greater
than actual rents (i.e.,PmQc> PcQc). And this factor is stronger than losses
originating from underconsumptionof housing (i.e., PmQm > PcQc).

5.11 More important(to the welfare of poor tenants)than the impact of rent
control on income distributionis its welfare impact relative to income levels.
As it turns out, rent control is also particularlyhostile to tenant households
in the highest quartile. Among the controlledhouseholds, the welfare benefits
of the poor representapproximately2 percent of income relativeto an equal loss
of 2 percent for householdsin the highest quartile. For the uncontrolledunits,
however, the cross-countrymodel estimatesthat losses to the poor amount to less
than 1 percent of income while the rich loses over 5 percent.
 - 60 -
FiLure 5.4: TENANTS' BENEFITS BY INCOME
QUARTILE

MedIan Benefit within Inco me Quar ile

500

*500 _..... ........

1000 .. . .......

_10 -.... .. ..

* 20002 0 0-
~ ~~~~~...........
.................................................................................................

2500 I I
Qf 02 0
a4
CS Contiolled 154
I 158 *130 *1136
XC Con Ito Ile d i85 3118 233 * 11
XC Uaeoantiolled 601 is7 *512 .241I

Distribution by Class of Tenure: Formal and Informal Markets

5.12 Summary cost-benefit measures for the controlled units were also
tabulated by tenure class. We define a tenure class as formal or informal based
on housing conditions. Essentially, any unit that has no sewage system, or only
a rustic one, and has no running water may be considered informal in the context
of Rio. There are 101 units that fall under this classification, of which forty-
six are controlled units. Our definition is perhaps overly restrictive since
other estimates of slums in Rio put the share of informal households at almost
twice our numbers. Nevertheless, this categorization of households highlights
a distinct subgroup: those living in the worst conditions within city slums.
Table 5.1 contrasts the characteristics of informal and formal households for the
whole sample.

5.13 As expected, the formal and informal household profiles are quite
distinct. Median length of tenure in the informal sector is half of those in the
formal, and income levels are substantially lower in the informal sector. Median
and mean incomes in formal households are more than twice and 3½ times that in
the informal, respectively (nearly three and four times, respectively, in the
controlled units). And informal sector households tend to be younger and to have
a greater number of dependents to support on their lower incomes. Rents are also
 - 61 -

Table 5.1: HOUSEHOLD SAMPLES COMPARED

Characteristics Informal Formal All Renters

Number of Observations 101 1,192 1,293

Mean Size of Household (Median) 2.7 (2.0) 2.9 (2.0) 2.8 (2.0)
Mean Number of Children (Median) 2.9 (2.0) 2.5 (2.0) 2.6 (2.0)
Mean Age (Median)1l 38 (36) 41 (39) 41 (39)
Mean Years of Schooling (Median) 2/ 2.7 (4.0) 4.4 (5.0) 4.2 (4.0)
Mean Length of Tenure (Median) 44 (12) 55 (24) 53 (24)
Mean Rent (Median) 1,114 (1,000) 5,040 (3,717) 4,475 (3,000)
Mean Income (Median) 8,541 (8,000) 29,654 (19,162) 28,005 (18,000)
Mean Rent/Income (Median) 3/ 0.15 (0.15) 0.13 (0.13) 0.21 (0.17)

1/ Head of household
2/ Head of Household
3/ Statistics are for Rent/Income variable. Note that sum or difference of medians (or of other
statistics) is not generally the median of the sum or the difference.

much lower in the informal units averaging about one third of formal sector
levels. However, simple rent comparisons are not very useful given the great
difference in the quality of housing offered in the two sectors.

5.14 Figure 5.6 shows the welfare impact of rent controlon controlledunits
in the two markets. Tenants experience losses in both formal and informal
markets. However, while low-income households generally benefit from rent
control, those living in informalhousing do not. In fact, median tenant losses
in the formal sector are less than one-thirdof informalsector losses. Part of
the welfare losses in the informal sector is due to underconsumptionof housing
and part to overpricing of housing services. Whereas rent control enables
tenants in the formal sector to pay rents which are 9 percentbelow market, those
in the informal sector pay 8.5 percent more than what the unit is worth.311
It is little wonder that slumlords register a welfare gain from rent control,
while formal housing landlords register a loss.

5.15 As a share of income, informal sector tenants are slightly more

penalized. Whereas tenants' losses in the formal sector represent 2 percent of
median income, losses in the informal sector are 3 percent of income.

Sj/ For the overall controlledsample, the first and second income quartilespay
rents which are, respectively,13 and 29 percent lower than the market. Part of
the story that we might be missing here relates to the relative enforcementof
rent control in the formal and informal sectors. Rent control might be less
strictly observed or altogetherdisregardedin the informalsector.
 62 -
Piturg l: DISTRIBUTION OF LOSSES IN TPE FORMAL
AND INFORXAL
UNZTS

Oil tlb. . . -- 250 ,I $ .

1~~~~~~~~~~~~....
............... ,., ... ... ................ .0.14
,

.cermfJ.ream
t|111
l....
......................................
ac
1**111 .... ................ ......................... . 6.-
.... e ei.................
lsr:omlee ................... .s.elsl^ue:Ilra
et
cd~~~~~~~~~~~~~~~~~~7 |
I I............. ....... ..... * |it|o
..... ......................
. 0.16"Etma
CoilemacallitMalurze:FormatSector coi( Bnctlio
Measures:Informal sector

tenure
choice i p 1 ouie di, to date,

who
rent theirIhomes.- MeanyI studms of tenuecoic havmele been t
caredemt, i

developed4 contie (e.g. Rosen 197,, Li. 197.tI. Omit,001, Miesuulytidta

stud

B. A Simfcle Model of Tenure Choice

5.16 The rental sector in Brazil has experienceda continuousdecline over

the last 25 years (see Figure 5.10). But would the rental sector have declined
anyway in the absence of controls? And by how much? Few studies have analyzed
tenure choice in developing countries, and no studies have, to date,
systematicallyanalyzed the effects of rent control on the percent of population
who rent their homes. Many studies of tenure choice have been carried out in
developedcountries (e.g., Rosen 1979, Li 1977). These studiesusually find that
income and stage of the life cycle are importantdeterminantsof tenure choice,
as is the relative cost of owning versus rentingefe But other than the study
of Korea by Lim et al. (1980),remarkablylittle empiricalwork has been done on
tenure choice. Limitedby the econometricsof the day, they aggregateda number
of tenure categoriesinto owning and renting,and estimatedan OLS tenure choice
model for that simple binary choice. Not surprisingly,the authors found that
income was a significantdeterminantof homeownership.

12/ In the U.S., the tax code has a strong effect on the relative price of
tenure and varies with income since the chief tax break, the mortgage interest
deduction, increases with income. This partly explains the strong demand for
homeownership for middle- and upper-incomeAmericans.
 - 63 -

