See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/391904623

How does income distribution affect the impact of public investment on

private investment? Empirical evidence from Brazil

Article in EconomiA · May 2025

DOI: 10.1108/ECON-02-2025-0043

CITATIONS READS

0 8

4 authors, including:

Marina Da Silva Sanches Clara zanon Brenck

University of São Paulo University of São Paulo
32 PUBLICATIONS 86 CITATIONS 4 PUBLICATIONS 5 CITATIONS

SEE PROFILE SEE PROFILE

Gustavo Pereira Serra

São Paulo State University
34 PUBLICATIONS 51 CITATIONS

SEE PROFILE

All content following this page was uploaded by Marina Da Silva Sanches on 21 May 2025.

The user has requested enhancement of the downloaded file.

 The current issue and full text archive of this journal is available on Emerald Insight at:
https://www.emerald.com/insight/1517-7580.htm

EconomiA
How does income distribution affect
the impact of public investment on
private investment? Empirical
evidence from Brazil
Marina da Silva Sanches and Tainari Taioka Received 25 February 2025
Revised 31 March 2025
Department of Economics, University of S~ao Paulo (USP), S~ao Paulo, Brazil Accepted 30 April 2025
Clara Zanon Brenck
Department of Economics, Federal University of Minas Gerais (UFMG),
Belo Horizonte, Brazil, and
Gustavo Pereira Serra
Department of Economics, S~ao Paulo State University (UNESP), Araraquara, Brazil

Abstract
Purpose – This article aims to investigate whether the effect of public investment on private investment differs
in regimes of relatively high and low-income inequality in Brazil from 1996 to 2022.
Design/methodology/approach – The linear vector autoregressive (VAR) model was applied to estimate the
effect of public investment on private investment. To investigate the role of income distribution in this result, we
employed a threshold vector autoregressive (TVAR) model.
Findings – The results reveal that the crowding-in effect only occurs in the relatively low-income inequality
regime: a 10% increase in public investment results in a 0.8% and 3.4% increase in private investment after one
and four quarters, respectively. Conversely, the response is not statistically different from zero in the relatively
high-inequality scenario. Thus, from a macroeconomic standpoint, diminishing inequality can enhance the
responsiveness of private investment to public investment.
Originality/value – Our results underscore the substantial macroeconomic potential of policies designed to
mitigate inequalities. These policies play a pivotal role in advancing social equity and propelling more inclusive
economic growth.
Keywords Public investment, Private investment, Income inequality, TVAR, Crowding-in effect
Paper type Research paper

1. Introduction
Investment is one of the main components of aggregate demand and is especially relevant to
long-term growth. Nevertheless, there is still controversy about the role of the state both as a
direct investor and as an inducer of private investment. To fill this gap, many studies have
carried out empirical estimates to capture whether public investment crowds in or crowds out

JEL Classification — C32, E22, E25, H54

© Marina da Silva Sanches, Tainari Taioka, Clara Zanon Brenck and Gustavo Pereira Serra. Published
in EconomiA. Published by Emerald Publishing Limited. This article is published under the Creative
Commons Attribution (CC BY 4.0) licence. Anyone may reproduce, distribute, translate and create
derivative works of this article (for both commercial and non-commercial purposes), subject to full
attribution to the original publication and authors. The full terms of this licence may be seen at http://
creativecommons.org/licences/by/4.0/legalcode
The authors express their gratitude to Rodrigo Orair for making available the monthly series of public
investment. The authors also thank Gilberto Tadeu Lima, Pedro Romero Marques, Rafael Ribeiro and an
anonymous referee for helpful comments in earlier drafts of this paper.
EconomiA
Funding: This research was supported by the Research Center on Macroeconomics of Inequalities Emerald Publishing Limited
(Made), University of S~ao Paulo (FEA/USP). e-ISSN: 2358-2820
p-ISSN: 1517-7580
Disclosure statement: The authors report there are no competing interests to declare. DOI 10.1108/ECON-02-2025-0043
ECON private investment. Recent studies on the Brazilian economy (Alves & Luporini, 2008;
Sanches & Rocha, 2010; Tadeu & Silva, 2013; Dos Santos et al., 2016; Reis, Ara� ujo, &
Gonzales, 2019; Bredow, Cunha, & L�elis, 2022; Fraga & Ferreira-Filho, 2023; Iasco-Pereira
& Duregger, 2023) found a crowding-in effect of public investment on private investment.
However, the literature has shown that the magnitude of the effect depends, for example, on the
target of the investment, such as strategic infrastructure sectors (Aschauer, 1989a, b; Greene &
Villanueva, 1991; Calder�on & Serv�en, 2004; Perrotti & S�anchez, 2011), economic conditions,
such as periods of economic expansion and recession, and low and high credit liquidity (Conte
Filho, 2013; Alves & Luporini, 2008; Soave, 2016).
Inspired by these studies and motivated by recent political changes in Brazil that have the
potential to affect income distribution and public investment - including the announcement of
a new public investment program aiming at enabling macro-sectoral conditions for economic
growth and development (Taioka, Sanches, Brenck, & Oliveira, 2023) and a recent expansion
of Programa Bolsa Fam�ılia (Bergamin, Serra, Sanches, Gomes, & Nassif-Pires, 2022) - we
noticed a gap in the empirical literature for Brazil on the relationship between public and
private investment in different income inequality. Understanding that the levels of income
inequality generate different stimuli in the dynamism of the economy (Carvalho & Rezai,
2016), our hypothesis posits that the inequality regime may influence the magnitude of the
ultimate impact of public investment on private investment. Thus, this article aims to
investigate whether the effect of public investment on private investment differs in regimes of
relatively high and low-income inequality.
Keynes (1964) argues for the importance of fiscal intervention in controlling the volume of
investment to avoid expressive fluctuations in employment [1]. Moreover, Keynes also
discusses the positive relationship between the marginal efficiency of capital (as a determinant
of private investment) and the propensity to consume. Additionally, Kalecki (1945), when
discussing ways of achieving full employment, argues that public investment does not exclude
the need to stimulate mass consumption. In this sense, our working hypothesis is that a
relatively equal income distribution enhances the public investment multiplier, leading to a
stronger crowding-in effect on private investment compared to an economic scenario with
relatively higher income inequality.
Using a linear Vector Autoregressive Model (VAR) for Brazil in 1996–2022, we find a
crowding-in effect, in line with the more recent literature: a 10% rise in public investment leads
to a 2.5% increase in private investment after four quarters. However, using a Threshold Vector
Autoregressive model (TVAR), we find that the crowding-in effect only occurs in the
relatively low-inequality regime: a 10% rise in public investment leads to a 3.4% increase in
private investments after four quarters. In contrast, the response in high-inequality scenarios is
not statistically different from zero. Our findings underscore the significant macroeconomic
implications of reducing income inequality: it has the potential to enhance the responsiveness
of private investment to public investment.
The remainder of this paper is organized as follows: Sections 2 and 3 review the literature,
focusing on the empirical literature about the relationship between private and public
investments in the Brazilian economy. In Section 4, we elaborate on the econometric
methodology and the data employed in this study. Section 5 presents the results, while Section
6 offers a discussion of them. Section 7 presents some robustness tests. Finally, in Section 8, we
draw our conclusions.