Flnn S56: URWBANTENUREAMDPERCENTURBAN Figure 5.7t URBAnTENUREAMDGNP PER CAPITA

to .. .. . ,, ,,It III
,l, ,, I1 ,
Frseml Urbam GNP Pst Capita iLl Seals)
I -I1.X. 3,6*: mte|l11UrII., I,. guaduall
*4,1;*:Cit slsl rig Igril,.
.43i.*4
iii I*SI. l.~~~~ III hr Caplu..
,I
~~~~~*.4 .1. siJi.1.4 Iii 5 Ih.. Ii.
l.. g..

5.17 A number of papers have presented evidencethat, in some cities, large

fractionsof low-incomehouseholdsown in the informalsector; as incomes rise,
they rent in the formal sector, and the richest again become homeowners. Yet
such patternshave not been scrutinizedor explainedcarefully. Strassman (1980)
suggests that the availability of services such as piped water may catalyze
investmentby some householdsand make the shift to rentingsuch units attractive
relative to current ownership of informal units without such amenities. In a
very stylized version of such a world, we would observe the lowest-income
households owning very low-qualityhousing, perhaps in the informal sector or
with little tenure security;past some threshold,householdswould begin into a
higher-quality rental submarket. Finally, at higher incomes and (perhaps)
overcoming financial constraints,households would be able to purchase such
housing.

5.18 Malpezzi (1990)

examined the pattern of these Table 5.2x SIMPLE CROSS COUNTRY
proportions across countries. TENURE REGRESSION
Figures 5.7 and 5.8 present
simple plots of the percent of Sample: 57 Countries
households renting against Dependent Variable: Percent of Urban PopulatLonRenting
Adjusted R-Squared: 0.13
percent of total population Degrees of Freedom: 52
found in urban areas and GNP Standard Prob>
per capita (the latter in CoefficLent Error ITI
logrithmic scale). His simple
regression model quadratic in Percent Urban/. 0.0112 0.0039 0.006
percent urban population and Percent Urban/k -0.0001 0.000037 0.002
GNP Per Capita/g 0.000028 0.000023 0.236
GNP per capita yields the GNP Per Capita/k -1.14E-09 1.33E-09 0.395
result presented in Table 5.2. Constant 0.1616 0.0997 0.111
/A 1985
5.19 Note that the percent k Squared
of urban renters first /- 1986
increases, then falls as
average per capita incomes
 - 64 -

rise; but that the effect is weak. A similar pattern can be found with percent
urban; the effect is stronger, but still not pronounced.

5.20 Future work can improve this model by constructingexplicit measures

of the relative price of each tenure form and the effect of controls on this
cost.331 But even this simple model can be used to predict the percent of
urban householdswho would rent in Brazil.

Fimure 5.8: PERCENT RENTERS IN BRAZIL,

ACTUAL AND PREDICTED

5.21 Plugging in values of the

relevantvariables for Brazil in 1960, 1§
the model predicts that in 1960 about
46 percent of urban Brazilian " - _
households would have rented their
homes; actually,about 43 percent did. ae/ ----------- :------
What was the experience of the next
twenty-fiveyears? In 1985 the model % --
predicts about 40 percent rental,
while only 29 percent actuallyrented. sex ---

5.22 Such simple calculationsdo

not prove that rent control caused liii
such a steep relative decline by ACS _ IL(%
itself,34 1 but they are consistent
with such an effect. Also, it is
worth noting that many of the markets which are used to calibrate this simple
cross-country model are themselves controlled; comparing Brazil and other
controlled markets to a better-specifiedmodel is an obvious area for future
research.

C. A Present Value Model of Housing Investmentin Rio de Janeiro

5.23 So far the results presentedhave focused on the costs and benefits to
individual landlordsand tenants. 'Whatare the effects of controls on supply?
In order to understandhow controls affect incentivesto landlords, and hence
supply, a simple present value model can be used to study how controlsaffect the
profitabilityof representativerental investments.

.13 Constructingsuch a variable is possiblebut requiresa fair level of effort

(Malpezziand Mayo, 1987b). To a household, the user cost of a rental unit is
the periodic rent paid plus any deposits or key money payments appropriately
discounted plus their own payments to others for housing services (e.g.,
household maintenance expenditures). User cost for owners is even more
complicated,since it must account for financing,depreciationand inflation.

14/ For example, BNH subsidizedhomeownershipfor a significantfractionof the

population during this period.
 - 65 -

5.24 Government subsidies, regulations,taxes and other interventions--

including rent control - - change the cash flows to landlords. Some
interventionsimpose costs (e.g., land use regulations,taxes, rent controls,
building regulations) and some benefits (e.g., land subsidies, tax relief,
financial subsidies) to landlords. The incidence of costs and benefits is
discussed in more detail in Malpezzi (1988) and Malpezzi (forthcoming).

5.25 Present values are a summary of the cash flow and its components.
Present values are computedby adding a stream of net costs and benefits from an
investmentafter discountingthem to account for the fact that a Cruzeiro today
is worth more than a Cruzeiro tomorrow.351

5.26 Consider a simple four period example:

PV - A 0 + A 1 /(l+r) + A 2 /(l+r) 2 + A 3 /(l+r) 3

where A representsthe net costs and benefits in each of four periods, and r is
the discount rate, or the opportunity cost of capital. For example, if an
initial investmentof 300 Cruzeirosis followedby three years of net returns of
150 Cruzeiros per annum, and the discountrate is 10 percent, the present value
is:

PV - -300 + 150/(1+0.1)+ 150/(1+0.1)2+ 150/(1+0.1)3- 73 Cruzeiros

The present value rule states that if the present value of the investment is
greater than zero, the investment yields more than the opportunity cost of
capital (the normal rate of profit for an investment of that type), and the
investmentshould be undertaken.

5.27 A closely related concept is the internalrate of return. This is the

discountrate at which the presentvalue of the cash flow would be zero (benefits
would equal costs, adjustedfor the timing of receiptsand expenditures). It can
be interpretedas a measure of profitability.

A Present Value Model of Housing Investmentin Rio de Janeiro

5.28 Rent controlIsmarket effects can be analyzedusing a simple cash flow

model of a representative rental investment. 36 Table 5.4 presents such a

15/ Even in the absence of inflation.

16/ The general method used is described in any corporate finance text (e.g.
Brealey and Myers 1981). Application of cash flow models to investment in
developingcountries is discussed in (e.g.) Gittinger (1982) and Mishan (1982).
Examplesof housing policy analysisusing such models includeDeLeeuw and Ozanne
(1981),Brueggeman(1985),and Malpezzi (1988). Similarmodels have been applied
to Ghana (Malpezzi,Tipple and Willis, 1990) and India (Malpezzi and Tewari,
1990).
 - 66 -

model.)! Each column represents a year's time. Landlord-developersare

assumed to build or purchase a unit in developmentperiod (year 0) and rent it
out for ten years. During this time landlordscollect rents and spend money on
maintenance and taxes. At the end of the ten-yearperiod, the unit (structure
and land) has some salvage value.2381

5.29 The example in Table 5.3 is meant to represent, appproximately,a

rental unit from 1980 in the strict:Ly
controlledsector, so that results from the
precedingchapter and from the household survey can be used to calibrate it. But
the model itself is quite general,and will be used with more recent data below.

5.30 This model is simple,391 yet it allows us to compare two different

rent regimes, labeled the baseline regime and the revised regime in Table 5.4
We can examine the interaction between controls, taxes, maintenance,
depreciation, profitability, and affordability in a simple but consistent
framework.

5.31 But the profitabilityof the rental investmentis only half the story.
The present value model can be complementedwith a demand side, calculating
affordability,willingness to pay, and consumers'surplus measures of benefit.
Figure 5.5 presents the layout of the demand side of the model.

5.32 Of course the model has limitations. It focuses on a "representative"

investment,and the exact numbers presented aren't exact for all or even most
units. But we can analyze more than one "representative"investment(including
differentstructure types, service levels, locations,and rents). We don't want
to focus on point estimatesbut rather on robust qualitativeconclusions.

5.33 The cash flow model is onLy as good as its inputs; "garbagein, garbage
out". But we can and have tested the model with a range of inputs not all
reported here, and while the exact numbers change, the qualitative conclusions
drawn from the simulationsreportedbelow remain robust.0',

1i/ Figures 7.1 and 7.2 are copies of the output of the model. The model itself
is constructedin a spreadsheet. (,opiesof the model are availableupon request
from the authors. Reading and using the model requiresan IBM compatiblePC with
640K RAM, a graphics card, and Lotus 1-2-3 version 2.1 or a compatible
spreadsheet.

1/ The salvage period is often discussed as if the owner sells the unit.
Actually it makes no difference to the analysis if the owner retains it; the
salvage value is the opportunitycost of doing so.

j2/ For example, we assume the landlords pay cash for the unit. While it
remains true that all durable assets must be financedin some way (even if self-
financed) ignoring finance is an appropriate simplificationfor the present
purpose. The analysiscouldbe readily extendedto evaluateproposalsfor rental
finance.

iQ/ We encourage interestedreaders to undertake their own analysis of other

representative investments and using other parameter values (especially
representingother changes in controls). The model itself is written in Lotus
1-2-3, and is availableupon request from the authors. Using the computermodel
requires the Lotus spreadsheet system, version 2.0 or higher, which is not
 - 67 -

Table 5.3: Key Model Inputs

CASE 1
--- MODEL PARAMETERS ---

GLOBAL &
BASELINE REVISED

Units in Structure 1 1
PRICES
Inflation Rate 100% 100%
Real Discount Rate 10% 10%
Relative Price Changes:
Land 3% 3%
Structure 2% 2%
Wages 0% 0%
CHANGESIN RENTS
Annual Rent Inflatio 90% 100%
Rents Reset Every x 10 10
Demand:
Average R/Y: 0.17 0.17
Elasticity: 0.60 0.60

MAINTENANCE & DEPRECIATION

Gross Dep to K Cost 5% 5%
Maint to K Cost 0.5% 0.5%
Tenant Maintenanc 1.5% 1.5%
Net Dep Rate 3.0% 3.0%

TENURE SECURITY
Secure-l, Evictable= 1 0

TAXATION

LL Income Tax Rate 30% 30%

LL Pays-0, Tenant-l 1 1
1-KV,O-Rent 1 1
Assesment Ratio 50% 50%
Property Tax Rate 1.00% 1.00%

available from the authors.

 Table 5.4: Cash Flow Mode1 of Rental Investment
Case 1: 1 BR Rlo Apartment1 1980
COMPARE CURRENT REGIME TO BLANKET LIFTING
Year 0 1 2 3 4 5 6 7 8 9 10 Present
General Price index 1.00 2.00 4.00 8.00 16.00 32.00 64.00 128.00 256.00 512.00 1024.00 V lues IRRJ

Market Value of Land 125,000 125,000

Market Structure Value 250,000 250,000
Financial Cost of Land (125,000) (125,000)
Financial Structure Co (250,000) (250,000)
…_____________________________________________________________________________________________________
BASELINE RENT CONTROL REGIME: Rents increase by 90 percent of inflation, tenure security
_______________________________

NOMINAL Monthly HE Rent 2,500 4,750 9,025 17,148 32,580 61,902 117,615 223,468 424,589 806,719
REAL Monthly BH Rent 2,500 2,375 2,256 2,143 2,036 1,934 1,838 1,746 1,659 1,576 12,819
NOMINALAnnual Gross Rent 30,000 57,000 108,300 205,770 390,963 742,830 1,411,376 2,681,615 5,095,069 9,680,631
REAL Annual Gross Rent 30,000 28,500 27,075 25,721 24,435 23,213 22,053 20,950 19,903 18,907 153,832

Salvage Value of LAnd e Highest & Best Use 167,99o 64,767

Salvage Value of Structure 6 Highest & Best Use 226,762 87,426