2. Related literature
There is no consensus in economic theory on the relationship between government spending
and private investment, with a more evident disagreement when comparing different schools
of thought. Regarding public investment more specifically, some authors argue that its
expansion crowds out private investment, as both types of investment might compete for
physical and financial resources. In this case, the increase in public investment could raise
production costs, for instance, through elevating interest rates or affecting expectations on EconomiA
taxation (if the public expenditure results in a higher government deficit), leading to a
reduction in private investment (Jacinto & Ribeiro, 1998; Cruz & Teixeira, 1999; Sonaglio,
Braga, & Campos, 2010; Reis et al., 2019).
On the other hand, the crowding-in effect occurs when private investment responds
positively to public investment. The empirical literature typically presents two channels
through which public investment boosts private sector investment (Bredow et al., 2022; Iasco-
Pereira & Duregger, 2023; Cruz & Teixeira, 1999; Sonaglio et al., 2010; Reis et al., 2019;
Jacinto & Ribeiro, 1998): (1) public investment stimulates aggregate demand through its
income multiplier effect, encouraging production and investment by the private sector (the so-
called “accelerator effect”); (2) Public investment provides better supply conditions to the
private sector (for example, through investments in infrastructure and human capital
formation), contributing to increasing labor productivity and reducing production and
transaction costs.
Although not exclusively, such a positive relationship between public and private
investment appears mainly in demand-led growth models. Tavani and Zamparelli (2017), for
instance, emphasize that the provision of public infrastructure positively affects labor
productivity by providing better economic conditions for innovation. Inspired by Mazzucato
(2013), the authors highlight the role of public investment in the innovative process and the
productivity of the economy. In this sense, Ciaffi, Deleidi, and Mazzucato (2024) empirically
show for a group of OECD countries that an expansion in public investment related to research
and development (R&D) generates a positive impact on business R&D investment.
Dutt (2013), in turn, despite constructing a demand-led growth model, considers supply
constraints such as the possibility of financial crowding-out effects in a scenario where the
government accumulates a public deficit. Even with such effects, however, it is possible to
verify a crowding-in outcome due to the direct and indirect effects of public investment on
economic activity. This positive effect may persist in the long term if productivity also
responds to public investment. The author emphasizes that even when accounting for negative
expectations regarding an increase in the public deficit, crowding out is not a rule – and, in fact,
it is logically plausible that public spending, especially in the form of investment, generates
positive short-term and long-term effects.
The empirical literature verifies that the relationship between public and private investment
differs for developed and developing economies. For instance, Soave, Gomes, and Sakurai
(2016) studied the relationship between public investment growth and aggregate demand for
48 countries between 1975 and 2009. The sample was divided into two subsamples: 24
developed countries and 24 developing countries, following the World Bank classification.
The results indicate a crowding-in effect in the long term. This effect is more significant for
developing countries, a result that is confirmed by Izquierdo et al. (2019), who point out that
the efficiency of an increase in public investment tends to be lower (higher) in countries where
capital scarcity is less (more) pronounced. Moreover, when public investment is applied in
strategic areas, such as infrastructure, it can enhance the positive effect on economic growth
and increase the stimulus for private investment (Greene & Villanueva, 1991; Calder� on &
Serv�en, 2004).
Table 1A, in Appendix A [2], summarizes the main findings of the empirical literature
analyzing the impact of public investment on private investment in the Brazilian context. In
general, studies that investigated that relationship for years up to the 1990s report the existence
of a negative relationship between public and private investment, such as Melo and Rodrigues
J�unior (1998), Jacinto and Ribeiro (1998), and Cruz and Teixeira (1999) - although the latter
reports a complementarity in the long term. Some other studies use data from the 1990s to the
mid-2000s and have found a complementary relationship between the two types of investment
(Alves & Luporini, 2008; Sanches & Rocha, 2010). An exception is Sonaglio et al. (2010),
who find evidence of a crowding-out effect for the same period.
ECON However, studies that use recent data and updated methodologies tend to find a
complementary effect between public and private investments, in contrast to the evidence from
previous literature (Tadeu & Silva, 2013; Dos Santos et al., 2016; Reis et al., 2019; Bredow
et al., 2022; Fraga & Ferreira-Filho, 2023; Iasco-Pereira & Duregger, 2023). In particular,
these more recent studies incorporate the period related to the Programa de Aceleraç~ ao do
Crescimento (PAC) - a Federal government infrastructure investment program - into the
database. Furthermore, some more recent studies use, as private investment, the variable of
investment in machinery and equipment (Bredow et al., 2022; Iasco-Pereira & Duregger,
2023; Alves & Luporini, 2008; Dos Santos et al., 2016) as it is the most relevant category for
productivity gains and income growth (Bredow et al., 2022), and thus considered more
accurate. However, Sanches and Rocha (2010) use investment (both public and private) in
construction, primarily due to data availability. Finally, studies like Tadeu and Silva (2013) and
Fraga and Ferreira-Filho (2023) emphasize that the crowding-in effect is mainly attributed to
public investment in infrastructure.
Besides the econometric literature on the relationship between private and public
investments, our study is related to the empirical evidence about social benefits multipliers.
The empirical literature has shown that social benefits in Brazil contribute to a substantial
fiscal multiplier (Sanches & Carvalho, 2022; Resende & Pires, 2021; Orair, Siqueira, &
Gobetti, 2016) and exert positive effects on household consumption and private investment
(Sanches & Carvalho, 2023), as well as to reduce income inequality (Hoffmann, 2013). In
particular, Sanches and Carvalho (2023) estimate that one unit spent on social benefits
generates 2.3 units in consumption and 1.58 units in private investment after two years. In this
sense, our results underline that literature, as redistributive policies, for instance, by reducing
income inequality, create economic conditions wherein one would expect private investment
to respond more to public investment.