Value of Final Rent e Discount Rate 189,075 72,897

Maintenance (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (7,681)
Rental Income Taxes (9,000) (8,550) (8,123) (7,716) (7,331) (6,964) (6,616) (6,285) (5,971) (5,672) (46,150)
Property Taxes (938) (469) (234) (117) (59) (29) (15) (7) (4) (2) (1,562)

Memo: Final Capital Value I Highest & Best Use 394,751 152,194
Memo; Actual Final Capltal Value 394,751 152,194
Landlord's Real Cash Flov,
Baseline Regim (375,000) 18,813 18,231 17,468 16,638
14,970 15,796
14,172 13,408 12,678 406,735 (124,366)1 4.61X
…-----------------------------------------------------------__---------------__-----------------------------------------------------------__-------
REVISED RENT REGIME: Real rents fixed at 902 of old initial rent: no security of tenure

REAL Monthly BH Rent 2,250 2,250 2,250 2,250 2,250 2,250 2,250 2,250 2,250 2,250 13,825
NOMINAL Monthly HE Rent 2,250 4,500 9,000 18,000 36,000 72,000 144,000 288,000 576,000 1,152,000
NOMINALAnnual Gross Rent 27,000 54,000 108,000 216,000 432,000 864,000 1,728,000 3,456,000 6,912,000 13,824,000
REAL Gross Rent 27,000 27,000 27,000 27,000 27,000 27,000 27,000 27,000 27,000 27,000 165,903

Salvage Value of Land e Highest & Best Use 167,990 64,767

Salvage Value of Structure e Highest & Best Use 226,762 87,426
Value of Final Rent e Discount Rate 270.000 104,097

Maintenance (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (1,250) (7,681)
Rental Income Taxes (8,100) (8,100) (8,100) (8,100) (8,100) (8,100) (8,100) (8,100) (8,100) (8,100) (49,771)
Property Taxes (938) (469) (234) (117) (59) (29) (15) (7) (4) (2) (1,562)

Memo: FLial Capital Value e Highest & Best Use 394,751 152,194
Memo: Actual Final Capital Value 394,751 152,194
Landlord's Real Cash Flow,
Revised Regime (375,000) 16,713 17,181 17,416 17,533 17,591 17,621 17,635 17,643 17,646 412,399 | 115,917)1 5.07Z
__--------------- __-----------------------------------------------------------
…----------------------------------------------------------- __-----------
OLD REGIME COMPARED TO NEW
_______________________________

Change in Rental Income (3,000) (1,500) (75) 1,279 2,565 3,787 4,947 6,050 7,097 8,093 12,071
Change in Central Taxation 900 450 23 (384) (769) (1,136) (1,484) (1,815) (2,129) (2,428) (3,621)
Change in Local Taxation 0 0 0 0 0 0 0 0 0 0 0
Change in Maintenance 0 0 0 0 0 0 0 0 0 0 0
Change in Final Value 0 0
Net Change in Landlord Revenue (2,100) (1,050) (53) (895) 1,795 2,651 3,463 4,235 4,968 5,665 8,450
Net Change in Government Revenue (900) (450) (23) 384 769 1,136 1,484 1,815 2,129 2,428 3,621
 Table 5.5: Demand SLde of Rental Investment Model
------------------------------------------------------------ __---------------__-----------------------------------------------------------__-----------

AFFORDABILITY ANALYSIS
_______________________________

Affordable e Real MONTHLY Incom of:

(Under Old Regime): 12,852 11,799 10,832 9,945 9,130 8,382 7,695 7,064 6,486 5,954
(Under Nev Regime): 10,782 10,782 10,782 10,782 10,782 10,782 10,782 10,762 10,782 10,782

Memo: Income Distribution Assuming All Incomes Rise 100X as Fast As Inflation
Quintile 1 59,498 59,498 59,498 59,498 59,498 59,498 59,498 59,498 59,498 59,498 59,498
QuintlLe 2 29,000 29,000 29,000 29,000 29,000 29,000 29,000 29,000 29,000 29,000 29,000
Quintile 3 18,000 18,000 18,000 18,000 18,000 18,000 18,000 18,000 18,000 18,000 18,000
Quintile 4 11,000 11,000 11,000 11,000 11,000 11,000 11,000 11,000 11,000 11,000 11,000
Quintile 5 5,519 5,519 5,519 5,519 5,519 5,519 5,519 5,519 5,S19 5,519 5,519
----------------------------------------------------------------- __----------__----------------------------------------------------------------__------

CONSUMER'S SURPLUS
_______________________________

MONTHLY Income% 14,000 14,000 14,000 14,000 14,000 14,000 14,000 14,000 14,000 14,000 14,000
MONTHLY PmQc 2,250 2,250 2,250 2,250 2,250 2,250 2,2.50 2,250 2,250 2,250
PcQc 2,500 4,750 9,025 17,148 32,580 61,902 117,615 223,468 424,589 806,719
PmQn 2,632 2,632 2,632 2,632 2,632 2,632 2,632 2,632 2,632 2,632
Subsldy (250) (2,500) (6,775) (14,898) (30,330) (59,652) (115,365) (221,218) (422,339) (804,469)
Beneflt (281) (2,531) (6,806) (14,928) (30,361) (59,683) (115,395) (221,249) (422,370) (804,500)
DWL 31 31 31 31 31 31 31 31 31 31
Efficiency 112X 101X 100X 1002 100X 100X 100X 100x 100X 1001
ANNUAL Subsidy (3,000) (30,000) (81,300) (178,770) (363,963) (715,830) (1,384,376) (2,654,615) (5,068,069) (9,653,631) (8,660,815)
ANNUAL Tenant Benefit (3,368) (30,368) (81,668) (179,138) (364,331) (716,198) (1,384,745) (2,654,983) (5,068,437) (9,653,999) (8,663,077)
ANNUAL Deadweight Loss 368 368 368 368 368 368 368 368 368 368 2,263

WILLINGESS TO PAY BY INCOME QUINTILE (First Year Only)

-------------------------------------------------------

Quintile I 1 2 3 4 5
Cum Pct 10X 302 50X 702 90X

Income (midpoint of
each qulntile) 5,519 11,000 18,000 29,000 59,498

WTP (first year) 1505 2277 3060 4074 6270

WTP/Income 27.32 20.7X 17.02 14.0X 10.5X
1st Yr Rents (Old Regi 2,500 2,500 2,500 2,500 2,500
1st Yr Rents (New Regi 2,250 2,250 2,250 2,250 2,250
0

WILLINGESS TO PAY BY INCOME QUINTILE (Fifth Year Only)

-------------------------------------------------------

Quintile 1 2 3 4 5
Cum Pct 102 30X 502 70X 90S

Income (midpoint of
each quintile) 5,519 11,000 18,000 29,000 59,498

WTP (Fifth Year) 1505 2277 3060 4074 6270

WTP/Income 27.3X 20.7Z 17.02 14.02 10.52
5th Yr Rents (Old Regi 2,036 2,036 2,036 2,036 2,036
5th Yr Rents (New Regi 2,250 2,250 2,250 2,250 2,250
 - 70 -

5.34 Key inputs to the model are underlinedin the tables. The first table
lists underlying assumptionsabout changes in market conditions, depreciation,
tax rates, and demand. Other numbers are calculatedby the model given these
assumptions. The landlord'sfinancial cost of building or acquiring the unit
(here 250,000 1980 Cruzeiros for the structure and 125,000 for the land)4'1may
be greater or less than their corresponding value but here we assume the
structure and land are worth what they cost, i.e., we focus on the "marginal"
unit. Baseline-controlledrents are assumed (in this example) to begin at a
nominal level of 2,500 Cruzeirosper month, and are partially indexed; they are
assumed to rise 90 percent as fast as inflation, consistent with recent
regulations on rental increases. The first comparison case assumes rents to
remain at 2500 Cruzeiros per month in real terms, i.e. they keep pace with
inflation.

5.35 Other assumptionsfor this first example include a general inflation

rate of 100 percent per annum; a real discountrate of 10 percent;real land and
structureprices, and wages, also rise only as fast as general inflation (i.e.
their relative price remains constant.) Gross depreciation of the unit is
assumed to be 5 percent per annum; spending more on maintenance is assumed to
reduce net depreciation one for one.421 Landlords pay an income tax of 30
percent on rents collected. Householdsat the median income are assumed to be
willing to spend 17 percent of their income on such a unit; the income
elasticity,.6, assumes that lower-incomehouseholdsspendhigher fractions,and
vice versa.

Gains and Losses from Four Componentsand Their Interaction

5.36 Five key componentsof landlordcash flow are: initial outlay, rents,
maintenance, taxes, and capital gains. Initial outlay does not change when
controls are removed, but the other four do. Rent control directly reduces
profitabilitybecause it reduces the rents a unit can command. But reduced rents
also affect maintenance (and depreciation),taxes, and capital gains. These
"indirect"effects can be large and should be taken into account. Figure 5.10
summarizesthe changes.

5.37 Rents. Brazil's rent control regime allows only partial indexationof
nominal rents and hence reduces real rents. In the example presented in Table
5.4 above, the reductionin real rents is modest comparedto similar calculations
for countrieswith stricter regimes. The present value of rent over ten years
decreases from about 166,000 Cs. to about 154,000, or a drop of only about 7
percent. In contrast, Malpezzi, Tipple and Willis' first calculation for
Kumasi's stricter regime estimated that the present value of uncontrolledrents
would be over eight times the present value of controlledrents, over a similar
period. In Bangalore,Malpezzi and Tewari estimated that the present value of
market rents would be about three times controlledrents over such a period.

AV/ Today's units would have much higher nominal values.

2!J One way the model could be extended would be to build in a more
sophisticatedproductionfunctionrelatingmaintenanceto depreciationin a non-
linear fashion.
 - 71 -
ZFaur X. Chana in Couyonant. of Landlord's
5.38 Taxatio. If taxes on rental Present Value
income are collected from landlords, pVy Coemp,a,ma
el CulbFlowI.ig
11l
rent control reduces these taxes as it "
reduces rent. This partially offsets a- - g m
the reduction in rent to landlords, but
also decreases government revenue.l3 . M
Not surprisingly, since effects on
rents are small, so are effects on
taxes. Controls reduce the present t
value of taxes increases from about
50,000 1980 Cs. to about 46,000.441

5.39 Maintenance. Landlordshave

the option to increase or decrease ' ,,,, Us. u. Fi lv,
maintenance. While good data are
lacking, we assume in these first
simulations that maintenance on a controlled unit is a minimal .5 percent of
structure cost. Since controls don't greatly affect rents we assume no change
in maintenance behavior (unlike in our parallel simulations for Kumasi and
Bangalore).

5.40 CapitalGains. Generally,capitalgains (and losses)stem from several

sources. First, structures and land can appreciatemore or less than general
inflation. We have already noted that the simulationpresentedhere assumesboth
land and structure prices move with inflation,i.e. an initially conservative
assumption. Second, the land and the structuremay originally(at period 0) be
worth more or less than the value of resourcesput into it. We've assumed that
the land for this example was valued at market prices, and that the original
value of the structureand the land (in the baselinecase) is worth its financial
cost to the landlord-developer. Third, the real value of structural capital
declines as the unit depreciates. As we've seen, depreciation depends on
maintenance,and in this version we've assumed a simple one-for-oneoffset.

5.41 Fourth, if the rent control regime changes, increasesin rents can be
fully or partially capitalized into value. This is worth discussing in some
detail, because it is the key to measuring the economic effects of tenure
securityprovisions.

5.42 Tenure Securityand Value The value of an asset at any particulartime

is simply the expected present value of all future uses of the asset. When we
talk about the future value of an asset (e.g. its salvagevalue in ten years, and
any associatedcapital gain or loss), we are implicitytalking about our current
expectationof the present value at that time of the rest of the asset's useful
life. Suppose that at the end of the notional ten-yearholding period landlords
can convert their unit to its "highestand best use," i.e. the use that maximizes

43/ If the household spends the extra cash on goods or services that are taxed,
the reduction in government revenue will be partially offset. We assume this
effect is small.

i&/ Assuming of course that taxes are fully collected.

 - 72 -

the present value of its worth in the current market.45J In an uncontrolled

market the "highestand best use" could be owner-occupiedhousing,or a shop, or,
indeed, it could still be worth more as a rental unit than as any of these
alternatives. The salvage value of such a housing unit is simply the present
value at that time in its highest use. In a rent-controlledmarket, because
controls reduce rents it is less likely that the present value of future
controlled rents would exceed other possible uses (such as owner occupation).
How likely or less likely partly depends, of course, on the size of rent
reductions from controls. The actual salvage value depends greatly on whether
the unit must actuallyremain a controlledrental unit, or whether the unit can
be converted to some other economicuse.

5.43 We can use a simpleversion of this in the present value model to study
the effects of strict (virtually complete) tenure security on landlord
profitability. That is, landlordscannot evict tenantsas a matter of course at
the end of a lease period. Our baseline "normal" tenure security is one where
landlords can evict at the end of the lease. If landlords are permitted to
convert to highest and best use, the salvage value is calculatedas an estimate
of the market price of the unit, i.e. our best estimateof the price at which the
unit would change hands between willing buyers and sellers if there were no
"unreasonable"restriction on its use.46/ If landlords are not permitted to
convertby completetenure securityregulations,the salvagevalue is the present
value of expected future controlled rents, which could be considerably
lower.'/ Again, note that in terms of future value it doesn't matter whether
or not they do convert but only that they could.

5.44 If landlords can convert to highest and best use at the end of the
simulationperiod, then there is no differencein capital gain between our first
two regimes.481 If strict complete tenure security were resurrected, capital
gains would be cut. The final value of the unit would be about half that in the
uncontrolled case (144,000versus 73,000, in present value 1980 Cr$).

A5 It only matters that landlords could convert their units. It is not

necessary to assume that the land:Lordactually sells the unit.

A6/ That is, permittinghouseholdsto convert to owner-occupiedhouses or shops-

cum-houses.

Aj/ For these particular simulationswe've assumed full capitalizationof the

differencebetween the average net income streams for the two regimes. We've
also assumed that changes in controlswere completelyunanticipatedat the time
of initial investment. These assumptionscould be compared to alternativesin
future work.

A./ Rememberwe have also assumed no differencein maintenancebehavior for the

moment. If controlledunits were undermaintainedthere would be corresponding
decreases in capital gains.
 - 73 -

Effects of Rent Control on Landlord Profitability

5.45 The overall net present value of the controlled and uncontrolled
investments,laid out in Table 5.6, as well as the correspondinginternalrate
of return, or discount rate at which the present value of the unit is zero (the
landlord-investorbreaks even).

Table 5.6: SUMMARY RETURNS I

Net Present InternalRate
Value @ 10% of Return

ControlledUnit, Normal
Tenure Security (Lease) (124,366) 4.61%

Controlled Unit,
With Complete Tenure
Security (203,664) -1.12%

UncontrolledUnit (115,917) 5.07%

5.46 At a real discount rate of 10 percent, the present value of all three
alternativesis negative, i.e. the highest of the three internalrates of return
is lower than our (arbitrary)a priori discount rate. Assuming some rental
investment would occur in Rio, this suggests one or more of the following.
First, our rough cost estimates could be too high for units at these rent
levels.491 Second, a real discount rate of 10 percent is quite a high
threshold (which we will discuss later under changes in risk), and the actual
discount rate could be lower, at least for uncontrolled units. Third, our
assumptions about increases in the relative price of land and structurescould
be conservative, ex ante if not ex post. And finally, landlords could have
expectationsabout increases in real rents as well.

5.47 Of course it would be straightforwardto adjust parametersto hit any

target rate of return, but we prefer to go with our first estimates for this
initial analysis. Since we really focus on the difference in returns from
controls and from changes in tenure security, the initial levels make little

Ai2/ Including depreciationand recurrentcosts as well as capital costs. Good

estimates of costs and rents in consistent 1980 Cs. are difficult to obtain,
given Brazil's rapid inflation.
 - 74 -

qualitative difference.U01 Here, the central message is clear: rent

indexation reduces returns to rental investment,but not by much As long as
landlords have the option of convertingto other uses and can reset rents to
market levels from time to time. Other simulations,not reported here, confirm
this result.

5.48 Another way of lookingat Brazil'scontrolsis to comparean indexation

scheme not unlike the current one to a restrictiverent freeze, A la Ghana and
some other countries. When the model was run with these policy assumptionsand
Brazilian inflation,the present value of costs was roughly400,000,but revenues
were less than 50,000. The cash flows were so bad that a rate of return could
not be calculated. Brazil'srent indexation(at current levels,with ability to
convertupon lease expiration)is closer to no controlsthan to strict Ghanaian-
style controls.

5.49 Of course these simulationsare indicativeat best. One problem not

often discussed in the literature on controls is that in real world housing
markets, rents do not increasein lockstepwith the overall price level, even if
supply is elastic. It is well documented, for example, that in developed
countriesrents increasefaster on the fringe of cities, and faster for new units
than for old.5 11 Low-income households tend to live in housing which
appreciates more slowly; this has been well documented in the U.S.521 If this
pattern holds for developing countries as well, controls will ceteris paribus
reduce rent increases of high-income tenants more than low-income tenants,
blunting any redistributiveeffect. Since controlswill depress returns to new
units more than to old, they could have a greater negative effect on supply
because the actual effect on the marginalunit is greater than our model predicts
for the "average"unit.

Changes in Risk and InvestmentDecisions

5.50 Several paragraphs abovrewe noted that initial assumptions about

discountrates were necessarilyarbitrary. While we cannot predict the level of
the discountrate for rental housing investmentwith confidence,at given assumed
discount rates the qualitativedifferencebetween regimes is clear and robust.

.20/ The point of building such a model is to ennable rapid comparisonof many
such assumptions. Copies of the spreadsheet implementationof the model are
readily available to users who want to analyze other assumptions. We are quite
explicit about the fact that trying to come up with the "true" parameter
estimates for such a model is a sterile approach. There is no "true" vector of
parameters. Such modeling must focus more on analyzing relevant ranges of
parametervalues. Many criticismsof this work are possiblebut "you do not know
the true discount rate" is not, in our view, admissible.

.aJ/ See Muth (1969) for the classic discussion.

5/ The relationshipbetween rents, values, and capitalizationrates has not

been adequately explored in developingcountries,to date.
 - 75 -

5.51 However, in the simulationsdiscussedabove the discount rata itself

was held constantbetween regimes. In fact, one of the most important effects
of controls may well be changes in discount rates for such investments,because
controls increase the risk of such investment.

5.52 Mechanically,modeling such effects is easy: changing the discount

rate for present value calculations,or an equivalentchange in the threshold
rate of return required to undertake a rental investment. But estimatingthe
effects of controls, tenure security and other regulationson the risk premium
landlordswould demand for rental housing is quite difficult,particularlysince
such premia will be driven more by expectations about future regulation than
current. It is quite conceivable,for example,that blanket decontrolundertaken
without reaching some political consensus on decontrol and without concomitant
measures to ensure an elastic supply responsewould not reduce risk premia at
all, since landlordsmight well expect more stringentregulatoryregimes further
down the road.

5.53 Such changes in risk and discountrates could well dominatedetails of

rent indexation (which do not change rates of return by an enormous amount, at
least as practiced in BrazilN- 1). They might not dominate changes in tenure
security legislation,which can have large effects on returns.

Changes in Rates of Return and Changes in Supnly

5.54 Given the relativelysmall effects of rent indexationon revenues,can

we infer that there are correspondinglysmall negative effects on supply? That
depends.

5.55 The traditionalhousing market literatureassumes that the supply of

housing services is very elastic.311 This would imply, of course, that small
reductions in rates of return would lead to large reductions in supply. The
assumptionof elastic supply has been subjectedto empiricaltests. The majority
of such tests have been carried out in the United States, and have supportedthe
hypothesis of elastic supply.55t However the elasticity of supply is not a
state of nature; it depends particularlyon the policy environmentin a country.
Countries which have well-functioninghousing and housing input markets, and
appropriate regulatory environments,will have more elastic supply than those
that do not. Stephen Mayo has estimated supply elasticities in several
developing countries.56J Table 5.7 supports the hypothesis that elasticities
are related to regulatorystringency.

li/ Obviously reductions from 90 percent indexationto (say) 50 or 60 percent

would be a different matter; or if inflation acceleratedeven further.

.4/ See Olsen (1969) for a clear exposition of the implicationsof elastic
supply of housing services.

.5_/ Muth (1960), Smith (1976) and Follain (1979) are the best known empirical
studies, and all support elastic supply.

ii6 His estimates are contained in Hannah et al. [(1989),pp. 84-90)].

 - 76 -

Table 5.7: ESTIMATED LONG-RUN PRICE ELASTICITIESOF

SUPPLY OF HOUSING SUPPLY

Assumption about LR demand elasticity: E7-1.O E,-1.5

Restrictive Regulatory Environments

Korea 0.10 0.40

Malaysia 0.14 0.46

Less Restrictive Regulatorv Environments

Thailand 6.64 10.21

U.S.A. 22.03 40.04

Source: Calculations by Stephen Mayo, Annex 1 of Hannah et al. (1989).

Price elasticitydepends on assumptionsabout long run demand elasticities;
estimates are presented for reasonablerange of such assumptions.

5.56 Direct estimatesof the price elasticityof supply of Brazilianhousing

are not available at this time. A reasonable conjecture is that Brazilian
housing markets are more elastic than Korean or Malaysian, since the regulatory
environment is not generally as stringent.

5.57 These estimates are for the total supply of housing, not rental
housing. Measureswhich reduce the return to rental housingwithout reducingthe
return to owner-occupiedhousing reduce the supply of rental housing, but will
be partially offset by increases in owner-occupiedhousing, to the extent owner-
occupied supply is also elastic. But the offset will not be perfect, even if
supply is very elastic, because many households, particularly low-income
households, find the transactioncosts of homeownershipprohibitive. That is,
it bears reminding why rental housing issues are important, qua rental,
especially if one is focused on the welfare of low-incomehouseholds.

5.58 If the supply elasticity of rental housing was as high as, say,
Thailand's total elasticity, and rent indexation reduced the present value of
typical rental investmentsby 9 percent (impliedby our calculationsabove) then
rental housing supply would be reducedby about half in the long run. If it were
half as high, rental housing supply would be reduced by a quarter. Some of this