3. Empirical evidence using non-linear approaches

Additionally, economic cycles are another important factor impacting the magnitude of the
crowding-in effect. For instance, using an STVAR (Smooth Transition Vector Autoregressive
model) for the period from 1947 to 2008, Auerbach and Gorodnichenko (2012) found
significant evidence that government spending in the US has a greater impact on output in
times of recession. Following a similar STVAR model, Orair et al. (2016) found comparable
results for Brazil between 2002 and 2016. According to the author’s findings, during economic
expansions, the response of output to fiscal impulses tends to be subdued, resulting in
multipliers lower than unity, with a maximum of 0.8. Conversely, in periods of economic
downturns, these multipliers rise to nearly 2, signifying a more substantial impact on output
during contractions in the economy.
Regarding the effect of government transfers in times of expansion and recession, using the
TVAR approach (Threshold Vector Autoregressive model) for data from 2008 to 2022,
Almeida, Ribeiro, and Schommer (2023) also verified a higher income transfer multiplier
during recessions (0.51 in the short term and 0.99 in the long term), although lower than unit, in
comparison to the expansion regime (0.11 and 0.31, respectively). As for the importance of
credit liquidity on fiscal multipliers, Soave (2016) analyzes Brazil from 1995 to 2014 using the
TVAR method, showing that, although output and income respond positively to fiscal shocks,
both in times of high and low liquidity, the multipliers are higher and the responsiveness to
shocks are more persistent in the liquidity-constrained regime. This indicates that in periods of
low liquidity in the financial market, a fiscal shock is more effective since consumption and
investment would be more tied to income and realized profit than to expected future values.
Carvalho and Rezai (2016) analyzed how low and high-income inequality regimes affect
the demand regime. The authors estimated a TVAR for the US between 1967 and 2010,
showing that inequality has a negative effect on output and can lead to a change in the demand
regime in favor of profits. Therefore, more equitable income distributions would positively
affect output. Hence, building upon the literature on crowding-in/out effects and economic EconomiA
cycles, in the subsequent sections we empirically address the following question: to what
extent can low- and high-income inequality regimes influence the magnitude of the impact of
public investment on private investment?