reduction would be picked up by increases in owner occupation,but how much is
impossible to predict even on such a rough "back of the envelope"basis.

5.59 These rough estimatesof supply effects also depend on the time frame.
All the supply elasticityestimatesand the rough calculationsabove are for the
 - 77 -

long run, which is long indeed in housing markets given the size and durability
of the asset. Most housing economistswould suggest the adjustmentprocessjust
describedwould take at least a decade. Hence controls trade off short-termand
fairly concreterent reductionsfor long-termand uncertainreductionsin supply.

5.60 While we have not developed a formal time series model of aggregate
supply and demand,Table 5.8 summarizesthe simplebivariaterelationshipbetween
housing supply and changes in controls.

Table 5.8: IMPACT OF RENT CONTROL REGIMES: 1912-85

RENT RENT INCREASE ECONOMIC SHARE OF

PERIOD CONTROL IN REAL TERMS GROWTH RENTAL HOUSING

1912-21 NONE NONE* WEAK 0.70

1921-27 WEAK MODERATE MODERATE -
1927-39 NONE NONE WEAK 0.49
1948-50 STRONG NONE* STRONG 0.47
1950-64 MODERATE NONE STRONG 0.43
1967-79 WEAK WEAK STRONG 0.30
1979-85 MODERATE NONE* WEAK 0.28

* Decrease.

5.61 In all, Brazilian rent control legislation might be considered

restrainedwhen contrastedwith those of other developingcountries. Table V-8
shows that the seventy years of rent-related legislation were dominated by
weak/none (54 percent) and moderate (33 percent) controls. The effects of
Brazilian rent control policy on long run housing supply still have to be
assessed empirically. Yet, impressionisticevidence suggests that even mild
forms of rent control, when combined with other market failures,might have a
strong and negative impact on rental housing supply.

5.62 We may derive two broad statementsfrom Table 5.8: (i) the absence of
rent controls (or the presence of weak controls) does not tend to be associated
with rises in real rental prices, especially in a weak economy; (ii) in the
long-run, rent control regimes (even non-stringentones) tend to affect real
rental prices and the share of rentals in total housing stock.

5.63 A simple statisticalanalysis of rental units offered for rent in the

city of Rio de Janeiro indicates that the supply of rental housing is rather
responsive to price, with elasticities (for one-bedroom apartments) ranging
between 2.7 in the high-income area of the city and 3.6 in the low-incomearea.
This suggests that rent controls, when strictly enforced, could have a
significant impact on the supply of rental housing.
 - 78 -
Fi2ure 5.10: REAL RENTAL PRICES AND SHARE OF RENTALS

1dci ofReal Rental Pilesa 1 Rio _ s

el 10Ev lars

2ffi@. ...........................................................- ............

1$0 ., .. ,,..,,.,,,,................................................ ..........

us~~~~~~~~~~~~~~~~~~~~~~~~s

i\..._............I .............
- ..,---.... .'..,,,,,,,,,,,...
..............
,,,,
'
lo ...... _- ...

1610 660 ISI 164 1650 1960 1676 ISIS 1505 e"i"""i"""""" """""^i"
yellt SI 506 945 664111M0 116
15*1 ISiS 550 10 6 151 1570 608 600
_ ~~~~~~~~~Yells
I- m. -- li -.-- ,I.1. _"|mt
-1.-0f HI_Rt+| fiaSemeJ
Lo Illega a
SJW.1 - -i-- -o-lCII.

la j ll4tll:,ttl
WIolest 1E, I 1.6,1 . l- It.ItstIht,s,t..
to.&hahr
s,e.amis, ISI.. hu.gragbis M PlAN,m,,elai
uiesao
IIS,,,, 1.66cr,.:555 . C01500. jat.

Alternative Decontrol Options

5.64 Among other cases studied (see above), alternatives for further
relaxation of controls were analyzed by running the model under four basic
alternative assumptions of decontrol. In the first instance, we assumed no
change in the rent control system; in the second, all controls are lifted, and
in the third and fourth options, we allow for different degrees of gradual
decontrol.

5.65 Do Nothing. This is the baseline case. Yearly increasesare permitted

up to 90 percent of general price inflation. We studied market-level
revaluationsallowed every five years and every ten years. It has already been
noted that while such controlsdon't reduce much profitabilityfor the "average"
unit if tenure security is limited to lease Reriods and conversion is
unregulated, they might reduce profitabilitymuch more for the marginal unit.
If so, analysis of average units wiLllunderstateeffects on supply.

5.66 Blanket Lifting. Conceptually,this is the simplestdecontroloption.

In this option, we are really concerned with the changes in rents for existing
"controlled"units rather than for new or "uncontrolled"units. If a household
is given a choice between remainingin an existingunit and moving to a new unit,
 - 79 -

Box 5.1: MARGINAL CHANGES,MAJOR EFFECTS

Pronosal Effect

Keep the rent control legislation Independentof whether or not the

constant. legislation is bad, keeping it
constantwill at least reduce risk
to landlords and allow long-term
planning and budgeting. Frequent
changes may increasediscount
rates.

Link indexed increases in rent to Improvesaffordabilityand reduces

the political resistanceto full
wage index indexationof rents.

Allow tenants to borrow from Improves affordabilityand allows

commercialbanks using their landlords to recover larger
deposits in the FGTS (see Annex A households
for a description of first-year
markups from FGTS) as collateral
households, thus improving low-
income rentals supply position.

Make the five year adjustmentin Improves affordabilityand reduces

rentals an administrativeprocess the risk to landlords and overall
rather than a judicial one. In its costs of administeringthe rent
simplest form, this would involve control legislation.
the periodic publication of a rental
housing prices index by size of
unit, microlocationand other
factors (see Chapter III subsection
on Hedonic indexes).

Reduce the adjustmentperiod from Reduces risk to landlords and

five to three years. lowers the first-yearmarkup thus
improving affordabilityand low-
rentals supply position.

however expensive, it cannot be made worse-off because it has the option to

remain. But it can be made worse-off if rents rise for its current unit.

5.67 Blanket decontrol, where all controls are lifted at one time, is
administrativelythe simplestdecontrol option. While data from 1980 show that
most controlledhousing had rents not too far removed from market price, over
time this has created some dispersion in PmQc/PcQc. Some rents in Rio have
probably fallen so far behind market values that rises could result in major
dislocations. Arnott (1981) indicatesthat the greater excess demand there is
in a market, the greater will be the disruption caused by blanket decontrol.
 - 80 -

Disruption under this alternative

could be large, especially if other Box 5.2: CLOSED DOORS
housing market imperfections such as The folloving passage appeared on "VeJa,I March 7,
land and zoning regulations initially 1990, pp. 78-90:
impede the supply response.
iIt is increasingly difficult for renters to live
in our major cities. And they form an army of
5.68 During any period of change, millions--only in Sao Paulo there are 4.5 million
some households will inevitably incur renters, in Rio de Janeiro some 3 million. The
higher rents and moving costs. While problem is not lack of physical
problem is to find owners
structures.
willlng
The
to rent them
rents increasing from less than 15 out.
percent of income may not seem
extraordinary to an outside observer, One of the consequences of the rent control law
and its subsequent distortions is the creation of
especially when low-income households ghost towns. In Sao Paulo, it is estimated that
typically spend large fractions of over 15,000 units are presently kept empty despite
their income for housing elsewhere, a tremendous shortage of housing. According to
Mr. Chindler, a realtor in Rio de Janeiro, for
the change from the current situation each ten apartments vacating today only one will
is substantial. One way to cushion find its way back to the rental market.
the blow and ensure political Realtors in the city of Curitiba (population 1
sustainability of decontrol is to million) had only 110 residentialproperties for
replace controls with better targeted rent last week. According to one estimate, it
housing subsidies for the poor. This would be necessary at least 2,000 such properties
to satisfy demand. The two largest real estate
is the approach that was used to relax agents in Recife (population 3 million) did not
postwar European controls. But large- have one single unit available. And one agency in
scale subsidy schemes are probably not Salvador (population 2.5 million)
for each rental advertisement it places, no less
discovered that
administratively or budgetarily than 600 inquiries are received."
feasible in Brazil at this time.

5.69 Decontrol New Construction

and Vacancies. Completely freeing rents for newly constructed units for all time
can only increase supply. As noted, if a household is given a choice between
remaining in an existingunit and moving to an expensive new unit, it can't be
made worse-off because it has the option to remain.

5.70 Permanent exemptionof new units from rent controls is most attractive
to the authors in the short-term on the grounds that it encourages new
construction. Similar exemption could be applied to improvement of existing
housing to encourageimprovementsin conditionswithout adverselyaffectingmany
existing low-incometenancies.
 - 81 -

5.71 Extending permanent exemptions to new vacancies57/is another action

which would not introduce displacementor weigh heavily on the budgets of low-
income households. After all, for the first year of the lease this corresponds
to the same situation under the present regime. Risks, and hence discount
rates, could actually decline.

5.72 Floating Up and Out. The most effective method for encouraging new
investmentwhile protectinglow-incomerenters may involvea combinationof full
indexationof increaseswith a "floating up and out" of controls. The latter
involves the transitionfrom controlledrents to market rents over a relatively
short period. Indexationhas been used in Brazil in the recent past and provides
a formula for determiningthe intermediaterent levels. For example,rents could
be increasedannuallyby, say, the ConsumerPrice Index plus a percentageof the
previous year's rent until the unit becomes vacant or until the revaluation
system kicks off (each five years). This phasing would smooth the path of
adjustmentgiving tenants who could not afford their current room at the market
rent some time to find suitable alternatives.

5.73 Since in Brazil the controlledsector is alreadyclose to what it might

be in the absence of controls, phasing-inwill make less difference than would
be the case under a rigid rent control regime.

.aJ. This is in contrastto the system of temporaryexemptions(i.e., that allows

rents on new leases to be set by market conditionswith any subsequentincreases
controlledby legislation)which is presentlyavailablein Brazil. These systems
result in several perverse incentives. Landlords have incentives to
undermaintainunits or even harass tenants in order to reclaim the unit and
increasetheir rental income. Tenants have incentivesto avoid moving to units
more in line with their current needs because they would give up existing rent
discounts and forego the often large initialmarkup payment. Such systems have
the potential to reduce mobility and decrease the efficiency of use of the
existing stock. In the Brazilian context, however, market flexibility is
introducedby the common and legal practiceof landlordspaying tenants to vacate
(i.e., contractingout).
 REFERENCES

Albuquerque, Marcos C.C. de. "Habitacao Popular: Avaliacao e Propostas de

Reformulacaodo Sistema Financeiroda Habitacao," Estudos Economicos,16
(1), 1986, pp. 77-121.

Arnault, Jane. OptimalMaintenanceUnder Rent Control with Quality Constraints.

AREUEA Journal, Summer 1975, pp. 67-82.

Arnott, Richard,with Nigel Johnston. Rent Control and Options for Decontrol in
Ontario. Ontario Economic Council, 1981.

Aschauer, David. Is Public ExpenditureProductive?. Federal Reserve Bank of

Chicago Staff Memorandum, 1989.

Azevedo, S. and L.A.G. de Andrade. Habitacao e Poder. Rio de Janeiro: Zahar,

1982.

Banco Nacional da Habitacao - Assessoria Tecnica da Presidencia. Mercado

Imobiliario--AEvidente Realidade Brasileira (Subsidiospara Sua Melhor
Compreensao). Rio de Janeiro, July 1982.

- Centro de PesquisasHabitacionais. Inguilinatoe Despeios. Rio de

Janeiro, 1971.

_ Locacao de Imoveis ResidenciaisAntes e DeRois da Lei 4494/64-GB.

Rio de Janeiro, 1974.

Belsley, David A., Edwin Kuh and Roy E. Welsch. Regression Diagnostics:
IdentifyingInfluentialData and Resourcesof Collinearity. Wiley, 1980.

Bender, Bruce. The Determinantsof Housing Demolitionand Abandonment.Southern

EconomicJournal, 46, July 1979, pp. 131-144.

Blay, Eva Alterman. "Dormitoriose Vilas Operarias: 0 Trabalhador no Espaco

Urbano Brasileiro." In Habitacao em Ouestao. Ed. L.P. Valladares.Rio
de Janeiro: Zahar, 1980.

Boersch-Supan,Axel. On Tenure Discounts and Rent Control.Mimeographed,1983.

Bolaffi, Gabriel. "A Casa das Ilusoes Perdidas: Aspectos Socio-Economicosdo

Plano Nacional da Habitacao," in Caderno CEBRAP 27. Sao Paulo:
Brasiliense,1977.

Bonduki, Nabil G. "Origens do Problema da Habitacao Popular em Sao Paulo."

Espaco e Debates-- Revista de Estudos Regionais e Urbanos, Ano 2, no. 5,
1982, pp. 81-111.

. "Habitacao Popular: Contribuicao para o Estudo da Evolucao Urbana

de Sao Paulo.I In Regensando a Habitacao no Brasil. Colecao Debates
Urbanos. Ed. L.P. Valladares. Rio de Janeiro: Zahar, 1983.
 - 83 -

Brealey, Richard and Stewart Myers. Princigals of CorRorate Finance. McGraw

Hill, 1981.

Brueggeman,William. Federal Rental Housing Production Incentives: Effect on

Rents and InvestorReturns. U.S. GovernmentAccounting Office, 1985.

Butler, Richard V. The Specificationof Hedonic IndexesforUrban Housing.Land

Economics, 58(1), February 1L982,pp. 96-108.

Clark, W.A.V., and Allan D. Heskin. The Impact of Rent Control on Tenure
Discountsand ResidentialMolbility.Land Economics,58(1), February1982,
pp. 109-117.

De Leeuw, Frank and Larry Ozanne. Housing. H. Aaron and J. Pechman (eds.),How
Taxes Affect Economic Behavior, Brookings,1981.

Dildine, Larry L. and Fred A. Massey. Dynamic Model of Private Incentives to

Housing Maintenance. Southern EconomicJournal, 1974, pp. 631-639.

Dillinger, William. Urban Property Taxation in Developing Countries. World

Bank, PPR Working Paper No. WPS 41, 1988.

Fallis, George and Lawrence B. Smith. "UncontrolledPrices in a Controlled

Market: The Case of Rent Controls." American Economic Review, 74(1),
March 1984, pp. 193-201.

Farah, Martha F.S. "Estado e Habitacao no Brasil: 0 Caso dos Institutos de

Previdencia," EsRaco e Debates--Revista de Estudos Regionais e Urbanos,
Ano 5, no. 16, 1985, pp. 73-83.

Ferchiou,Ridha. The IndirectEffects of New Housing Constructionin Developing

Countries. Urban Studies, 19, 1982, pp. 167-76.

FINEP. A Ouestao HabitacionalAtraves da Historia. 1984.

Follain, James R. and Stephen Malpezzi. Dissecting Housing Value and Rent:
Estimatesof Hedonic Indexes for Thirty-NineLarge SMSAs. Washington,D.
C.: The Urban Institute,1980.

Follain,James R.. The Price Elasticityof the Long Run Supply of New Housing
Construction. Land Economics,55, 1979, pp. 190-99.

Fundacao Joao Pinheiro, Centro de Economia Aplicada. A Ouestao dos Precos de

Alugueis no Brasil--Umexame Preliminar. Belo Horizonte,1984.

FundacaoInstituto Brasileirode Geografiae Estatisticas- IBGE. Estatisticas

Historicas do Brasil. Rio de Janeiro, 1987.

Gittinger,J. Price. EconomicAnalysisof AgriculturalProiects. Johns Hopkins

University Press, 1982.
 - 84 -

Gyourko,Joseph and Peter Linneman. MeasurementProblemsin the Distributionof

Subsidies.

Hamilton, Rabinovitz, Szaanton and Alshuler; and the Urban Institute. The Los
Angeles Rent StabilizationSystem: Impactsand Alternatives. April 1985.

Ingram, Gregory K.. Land in Perspective:Its Role in the Structure of Cities.

M. Cullen and S. Woolrey (eds.), World Congress on Land Policy. 1980
(Lexington,Mass,: DC Heath 1982).

Johnson, Thomas E. Jr. Upward Filtering of the Housing Stock. Habitat

International,11(1), 1987, pp. 173-90.

Johnson, D. Gale. "Rent Control and the Distribution of Income, American

EconomicReview, (41) 1951.

Kahnert, Friedrich. Improving Urban Employmentand Labor Productivity. World

Bank DiscussionPaper No. 10, 1987.

Kiefer, David. Housing Deterioration,Housing Codes and Rent Control.Urban

Studies, February 1980, 53-62.

Kravis, Irving B., Alan Heston and Robert Summers. World Product and Income:
InternationalComparisonsof Real Gross Product. Johns HopkinsUniversity
Press, 1982.

Lemer, Andrew. The Role of Rental Housing in DevelopingCountries: A Need for

Balance. World Bank Urban DevelopmentDiscussionPaper No. 104, January
1987.

Lima, Luiz Eduardo Pinto. 'The Financingof Housing in Brazil,"Housing Finance

International. February 1987, pp. 16-21.

Linneman, Peter. "The Distribution of Tenant Benefits from Rent

Control."Journalof Urban Economics

Maclennan, Duncan. The 1974 Rent Act--Some Short Run Supply Effects. The
EconomicJournal, 88 (June 1978), 331-340.

Maddala, G. S. Limited-Dependentand Qualitative Variables in Econometrics.

CambridgeUniversity Press, Cambridge,1983.

Malpezzi, Stephen. Analyzing Incentives in Housing Programs: EvaluatingCosts

and Benefits with a Present Value Model. INU Discussion Paper No. 23,
1988.

Malpezzi, Stephen and Vinod Tewari. Costs and Benefits of Rent Regulation in
Banzalore. India. INURD, Mimeo, 1990.

Malpezzi, Stephen. Present Value Analysis of Housing Programs and Policies.

A.G. Tipple and K.G. Willis (eds.), Housing the Poor in the Developing
World: Methods of Analysis. Case Studies and Policies, Routledge
Publishing, Forthcoming.

Malpezzi, Stephen and Gwendolyn Ball. Rent Control in DeveloDingCountries:A

Synthesis. Mimeographed,1990.
 - 85 -

Malpezzi, Stephen and Peter Rydell. Rent Control in Developing Countries: A

Frameworkfor Analysis. World Bank Urban DevelopmentDiscussionPaper No.
102, September 1986.

Malpezzi, Stephen and Stephen K. Mayo. "The Demand for Housing in Developing
Countries." Economic Development and Cultural Change, July 1987.

Malpezzi, Stephen,Larry Ozanne and Thomas Thibodeau. CharacteristicPrices of

Housing in 59 SMSAs. Urban InstituteContract Report 1367-1, 1981.

Malpezzi, Stephen, Stephen K. Mayo and Ricardo Silveira. Measuring the Costs
and Benefits of Rent ControL: Case Study Design. INU, 1988.

Malpezzi, Stephen and Stephen K. Mayo, with David J. Gross. Housing Demand in
DevelopingCountries. World Bank Staff Working Paper No. 733, 1985.

Malpezzi, Stephen. "An Introduction to Regression Diagnostics and Robust

Estimation." Processed, 1985.

Malpezzi, Stephen. Rent Control and Housing Market Eguilibrium: Theory and
Evidence from Cairo. Evt. Ph.D. Dissertation, George Washington
University, 1986.

Malpezzi, Stephen. Urban Housing and Financial Markets: An International

Comparison. Paper prepared for the Rowntree Trust conference, York,
England, June 1989.

Malpezzi, Stephen. Rental Housing in Developing Countries: Issues and

Constraints. Paper presented to the UNCHS Expert Group Meeting on Rental
Housing in DevelopingCountries,October 1989. Revised March 1990.

Malpezzi, Stephen,A. Graham Tipple, and Kenneth G. Willis. Costs and Benefits
of Rent Control in Kumasi. Ghana. World Bank Discussion Paper No. 74,
1990.

Mathur, G.C. "Strategiesfor Provision of Shelter for the Urban Poor."

Mayo, Stephen K, StephenMalpezzi, and David J. Gross. "ShelterStrategiesfor

the Urban Poor in Developing Countries." World Bank Research Observer,
Vol. 1, July 1986.

Mayo, StephenK. "Theoryand Estimationin Economicsof Housing Demand."Journal

of Urban Economics,July 1981.

Mayo, Stephen K. and David J. Gross. Sites and Services -- And Subsidies:The
Economics of Low Cost Housing in Developing Countries. World Bank
Economic Review, 1(2), 1987, pp. 301-35.

Mishan, E.J.. Cost Benefit Analysis. George Allen and Unwin, 1982.

Moorehouse,John C. "OptimalHousing MaintenanceUnder Rent Control," Southern

Economic Journal, 1972 (39).
 - 86 -

Muth, Richard F.. The Demand for Non-Farm Housing. Arnold Harberger, The
Demand for Durable Goods, University of Chicago Press, 1960.

Muth, Richard F.. Cities and Housing. University of Chicago Press, 1969.

Olsen, Edgar 0. "An EconometricAnalysis of Rent Control,"Journal of Political

Economy, 1972.

Olsen, Edgar O.. A CompetitiveTheory of the Housing Market. AmericanEconomic

Review, 1969, pp. 612-622.

Olsen, Edgar O.. The Effects of a Simple Rent Control Scheme in a Competitive
Market. Rand Working Paper P-4257, 1969.

Olsen, Edgar O.. What Do EconomistsKnow About Rent Control?. The Journal of
Real Estate Finance and Economics,1(3), November 1988, pp. 295-308.

Olsen, Randall J. "A Least Squares Correction for Selectivity Bias,'

Econometrica,(48) November 1980.

Ozanne, Larry and Stephen Malpezzi. "The Efficiencyof Hedonic Estimationwith

the Annual Housing Survey: Evidence from the Demand Experiment."Journal
of Economic and Social Measurement,1986.

Paul, Samuel. "Housing Policy: A Case of Subsidizingthe Rich?," Economicand

Political Weeklyv,September 23, 1972.

Pena, Daniel, and Javier Ruiz-Castillo. DistributionalAspects of Public Rental

Housing and Rent Control Policies in Spain. Journal of Urban Economics,
15, 1984, pp. 350-370.

Pessoa, Alvaro, ed. Direito do Urbanismo--Umavisao Socio-Juridica. Rio de

Janeiro: Livros Tecnicos e Cientificos,1981.

Renaud, Bertrand. AffordableHousing. Housing Sector Performanceand Behavior

of the Price-to-Income Ratio: International Evidence and Theoretical
Analysis. Paper Presented to the Center of Urban Studies and Urban
Planning, University of Hong Kong, October 1989.

Ribeiro, Luis Cesar de Queiroz. "Formacaodo Capital Imobiliarioe a Producao

do Espaco Construido no Rio de Janeiro--1870/1930," Espaco e
Debates--Revistade Estudos Regionais e Urbanos, Ano 5, no. 15, 1985,
pp. 5-32.

Rolnick, Raquel. "De Como Sao Paulo Virou a Capital do Capital," in Repensando
a Habitacao no Brasil. Debates Urbanos. Ed. L.P. Valladares. Rio de
Janeiro: Zahar, 1983.

Rydell, C. Peter, and Kevin Neels. Rent Control,Under-maintenance,andHousing

Deterioration. Mimeographed,1982.
 - 87 -

Rydell, C.P., C.L. Barnett, C.E. Hillestad, M.P. Murray, K. Neels, and R. H.
Sims. The Impact of Rent Control on the Los Angeles Housing Market. Rand
Note N-1747-LA,August 1981.

Smith, Barton . The Supply of Urban Housing. Quarterly Journal of Economics,

1976.

Struyk, Raymond J. "The Distributionof the Tenant Benefits from Rent Control
in Urban Jordan." Land Economics,1988.

Sandilands,Roger James, Monetary Correction and Housing Finance in Colombia.

Brazil and Chile. Sarnborough,England: Gower Publisher, 1980.

United Nations, Review of Rent Control in DevelopingCountries. Department of

Economics and Social Affairs, ST/ESA/85,1979.

United Nations. Review of Rent Control in DevelopingCountries.Departmentof

Economic and Social Affairs, ST/ESA/85,1979.

United Nations. The Prosiects of World Urbanization. U.N. Publication

ST/ESA/SER.A/101,1987.

Urban Edge. The. Making Shelter Projects Replicable. World Bank, 9(10),
December 1985.

Vitaliano, Donald F.. The Short:Run Supply of Housing Services Under Rent
Control. Urban Studies, 22, 1985.
 - 88 -

ANNEX A

A SURVEY OF HOUSING INSTITUTIONSIN BRAZIL

A. Early Housing Institutions:1889-1930

The Brazilian republicanperiod begins with the fall of the empire in

1889, just one year after the abolition of slavery. It was a time of growing
political demands by an emergingurban constituency,and housing policy was one
of the first concerted attempts by the governmentto address this new political