4. Data and methodology

4.1 Data
We employed quarterly time series from 1996 to 2022 for the following variables:
(1) Public Investment: general government investment (comprising federal, state, and
municipal governments), excluding state-owned companies. We obtained the
quarterly series from the National Treasury of Brazil (STN), which provides public
investment data in quarterly frequency from 2010 onward. For periods before 2010, we
extracted the public investment series from the study conducted by Dos Santos et al.
(2012), who estimated the series in monthly frequency. By combining both series,
following the approach of Bredow et al. (2022), we achieved a comprehensive
quarterly dataset. The public investment series is presented in millions of Reais,
adjusted for inflation using the National Construction Cost Index (INCC-DI), from the
Getulio Vargas Foundation (FGV) since public investments focus on infrastructure
(Miguez, 2016), following Bredow et al. (2022). The series is also seasonally adjusted
by the X-13 Arima Method [3].
(2) Private Investment: it is an index (1995 5 100), given in real terms, and also seasonally
adjusted, calculated by the Institute of Applied Economic Research (IPEA). As it is a
monthly index, we transformed it into a quarterly frequency by taking the average of
three months (as in Bredow et al., 2022). It refers to gross fixed capital formation in
machinery and equipment. Similar to Bredow et al. (2022) and Iasco-Pereira and
Duregger (2023), the assumption adopted here is that investments in machinery and
equipment are primarily carried out by the private sector (see Dos Santos et al., 2016).
The variable for investment in machinery and equipment is used by the recent literature
(Bredow et al., 2022; Iasco-Pereira & Duregger, 2023; Alves & Luporini, 2008; Dos
Santos et al., 2016), since it is considered the most relevant category for productivity
gains and income growth (Bredow et al., 2022).
(3) Gini index for disposable income: this annual series was obtained from the
Standardized World Income Inequality Database (SWIID) (Solt, 2020). It was
transformed into quarterly frequency using cubic interpolation [4], following Carvalho
and Rezai (2016). The Gini coefficient refers to post-tax and post-transfer income
(disposable income).
(4) Installed capacity utilization of the manufacturing industry: The data were obtained
from the Time Series Management System of the Central Bank of Brazil (BCB) and
made available at a quarterly frequency by FGV, from their Conjunctural Survey of
Manufacturing Industry.
(5) Real exchange rate index (US Dollar to Brazilian Real): calculated by the Department
of Statistics of the BCB (BCB-DSTAT) using the Extended National Consumer Price
Index (IPCA) of the Brazilian Institute of Geography and Statistics (IBGE). We use the
quarterly average of the monthly data available.
(6) Real interest rate: we use the monthly Brazilian federal funds rate (Selic rate)
accumulated in annual terms (basis of 252 days) from the Open Market Operations
Department of the BCB (BCB/Demab) and the 12-month percentage change in the
IPCA. The data used in the estimation comprises the average of the monthly data for
each quarter.
ECON (7) Primary Commodity Price Index: monthly data from the International Monetary Fund
(IMF). We use the average for the quarter. The time series available starts in January
2003.

4.2 Methodology
Following the recent literature investigating the relationship between private and public
investment in Brazil, we employ a vector autoregressive model to analyze the dynamic
impulse-response functions derived from the empirical model (Dos Santos et al., 2016; Reis
et al., 2019; Bredow et al., 2022). Before analyzing the hypothesis of this article, that is, how
the level of income inequality influences the public and private investment relationship, we
employ an estimation of simultaneous dynamic equations in a standard linear VAR model, as
shown in Sims (1980) (see Dos Santos et al., 2016; Bredow et al., 2022), using public
investment and private investment as our endogenous variables vector. Both variables were
log-transformed and differenced in the first order since the Augmented Dickey-Fuller test
showed they are integrated of order one. Based on the information criteria, we selected 4
lags [5].
Next, to consider a possible non-linearity in the relationship between the two endogenous
variables when considering income inequality levels, we employ a Threshold Vector
Autoregressive model (TVAR), as in Carvalho and Rezai (2016), Almeida et al. (2023), and
Soave (2016). Following Carvalho and Rezai (2016), we utilize the Gini index as our threshold
variable, enabling us to derive model results for relatively low- and high-income inequality
levels.
The two-dimensional TVAR aims to estimate the non-linearity of the dynamic relationship
between the endogenous variables. The threshold, or value among the possible values of the
transition variable, is thus defined so that the sum of the squared residuals can be minimized,
and the estimated coefficients will differ in the regimes. Tsay (1998) proposed an extension of
the regime shift autoregressive model (threshold) to the multivariate context, giving rise to
the TVAR.
The TVAR model can be represented as follows (Almeida et al., 2023):
! !
Xp X p
Yt ¼ α1 þ β1;i Yt−1 I½St ≤ θ� þ α2 þ β2;i Yt−1 I½St > θ� þ vt (1)
i¼1 i¼1

vt ¼ I½St ≤ θ�v1;t þ I½St > θ�v2;t (2)

Where (Yt) is a vector of endogenous variables and (St) is the threshold variable and θ the
threshold. (βj,i) is the matrix of lagged coefficients associated with period (i) and regime (j),
where: j 5 1 and j 5 2 stand for the relatively low- and high-inequality regimes, respectively.
(I) is an indicator that can be set to (1) if the condition in brackets is true or (0) if the condition is
false. (vi,j) is a vector of random errors, and (αj) is a vector of constant terms for the regime (j).
It should be noted that non-linearity is a property of the TVAR model, but within each regime,
the model will be linear.
We performed an estimation for a two-dimensional TVAR [6] for the period 1996–2022,
with public investment and private investment as endogenous variables. Again, both variables
were log-transformed and differenced in the first order. We adopted one lag for the vector
autoregressive model, following two of the three information criteria (BIQ and HQ) [7]. The
Gini index for disposable income was utilized as the threshold variable, lagged by one period
and differenced in the first order. The threshold value was automatically determined through a
grid search, minimizing the sum of squared residuals [8]. We obtained linear accumulated
impulse-response functions to Cholesky standard deviation innovations for each TVAR
regime.
 We also conduct tests for the robustness of the VAR model by including the following EconomiA
exogenous control variables (see Section 7): installed capacity utilization, real exchange rate,
real interest rate, and primary commodity price index. All variables used were log-transformed
and first-order differentiated, as the Augmented Dickey-Fuller test showed they are first-order
integrated. Following three information criteria (AIC, HQ, FPE), we adopted four lags for the
vector autoregressive model.
Furthermore, we used the local projections model based on Jord�a (2005) to perform the
robustness test of the model. The non-linear version of this methodology uses a smooth
transition function as in Auerbach and Gorodnichenko (2012) to separate the data into two
regimes. Unlike the TVAR model, the local projections package in R allows us to estimate the
confidence interval for the two inequality regimes [9]. The local projections model can be
represented as follows (Auerbach & Gorodnichenko, 2012):