force. The housing sector during this period was purely market-oriented,with
nationalhousing policy constrainedby the overall limited role of the state.

In 1920, 'Law-Decree'(Decreto-lei)DL 4209 allowed the governmentto

build housing projects to be rented to wage laborers, while DL 14813 of 1921
regulated fiscal benefits to private individualsinvesting in rental housing.
In 1923, DL 4682 created the first 'Pensionand Retirement Fund' in the nation
(for railroad workers), setting a milestone in the development of a social
security system in Brazil, and later providing a source of financing to the
housing sector.

The private rental sector was crucial in the provision of housing to

the fast-growingurban populationsof Rio de Janeiro and Sao Paulo (whererentals
accounted for about 70 percent of all housing in the 1920s). Two basic rental
systems attended to the low-income population: the collective dwellings,
popularly known as corticos; and working class villages, known as vilas-
operarias. Corticoswere rented to the poorest segment of the working class, and
were generally overcrowdedand unsanitary. Vilas operarias tended to house the
more trained portion of the working class; althoughmiddle-incomeprofessionals
often rented there as well. These accommodationsgenerated labor dependencyon
the employer, and contributed to the labor disciplinerequiredby the emerging
factory system. Moreover, judging from the number of similar dwellings
reproducedby private investorswithout industrialaffiliation,these villages
seem to have been highly profitable.

Although the governmentof the Old Republicrarely interfereddirectly

in the housing sector, it had some impact on the quantity and quality of the
housing stock through its regulatory functions. The early Brazilian urbs was
apparentlyquite an unhealthy place in which to live, with a number of serious
epidemicson record. At first, local governmentsattemptedto address the health
problem through tax incentiveschemes such as the Sao Paulo municipalLaw No.1098
of 1908, which proposed tax exemptions for structuresto be rented or sold to
low-incomehouseholdsif built within acceptablenorms of hygiene. Most housing
regulationswere legislated to address health problems, as for instanceRio de
Janeiro's Law No.391 of 1903 which at the time ruled the construction and
refurbishmentof buildings. Many of these individualhealth-relateddirectives
were slowly broadened, and later incorporatedinto comprehensiveconstruction
codes; those of Sao Paulo in 1920 are the earliest on record.
 - 89 -

B. PopulistHousing Institutions:1930-64

After the revolutionof 1930, the first institutionalchange to affect

the housing sector was the creation of the 'Pensionand Retirement Institutes'
(IAPs). Although the use of a portion of these pension funds to financehousing
had been authorized by DL 20465 in 1931, only 118 residentialunits had been
financed through this mechanism by 1937. That same year, DL 1749 altered the
conditions of financing through these funds (reducing interest rates and
extendingthe amortizationperiod) and increasedthe share of IAPs' funds devoted
to housing financing to 50 percent. Still, only 279 developments (benefitting
47,789 households) were financed by the IAPs between 1937-64. Nearly three-
fourths of housing financingby the IAPs took place between 1937-50;after that,
a large share of the funds began to be used to financepublic megaprojects,such
as the constructionof Brasilia (Farah, 1985).

PresidentDutra introducedanother institutionto complementthe IAPs.

In 1946, DL 9218 created the 'PopularHousing Foundation'(FCP). While IAPs were
intended to finance housing for the middle class, the FCP sought to extend
housing loans at subsidized terms to all income classes, with municipalities
contributingin the pro-rision of land and urban infrastructurefor the low-income
segment of the portfolio. Insufficientability to raise funds, and large loan
subsidies, ensured a poor performance in FCP's attempt to reduce the formal
housing deficit.

The posture of populistgovernmentstowards informalhousing generally

favored the completerelocationof slum dwellersto government-builtor -financed
structures. Unfortunately,all low-incomehousing programs of the populistera
eventuallyfaced financialcollapse and were only able to reach a small segment
of the population they were created to serve.

By late 1950s and early 1960s, it had become increasinglydifficult

for the government to raise funds as a result of high inflationand usury laws
which imposed a 12 percent interest ceiling on all loans.

"Ironically,the usury laws...affectedthe financingof the

government sector more than the private sector since the
governmentwas not preparedto resort to the various devices
such as special commissions, discounts and illegal side-
payments used by the private sector to circumvent these
laws. .. 581

Inability to borrow, and the incessantneed to maintain popular support through

heavily subsidizedprograms,forced the governmentto accommodateby expansionary
monetary policy, thus fueling even higher inflationand making it even harder to
borrow.

18 Roger James Sandilands,MonetaryCorrectionand HousingFinance in Colombia.

Brazil and Chile (Sarnborough,England: Gower Publisher,1980), p. 114.
 - 90 -

Table A-1: THE DISTRIBUTIONOF CREDIT IN BRAZIL

(Billionsof 1974 Cruzeiros)

Short-Term Credit to Long-Term Credit to

Private Public Private Public
Year Sector Sector Total Sector Sector Total

1952 43.21 1.32 44.53 5.69 6.11 11.80

1953 40.84 1.66 42.50 5.49 5.80 11.29
1954 41.74 1.50 43.24 5.38 6.53 11.91
1955 40.90 1.08 41.98 5.79 6.44 12.23
1956 38.68 0.80 39.48 5.71 5.47 11.18
1957 45.40 0.90 46.30 7.51 5.65 13.16
1958 42.51 0.63 43.14 7.07 5.59 12.66
1959 38.58 0.73 39.31 6.65 5.10 11.75
1960 40.84 0.94 41.78 6.45 4.48 10.93
1961 36.56 1.79 37.35 5.36 3.20 8.86
1962 40.14 0.57 40.71 6.45 3.73 10.18
1963 33.78 0.62 34.40 5.65 2.65 8.30
1964 33.26 0.83 34.09 5.16 1.64 6.80

1974 226.21 2.90 229.11 88.85 13.57 102.42

The private sector did not escape the credit crunch unscathed. Even
though it was relatively easy to circumvent the usury law with short-term,
unsecured credit, it was nearly impossible to do so with long-term loans such as
those required in housing finance. Table A-1 illustratesthe combinedimpact of
the government's usury laws and expansionarymonetary policy on the overall
credit availability.

The overall performance of the populist governments in the housing

sector was disastrous. An inflexiblerent control legislation,coupledwith the
inabilityto mobilize resourcesto financehome ownershipvirtuallyparalyzedthe
housing sector in the early 1960s:

"...between1930 and 1964, a mere 100,000housing units

had been financed [mainly through the IAPs] ... this was an
abysmal record for a country which in 1960 had 70 million
inhabitants and whose urban population was growing by about
5 percent each year."''

Towards the end of the Old Republic, there was a return to market
negotiationin rental agreements. The 1930swere characterizedby decliningreal
rents, and by generally low inflation,which allowed the laissez-faireattitude
towards rental housing to continue.

.i/ Roger James Sandilands, Ibid, p. 115.

 -91-

Table A-2: INDEX OF LICENSES TO CONSTRUCT Table A-3: INDEX OF CONSTRUCTION

IN RIO DE JANEIRO: 1935-63 COSTS: 1944-64
(1963 - 100) (1963 - 100)

Construction Constrruction
Number of Area Year Cost Index Cost Index
Year Licenses Licensed (Real)

1935 50 1944 2.8 138.2

1936 68 - 1945 3.1 139.3
1937 - 26 1946 3.6 132.4
1938 - 23 1947 4.0 144.2
1939 - 25 1948 4.2 137.8
1940-47 - - 1949 4.7 137.3
1948 105 36 1950 4.8 124.4
1949 120 37 1951 5.4 124.8
1950 176 50 1352 5.8 118.7
1951 160 77 193\ 6.4 110.5
1952 135 88 1954 8.5 115.8
1953 98 80 1955 9.2 112.3
1954 96 68 1956 11.8 115.1
1955 122 48 1957 12.6 114.8
1956 133 62 1958 16.1 118.3
1957 113 77 1959 19.5 102.4
1958 123 81 1960 27.4 110.6
1959 117 83 1961 39.3 107.2
1960 126 75 1962 60.9 109.5
1961 138 77 1963 100.0 100.0
1962 124 102 1964 204.1 106.2
1963 100 100
_____________ _______
_______ ______Source: IBGE, 1987.
Source: IBGE, several years.

The government had hoped that an

infusion of funds into the housing market via the IAPs would help reduce the role
of rental housing, which was seen as inferior to homeownership, and as a breeding
ground for social conflicts and health epidemics. The government's attitude
toward rental housing is also evident in its urban plans: urban reconstruction
and growth contributed to the decline of the inner city corticos,601 while the
recently improved public transportation system allowed the expansion of the city
into land more affordable for ownership in the urban periphery.

C. Modern Housing Institutions: 1964-Present

A coup in 1964 ended nearly 20 years of democratic institutions and

brought to power a series of military governments. Housing was one of the major
social problems inherited by the military government. Although we do not have
estimates of the formal housing deficit in the populist years, we can assume it
was sizable, given the fast expansion of slums in the urban peripheries. The
housing deficit continued to increase in the 1960s, despite a serious effort made

60/ Except for the city of Sao Paulo where corticos continue to this date to be
an important component of urban housing, sheltering as much as 40 percent of the
population.
 - 92 -

to mobilize resources. The deficit grew from about 945,000 dwellings in 1960 to
1.9 million in 1964, and to 2.5 million in 1970 (Souza, 1978). Crude estimates
of the housing deficit in 1986 range between 4 and 6 million units.

In August 1964, just a few months after the military coup, the
National Housing Bank (BNH) was created by DL 4380, with the responsibilityof
executing the National Housing Plan (PNH) and managing the National Housing
System (SFH).

The BNH, established with an initial capital of 1 million cruzeiros

(US$910,000),was originallyexpected to be funded mostly by a 1 percent tax on
wages. Later, in the second half of 1966, the 'Lengthof ServiceGuaranteeFund'
(FGTS)was introducedto expand the pool of funds availablefor housing programs.
The FGTS is a scheme whereby workers contribute8 percent of their wages into an
interest-earning(3 percent), inflation-indexedfund with quarterly monetary
corrections,from which they can withdraw either at retirement, during periods
of unemploymentor illness,or for the purchase of a house.

Additional funding for housing financing could also be obtained

directly from the federal budget and from foreign borrowing, as well as from
profits of the system and voluntary savings in one of the institutionscomprising
the 'BrazilianSystem for Savings and Loans' (SBPE). The SBPE is a sub-system
of the SFH composedof privately-ownedSavingsand Loan Associations,Real Estate
Credit Companies, and the Federal and State Savings Banks. The SBPE was
controlledby BNH, which also determinedthe financialnorms of operationsof all
the members of the system.

When the FGTS was created, its deposits at BNH accounted for nearly
70 percent of total savings in the Housing Finance System. In the 1970s,
however,savings passbooks (offeringa 6 percent returnplus monetarycorrection)
became increasinglypopular as the common man's hedge against inflation,and
quickly surpassed FGTS in importance. By 1973, voluntary savings had already
surpassed FGTS as the major source of funding of the SFH, with 42 percent of
total savings, while FGTS and real estate bonds (issuedby Real Estate Credit
Companies and BNH) accounted for 39 and 19 percent, respectively.

BNH was61 1 a second-linebank and actual housing loans to the public

were made through one of the SFH's many financialagents. Each agent in the SFH
specializes,to some extent, in a certain class of housing. Except for the
Federal and State Savings Banks, SBPE finances loans mostly to middle-class