expð−γzt Þ
Fðzt Þ ¼ (3)
ð1 þ expð−γzt ÞÞ0

varðzt Þ ¼ 1; Eðzt Þ ¼ 0 (4)

where (zt) is standardized so that γ (>0) is scale-invariant, the observations for the two regimes
are the product of the transition function and the endogenous variables:
Regime 1: yt−l ð1 − Fðzt−1 ÞÞ; l ¼ 1; . . . ; p
Regime 2: yt−l Fðzt−1 Þ; l ¼ 1; . . . ; p
Public investment is ordered first, as it is considered the most exogenous [10] (Dos Santos
et al., 2016; Bredow et al., 2022). As highlighted in the fiscal multiplier literature, when using
high-frequency data, there is little or no fiscal policy response to unexpected shocks in
aggregate demand (or components, such as private investment) within the same quarter since
policymakers take more than a quarter to react to the macroeconomic conditions and decide the
next steps of fiscal policy (Blanchard & Perotti, 2002).
A common criticism of VAR models is that they require a predefined ordering of variables,
often addressed through the Cholesky decomposition. The ordering usually depends on
economic theory or institutional knowledge for identification (Stock & Watson, 2001; Jalles,
2017). For instance, Blanchard and Perotti (2002) argue that government expenditure
should be ordered first when estimating fiscal multipliers with high-frequency data, as
governments do not immediately respond to GDP changes. Another major limitation of VAR
models is their sensitivity to misspecification - such as omitted variables, incorrect lag
selection, or incorrect orthogonalization of innovations - which distorts impulse response
functions (Hendry, 1995; Ericsson, Hendry, & Mizon, 1997). Since impulse responses rely on
increasingly long-horizon forecasts, specification errors accumulate over time (Braun &
Mittnik, 1993). To mitigate these issues, we include control variables and test different
specifications of the model.
Additionally, VAR models face identification and size-related challenges, as estimating too
many parameters leads to imprecise results (L€utkepohl, 2005), which is why recent literature
has shifted towards Local Projections for impulse response estimation (Gupta, Talles, Mullas-
Granados, & Schena, 2017; Heimberger, 2020). Given these advantages, we apply Local
Projections in our nonlinear model as a robustness test. Finally, since linear VAR models fail to
capture nonlinear responses to shocks, we also employ Threshold VAR (TVAR) models to
address potential asymmetries (Donayre & Wilmot, 2016).
ECON 5. Results
5.1 Linear VAR
Employing a standard linear bi-dimensional VAR, we estimate that an increase of 10% in
public investment generates a 2.5% increase in private investment after four quarters. The
immediate response is 1.48%. It is noteworthy that these results are close to the ones found by
Bredow et al. (2022) (see Table 1A, in Appendix A). The response is statistically significant at
the 10% level for all periods (Figure 1). Tests on the residuals of this model [11] are available in
Appendix B and show that the estimation is stable and free of problems such as
heteroscedasticity and residual autocorrelation.

5.2 TVAR
The coefficients from the TVAR estimations are presented in Table 1. Both equations for
private and public investments are provided, but our analysis will primarily concentrate on the
private investment equation. This focus aligns with our investigation into the crowding-in
effect across two inequality regimes. The threshold parameter estimated by the method for the
Gini coefficient was 0.4656, with 21.7% of the observations lower than this value (“low-
inequality regime”) and 78.3% of the observations higher than the threshold value (“high-
inequality regime”).
As indicated by the private investment equation, we observed a crowding-in effect in the
relatively low-inequality regime: private investment responds positively to a shock in public

Figure 1. Accumulated response of private investment to a standard deviation shock in public investment in the
linear VAR. Note: The dashed lines correspond to confidence intervals of 90%. Source: Authors’ elaboration

Table 1. Results of the two-dimensional TVAR estimation for Brazil (1996–2022) using the Gini index for
disposable income as a threshold

Private investment equation Public investment equation

Low High Low High
inequality inequality inequality inequality

Private investment (�1) 0.0281 �0.0731 �0.4822** 0.4349***

Public investment (�1) 0.2477** �0.0480 0.2041 0.0345
Intercept �0.0294 0.0160 �0.0258 0.0026
Note(s): Significance: ***1%, **5%, *10%
Source(s): Authors’ elaboration
investment. This impact is statistically significant at a 5% significance level, as detailed in EconomiA
Table 1. In contrast, within the relatively high-inequality regime, public investment does not
significantly impact private investment.
Figure 2 illustrates the cumulative impulse-response function of private investment to a
one-standard-deviation shock in the public investment variable for both relatively low- and
high-inequality regimes. The graph underscores a notably positive response of private
investment to a shock in public investment within the relatively low-inequality regime,
showcasing a significant disparity between the two regimes over time, despite similar short-
run responses. In other words, our findings suggest the presence of a crowding-in effect when
income inequality is lower.
More precisely, the cumulative response depicted in Figure 2 reveals that a 10% increase in
public investment results in a 3.4% surge in private investments after four quarters in the
relatively low-inequality regime. In contrast, the corresponding response is not statistically
significant in the relatively high-inequality scenario. In the short run, however, both regimes
display a comparable reaction, with a 10% increase in public investment leading to an
approximately 0.8% increase in private investment. Even so, as we notice in Table 1, only the
immediate response of private to public investment in the relatively low-inequality regime is
statistically significant.
Although it is not our focus in this article, Table 1 shows a positive response of public
investment to private investment for the relatively high-inequality regime and a negative in the
case of relatively low-inequality. Appendix C shows the impulse response functions for this
exercise.

6. Discussion
Our findings are consistent with recent literature on the Brazilian economy, revealing a
crowding-in effect. Importantly, our results parallel those of Bredow et al. (2022), who
employed a linear VAR model for Brazil using similar time series data from 1996 to 2018.
While their study found that a 10% expansion in public investment leads to a 2.04% increase in
private investment over time (after four quarters), our analysis reveals a 2.5% increase for the
same period. Moreover, Bredow et al. (2022) find an immediate effect (in the first quarter) of
1.64%. Our results for the linear VAR indicate a similar impact (1.48%).
When allowing for different regimes using the TVAR approach, we find that the crowding-
in result only appears in the relatively low-inequality regime: a 10% expansion in public

Figure 2. Accumulated response of private investment to a standard deviation shock in public investment in
relatively low- and high-inequality regimes. Source: Authors’ elaboration
ECON investment leads to a 0.8% and 3.43% increase in private investment after one and four
quarters, respectively. Conversely, the response is not statistically different from zero in the
relatively high-inequality scenario.
It should also be noted that the most recent literature, which includes in the sample the sharp
decline in the Gini index period in the mid-2000s, has found positive effects of public
investment on private investment (see Table A1 in Appendix A). Therefore, our result of the
crowding-in effect being more pronounced in the relatively low-inequality regime aligns with
and reinforces the findings in that literature, underscoring a positive relationship between
reduced inequality and investment.
In the relatively low inequality scenario, lower-income brackets, exhibiting a higher
inclination to consume (as observed in Palomo, Carvalho, & Toneto, 2022, for Brazil),
contribute to a more dynamic economy since there is a redistribution from the class with a
higher propensity to save to the class with a higher propensity to consume (Carvalho & Rezai,
2016). This dynamic setting not only stimulates household consumption but also drives private
investment, which responds positively to increased aggregate demand, commonly referred to
as the accelerator effect (Hein & Vogel, 2008).
Since the lower-income groups exhibit a higher marginal propensity to consume (Palomo
et al., 2022; Kalecki, 1952) and a higher average propensity to consume (Serra & Sanches,
2025), a lower-inequality scenario enhances the stimulus effect of public investment - known
for its high multiplier impact on the economy (Sanches & Carvalho, 2022; Orair et al., 2016) -
on private investment. This occurs because the overall dynamic effects on economic activity
are stronger. Moreover, social benefits - expenditures that potentially reduce inequality - also
have a significant multiplier effect (as estimated for various countries by Cardoso et al., 2025),
as they further stimulate both consumption and investment dynamics (Sanches & Carvalho,
2023). These findings suggest that in a lower-inequality regime, the economy operates with
greater dynamism, as income redistribution fuels economic activity. Consequently, the
crowding-in effect of public investment on private investment becomes even more
pronounced in this context.

7. Robustness check
7.1 Alternative methodology: non-linear local projections
Figure 3 shows that our main conclusions still hold when we use an alternative non-linear
methodology. The local projections method produces results that are similar to the TVAR’s

Figure 3. Responses of private investment to a standard deviation shock in public investment - low and high
inequality, respectively. Note: The gray lines correspond to confidence intervals of 90%. Source: Authors’
elaboration
ones. The impact was positive and significant for both regimes in the first period. After the EconomiA
second period, the impulse-response function for the low inequality regime was significant at
10% throughout the entire period after the shock, until period 8, while it was not significant for
the high inequality regime [12].

7.2 Control variables

We check the influence of control variables to account for possible omitted variable bias, based
on relevant literature: degree of capacity utilization, real exchange rate, commodity price
index, and interest rate. We observe that the findings from Section 5 remain robust even after
including control variables: the response of private investment to a shock in public investment
continues to be statistically significant at 10% across most periods. The results can be found in
Appendix D [13].