households;whereas the housing companiesknown as COHABS (partlyowned by state
and local governments),and the mutual aid associationsknown as COOPHABS,target
the lower-incomepopulation. BNH also financedsanitationprojects through the
Financial System of Sanitation (SFS), urban development through the Financial
System for Urban Development (SFDU),and purchase or productionof construction
materials through complementaryprograms known as FIMACO and RECON.

i1/ BNH was closed in 1986.

 - 93 -

Table A-4: HOUSING FINANCING THROUGH THE SHF

BY PROGRAM: 1964-86
('000 Units)

Other /a Other lb
Year COHABS COOPS Housing SBPE Recon Programs Total

1964-67 76.4 11.8 20.0 31.4 0.4 0.6 140.7

1968 44.6 15.3 29.7 47.1 2.6 - 139.4
1969 57.7 26.4 21.7 41.9 10.3 - 158.2
1970 21.8 29.0 21.0 72.0 10.2 0.2 154.2
1971 18.0 25.2 16.6 50.5 10.8 - 121.2
1972 12.0 19.7 5.3 66.7 15.4 1.6 120.1
1973 17.8 22.1 9.7 76.6 22.2 - 148.4
1974 7.8 4.9 3.5 60.3 19.0 1.0 96.3
1975 45.4 11.2 5.4 64.5 13.7 7.1 147.3
1976 83.0 33.9 7.2 82.2 18.2 17.9 242.5
1977 92.7 34.0 9.1 58.0 17.4 7.2 218.6
1978 192.4 35.8 14.8 58.1 19.3 3.3 323.8
1979 139.8 69.7 35.7 109.8 22.2 3.5 380.7
1980 115.2 44.8 113.6 268.7 20.2 - 562.4
1981 80.0 39.3 87.1 231.0 13.3 0.2 450.9
1982 125.3 34.9 12.6 249.8 19.7 - 555.8
1983 16.5 14.0 20.5 121.8 9.1 - 181.8
1984 63.7 15.5 40.2 96.7 1.0 - 214.0
1985 277.0
1986lc 50.6

/I Includes Mortgage Market, Prohasp, Ficam, Prosindi, Joao de Barro,

Promorar.
lb Includes Profico and Prodepo.
/a First Semester
Source: Albuquerque, 1986 and BNH estimates.

SFH policy towards low-incomehousing provision may be divided into

four distinctphases. During the establishmentphase (1964-69),over 40 percent
of the units financed (178,227)were targeted at low-incomehouseholds (up to
three minimum wages). Azevedo and Andrade (1982) describe this policy as an
attemptby the military governmentto legitimizethe revolutionbefore the urban
population,which had become politicallyorganizedduring the populistyears and
now constituteda potential focus of political conflict.

As the revolutiontook roots, and more importantly,as the number of

loan defaults began to rise, the consolidated portfolio of the SF1Hshifted
towards the more profitableand less risky middle class market. The period 1970-
74 is characterizedby the loss of dynamism of the COHABs, with the number of
low-incomeunits financed reduced to 77,309 (12 percent of the total).

In the third phase (1975-79), with COHABs now allowed to finance

houses for clients earning up to five minimum wages, delinquency began to
improve,and the total number of COHAB units rose to 551,203--over40 percent of
the total number of low-incomehousing built in the entire period 1964-85.
 - 94 -

Between 1980-85, a slow down in low-income housing financing took

place once again and COHABs' share of total house loans decreased to about 20
percent.

Table A-5: INCOME TO MORTGAGERATIOS IN LOW-INCOME HOUSING

BY CLASS OF INCOME: 1965-85

Value of unit
(in Minimum Minimum Wages
Year Wages) Two Three Four Five

1965 33.84 10.5 7.0 5.2 4.2

1966 - - - - -
1967 63.65 18.0 12.0 9.0 7.2
1968 89.04 25.0 16.7 12.5 10.0
1969 67.10 19.0 12.7 9.5 7.6
1970 40.79 12.5 8.3 6.3 5.0
1971 79.37 23.0 15.3 11.5 9.2
1972 68.95 19.5 13.0 9.8 7.8
1973 41.39 12.5 8.3 6.3 5.0
1974 - - - -
1975 78.53 23.5 15.7 11.8 9.4
1976 71.92 21.5 14.3 10.8 8.6
1977 75.71 24.5 16.3 12.3 9.8
1978 49.70 14.5 9.7 7.3 5.8
1979 62.30 21.0 14.0 10.5 8.4
1980 - - - - -
1981 50.76 18.5 12.3 9.3 7.4
1982 56.32 21.5 14.3 10.8 8.6
1983 - - - - -
1984 - - - - -
1985 50.03 19.0 12.7 9.5 7.6

Source: ABECIP - BNH/DPLAC, 1985.

The staff at BNH implicitlyrecognizedthat the system was unable to

meet the needs of a substantialshare of the lowest-incomepopulation,and that
self-constructionin the informal real estate sector was compensatingfor the
housing deficit. In addition to trying to induce SBPE agents to lend to low-
income households, BNH also initiated two programs in the second half of the
1970s, in an attempt to utilize the potentialof informalhousing construction:
PROFILURB,which financed the purchase of parceled plots of land by households
earning less than three minimum wages; and PROMORAR,which introduceda shift in
policy towards slum upgrading/reconstruction, again for householdswith incomes
below three minimum wages.

Originally, the financialsoundness of the SFH was based on periodic

monetary correctionsapplied equally to assets and liabilities. However, soon
after the creationof the system, the governmentbegan to introducemodifications
in order to accommodatesalariedworkers whose incomes were adjusted only once
a year, and at a rate inferiorto the one of the 'StandardUnit of Capital' (UPC)
used to index mortgages. Between 1965-69,for instance, qualifyinglow-income
 - 95 -

households were given the option to enroll in a mortgage program, which in

essence, indexed monthly installmentsto yearly changes in the minimum wage,
while making quarterlycorrectionson the outstandingdebt accordingto the UPC.
The resulting disequilibriumwas translated into extensions of the mortgage
period of up to 1.5 times the initial contracted time, with the balance being
absorbed through the 'Fund for compensationof Salary Variation' (FCVS)--afund
generated by contributionsfrom borrowers and lenders.

In 1970, BNH created the new 'Plan for Salary Equivalence' (PES),
ending the practice of stretching mortgage periods to accommodate for the
differencebetween changes in the minimum wage and in the UPC. The number of
payment periods was fixed once again, and expectationsof inflationwere built
into the calculationof installmentpayments. Any shortfallswere to be absorbed
by the FCVS.

Between 1977-83, mortgage

payments were indexed by the UPC. At Table A-6: VARIATIONS IN MINIMUM WAGE
the end of this period, the high MONETARY CORRECTION ANDMONETARY
PAYMENTS: 1965-84
unemployment and sharp reductions in
real wages that resulted from the
1981-83recession raised the number of
loan defaults to a global maximum in YEMR WAGE CORRECTION PAYMENTS
the history of the SFH. By 1984,
about 50 percent of accounts were one- 1965 57.14 63.00 57.14
month arrears and over 20 percent were 1966 27.27 39.20 27.27
three-month arrears. As high 1967 25.00 23.23 25.40
inflation and slow wage adjustments 1968 23.43 25.00 23.43
1969 20.37 18.51 20.37
reduced the ability of borrowers to 1970 20.00 19.60 20.00
repay, real housing prices were 1971 20.56 22.67 20.51
falling, often to levels below the 1972
1973
19.15
16.07
15.30
12.84
19.15
14.70
value of their liabilities, a 1974 20.77 33.31 14.40
situation which seriously threatened 1975 41.40 24.21 34.00
the integrity of the system. The 1976
1977
44.14
44.06
37.23
30.09
26.72
36.97
government intervened once again (in 1978 41.00 36.24 30.51
July 1983) to reduce the real value of 1979 45.38 47.19 39.76
installment payments by introducing an 1980 82.96 50.77 55.06
option for installment adjustments 1983 96.20 97.76 89.03
according to the minimum wage index, 1982 100.39 156.58 130.42
later (in July 1984) with options for 1984 179.43 215.27 191.05
adjustments corresponding to 80 Source: Albuquerque, 1986.
percent of the variation in the
minimum wage, and finally (in July
1985) by allowing a 112 percent
adjustment for a period when inflationwas 246 percent. As a result of these
measures, the short-runviability of the system was restored,but the future of
the SFH was gravely compromised,as the SBPE was forced to "bank" the shortfalls
in real installmentsuntil the end of the mortgage terms, when the FCVS would
balance the accounts.

The performance of the SFH in the period 1980-85 was mixed. The
system was able to finance about as much housing as it did in all the preceding
years of its history,but the total number of units financedseems to have fallen
 - 96 -

drasticallyshort of the needs of the population. By one estimate (DEPLAC-1000,

1985), some 7 million new houses were needed between 1980-85,when only about 2
million were financedby the formal system. Moreover,while the regionalmix of
housing was very much in balance with the needs of the population, the same
cannot be said for the income mix. Despite the fact that between 1980-85 more
than 90 percent of all housing financedby BNH was geared towards low-income
housing (COHABS,LOTS, PROMORAR),only about half of the total SFH loans went to
"areas of social interest;"even though familieswith less than three and five
minimum wages represented 70 and 85 percent, respectively,of the estimated
housing need.

Table A-7: ESTIMATES OF NEED FOR NEW HOUSING

BY REGION AND CLASS OF INCOME
(1980-85, '000 Unita)

Income Class (Number of Minimum Wages)

Region Up to 1 1 to 2 3 toS 5 to 10 10+ Total

North 83 175 49 32 15 354

Northeast 866 717 150 81 46 1,860
Southeast 652 1,403 536 392 237 3,220
South 196 490 154 104 60 1,004
Center-West 163 325 85 64 40 677

Brazil 1,960 3,110 974 673 398 7,115

Source: DPLAC-1000, 1985.

In 1985, the New Republic inheritedfrom the military regimes a large

external debt and a high inflationlevel. One year after the return to civilian
government a stabilizationprogram was put into action.

Housing was one of the sectors of the economy most affectedby the so
called Cruzado StabilizationPlan. Three aspects of the Plan weakened further
the viability of the housing system: (i) mortgage payments were updated, then
frozen for a year; (ii) monetary correctionon savings accountsand on the FGTS
were maintained; and (iii) debt balances were not to be corrected for a period
of one year.

Because the volume of deposits in savings accounts is a function of

nominal interest rates (high nominal rates created the illusionthat depositors
were becoming wealthier),as inflation tumbled in the first months of the Plan,
savers withdrew their money, thus, reducing the profitabilityof the SBPE, and
severely limiting its ability to finance new houses. When inflationbegan to
accelerate again, the system was strained to its limits as the increase in
interestpayments on the liabilityaccount greatly outpacedthe mortgagepayment
receipts,which were nominally frozen. At the same time, the long-runviability,
of the SFH began to deteriorate,as the real value of the loan balances--which
could not be corrected for one year--continuouslydeclined.
 - 97 -

In an effort to salvage the system, the government intervened

primarily by closing BNH and transferringits responsibilitiesto the Federal
Savings Bank (operations)and to the Central Bank (regulatoryframework). These
measures seem to have been largely ineffectiveto improve the efficiencyof the
system. Major decisions required to restore the financial viability of the
system were postponed indefinitely. Meanwhile, adjustment in mortgage
installmentscontinue to lag behind inflation,with the problem of financing
unamortized balances still to be faced, when mortgage-terms end and the
government is forced to complementthe FCVS in order to maintain the viability
of the SBPE.

Table A-8: INDEX OF LICENSES FOR RESIDENTIAL Table A-9: INDEX OF CONSTRUCTION
CONSTRUCTION IN RIO DE JANEIRO: 1965-86 COSTS IN BRAZIL: 1964-86
(1967 - 100) (1967 - 100)

2
YEAR NUMBER AREA (M )
____________________
__ YEAR NOMINAL REAL

1965 _ 90
1966 - 86 1964 36.5 85.3
1967 100 100 1965 52.4 91.1
1968 172 348 1966 71.0 88.8
1969 99 148 1967 100.0 100.0
1970 182 292 1968 132.3 105.5
1971 72 125 1969 149.0 101.5
1972 70 190 1970 176.9 100.6
1973 80 217 1971 199.2 95.5
1974 67 202 1972 238.7 98.0
1975 68 234 1973 288.4 101.8
1976 91 320 1974 380.4 100.4
1977 64 208 1975 472.1 95.8
1978 97 286 1976 748.8 102.5
1979 117 322 1977 1,083.9 107.1
1980 162 393 1978 1,484.7 104.5
1981 166 402 1979 2,420.3 96.3
1982 163 433 1980 5,156.8 98.4
1983 145 397 1981 9,598.3 95.0
1984 63 175 1982 19,961.9 99.2
1985 42 150 1983 49,696.6 78.9
1986* 49 184 1984 155,747.5 75.9
1985 597,398.3 85.3
Source: CEHAB-RJ. 1987. 1986 1.082.872.9 94.7