8. Concluding remarks
This paper has analyzed how relatively high- and low-inequality regimes impact the
magnitude of the response of private investment to public investment. Our findings support the
conclusions drawn from recent literature, indicating that public investment has a crowding-in
effect on private investment—that is, the latter responds positively to the former.
More specifically, our linear VAR baseline estimation concludes that an increase of 10% in
public investment generates a 2.5% rise in private investment after four quarters. Using a
TVAR model, we have demonstrated that this crowding-in effect only occurs in the relatively
low-inequality regime: a 10% increase in public investment results in a 3.4% increase in
private investment after four quarters. On the other hand, in the relatively high-inequality
regime, this response is not statistically different from zero.
A surge in public investment, for example, produces a significant multiplier effect,
typically higher than one, as evidenced in Brazil (refer to Sanches & Carvalho, 2022; Orair
et al., 2016; Pires, 2014; Resende & Pires, 2021). According to our findings, this economic
stimulus positively impacts private investment, particularly in scenarios where inequality is
lower, fostering a more dynamic aggregate demand environment. In such circumstances, the
impetus for investment is heightened, driven by the increased dynamism in consumption
stemming from a more equitably distributed income.
Our results indicate that the crowding-in effect only appears in the relatively low-inequality
regime, which indicates the importance of redistributive policies to mitigate income inequality
and enhance the responsiveness of private to public investment. This aligns with empirical
findings in the literature highlighting the relevance of the social benefits multiplier effect.
Since social benefits are targeted toward individuals in lower-income groups, who exhibit a
higher propensity to consume, their multiplier effect is relevant, making the economy more
dynamic (Cardoso et al., 2025; Sanches & Carvalho, 2023; Resende & Pires, 2021; Orair et al.,
2016). This evidence may contribute to elucidating our finding that private investment
responds significantly to the stimulus in public investment (crowding-in effect) when income
inequality is lower.
Furthermore, our study suggests that the state has a key role as a driver of investment and
that its role is crucial in implementing and coordinating investment programs. Programs aimed
at greater income distribution are crucial to improving the welfare of low-income populations
and boosting the effects of macroeconomic stimuli, as public investment is more effective at
stimulating private investment in a less unequal economy.
In alignment with the literature on fiscal multipliers of social benefits, our results
underscore the substantial macroeconomic potential of policies designed to mitigate
inequalities. In this context, these policies play a pivotal role in advancing social equity and
propelling more inclusive economic growth.
ECON Notes
1. See Seccareccia (2012) for an analysis of Keynes’ proposal for the socialization of investment.
2. All appendices are available in the supplementary materials section.
3. Available in Eviews 12.
4. Available in Eviews 12 (“Cubic Match Last”).
5. The criteria AIC, HQ and FPE indicated 4 lags. SC criteria indicated 1 lag. However, the estimation
using 1 lag showed heteroscedasticity problems.
6. We performed the TVAR estimation using the “tsDyn” package in R.
7. Following the parsimony principle, we chose 1 lag for the estimation, according to the criteria
mentioned. The AIC criteria indicated 6 lags. Given that our sample is not very large, we opted to
include one lag in order to have a higher degree of freedom. However, when we estimate the model
using 6 lags, the substantial difference between the low and high inequality regimes persisted.
8. We adjusted the trimming parameter indicating the minimal percentage of observations in each
regime according to the LR Test, proposed by Lo and Zivot (2001) (which is based on Hansen (1999),
but for the multivariate case), and to increase the number of observations for the low-inequality
regime (which has lower observations). Setting the parameter to 0.2 (Carvalho & Rezai (2016) used
0.25), the LR test reveals non-linearity, leading to the rejection of the null hypothesis of a linear model
at a 5% significance level. We tested other parameter values, such as 0.15 and 0.10, for which the LR
test also detects non-linearity. Our results barely changed with this variation.
9. The package used was “pirfs” in R.
10. We also conducted estimations using an alternative specification, ordering the private investment
variable first. The results remained robust to this change, with the difference between the two regimes
persisting.
11. Since the residual tests are not available in the tsDyn package in R for the TVAR model, we
performed these tests only for the linear VAR models.
12. The switching function is similar to the work by Auerbach and Gorodnichenko (2012). We used
gamma as 2, but our conclusions are robust to other values. We used the lags for the model chosen by
the AIC criteria. We obtained similar results when using the BIC criteria.
13. Since the possibility of including control variables is not available in the tsDyn package in R for the
TVAR model, we have included controls only for the linear model.

References
Almeida, E. S., Ribeiro, E. P., & Schommer, S. (2023). Transfer^encias fiscais de renda, gastos e seus
omica brasileira. Anais do 51º Encontro Nacional de
efeitos multiplicadores na atividade econ^
Economia – Anpec.
Alves, J., & Luporini, V. (2008). Determinantes do investimento privado no Brasil: Uma an�alise de
painel setorial. Anais do 36º Encontro Nacional de Economia – Anpec.
Aschauer, D. A. (1989a). Is public expenditure productive?. Journal of Monetary Economics, 23(2),
177–200. doi: 10.1016/0304-3932(89)90047-0.
Aschauer, D. A. (1989b). Does public capital crowd out private capital?. Journal of Monetary
Economics, 24(2), 171–188. doi: 10.1016/0304-3932(89)90002-0.
Auerbach, A. J., & Gorodnichenko, Y. (2012). Measuring the output responses to fiscal policy.
American Economic Journal: Economic Policy, 4(2), 1–27. doi: 10.1257/pol.4.2.1.
Bergamin, J., Serra, G., Sanches, M., Gomes, J., & Nassif-Pires, L. (2022). Quais os poss�ıveis impactos
do Novo Programa Bolsa Fam�ılia? Uma an�alise de seus efeitos sobre a pobreza e o PIB. Revista
Iniciativa Econ^ omica, 8(1), 1–17.
Blanchard, O., & Perotti, R. (2002). An empirical characterization of the dynamic effects of changes in
government spending and taxes on output. Quarterly Journal of Economics, 117(4), 1329–1368.
doi: 10.1162/003355302320935043.
Braun, P., & Mittnik, S. (1993). Misspecifications in vector autoregressions and their effects on impulse EconomiA
responses and variance decompositions. Journal of Econometrics, 59(3), 319–341. doi: 10.1016/
0304-4076(93)90029-5.
Bredow, S., Cunha, A., & L�elis, M. (2022). Investimentos p�
ublicos e privados no Brasil entre 1996 e
2018. Revista de Economia Pol�ıtica, 42(4), 957–976.
Calder�on, C., & Serv�en, L. (2004). The effects of infrastructure development on growth and income
distribution. Central Bank of Chile, Working Paper No. 270.
Cardoso, D., Carvalho, L., Lima, G. T., Nassif-Pires, L., Rugitsky, F., & Sanches, M. (2025). The
multiplier effects of government expenditures on social protection: A multi-country study.
Development and Change, 56(1), 172–224. doi: 10.1111/dech.12869.
Carvalho, L., & Rezai, A. (2016). Personal income inequality and aggregate demand. Cambridge
Journal of Economics, 40(2), 491–505. doi: 10.1093/cje/beu085.
Ciaffi, G., Deleidi, M., & Mazzucato, M. (2024). Measuring the macroeconomic responses to public
investment in innovation: Evidence from OECD countries. Industrial and Corporate Change, 33
(2), 363–382. doi: 10.1093/icc/dtae005.
Conte Filho, C. (2013). Uma an�alise emp�ırica acerca do efeito do investimento p�
ublico sobre o
investimento privado no Brasil: 1971-2008. PhD Dissertation, Universidade Federal do Rio de
Janeiro.
Cruz, B., & Teixeira, J. (1999). The impact of public investment on private investment in Brazil, 1947-
1990. CEPAL Review, 1999(67), 75–84. doi: 10.18356/fbcbe22d-en.
Donayre, L., & Wilmot, N. (2016). The asymmetric effects of oil price shocks on the Canadian
economy. International Journal of Energy Economics and Policy, 6(2), 167–182.
Dos Santos, C., Orair, R., Gobetti, S., Ferreira, A., Rocha, W., Silva, H., & Britto, J. (2012).
Estimativas mensais da formaç~ao bruta de capital fixo p�ublica no Brasil (2002-2010). Economia
Aplicada, 16(3), 445–473. doi: 10.1590/s1413-80502012000300005.
Dos Santos, C., Modenesi, A., Squeff, G., Vasconcelos, L., Mora, M., Fernandes, T., . . . Braga, J.
(2016). Revisitando a din^amica trimestral do investimento no Brasil: 1996-2012. Revista de
Economia Pol�ıtica, 36(1), 190–213. doi: 10.1590/0101-31572016v36n01a11.
Dutt, A. K. (2013). Government spending, aggregate demand, and economic growth. Review of
Keynesian Economics, 1(1), 105–119. doi: 10.4337/roke.2013.01.06.
Ericsson, N. R., Hendry, D. F., & Mizon, G. E. (1997). Exogeneity, cointegration and economic policy
analysis. Journal of Business and Economic Statistics, 16(4), 370–387. doi: 10.1080/
07350015.1998.10524778.
Fraga, J., & Ferreira-Filho, H. (2023). The effects of infrastructure and public investment on the
elasticity of private investment: An empirical investigation for Brazil. Brazilian Journal of
Political Economy, 43(1), 275–298. doi: 10.1590/0101-31572023-3383.
Greene, J., & Villanueva, D. (1991). Private investment in developing countries: An empirical analysis.
IMF Staff Papers, 38(1), 33–58. doi: 10.2307/3867034.
Gupta, S., Talles, J., Mullas-Granados, C., & Schena, M. (2017). Governments and promised fiscal
consolidations: Do they mean what they say?. IMF Working Paper No. 17/39.
Hansen, B. (1999). Testing for linearity. Journal of Economic Surveys, 13(5), 551–576. doi: 10.1111/
1467-6419.00098.
Heimberger, P. (2020). The dynamic effects of fiscal consolidation episodes on income inequality:
Evidence for 17 OECD countries over 1978-2013. Empirica, 47(1), 53–81. doi: 10.1007/s10663-
018-9404-z.
Hein, E., & Vogel, L. (2008). Distribution and growth reconsidered: Empirical results for six OECD
countries. Cambridge Journal of Economics, 32(3), 479–511. doi: 10.1093/cje/bem047.
Hendry, D. F. (1995). Dynamic econometrics. New York: Oxford University Press.
Hoffmann, R. (2013). Transfer^encias de renda e desigualdade no Brasil (1995-2011). in T. Campello, &
M. Neri (Eds), Programa Bolsa Fam�ılia: uma d�ecada de inclus~ ao e cidadania (pp. 207–216).
IPEA, Bras�ılia.
ECON Iasco-Pereira, H., & Duregger, R. (2023). Public investment, infrastructure, and private investment in
Brazil: Is there a crowding-in effect?. Anais do 51º Encontro Nacional de Economia – Anpec.
Izquierdo, A., Lama, R., Medina, J. P., Puig, J., Riera-Crichton, D., Gramont, C. A. V., & Vuletin, G. J.
(2019). Is the public investment multiplier higher in developing countries? An empirical
exploration. IMF Working Paper No. 289.
Jacinto, P., & Ribeiro, E. (1998). Co-integraç~ao, efeitos crowding-in e crowding-out entre investimento
p�ublico e privado no Brasil: 1973-1989. Teoria e Evid^encia Econ^ omica, 6(11), 145–158.
Jalles, J. (2017). How do fiscal adjustments change the income distribution in emerging market
economies?. International Journal of Emerging Markets, 12(2), 310–334. doi: 10.1108/ijoem-
10-2015-0219.
Jord�a, O. (2005). Estimation and inference of impulse responses by local projections. American
Economic Review, 95(1), 161–182. doi: 10.1257/0002828053828518.
Kalecki, M. (1945). Full employment by stimulating private investment?. Oxford Economic Papers, os-
7(7), 83–92. doi: 10.1093/oxepap/os-7.1.83.
Kalecki, M. (1952 [1978]). Teoria da din^
amica econ^
omica. S~ao Paulo: Abril Cultural. (Coleç~ao Os
Economistas).
Keynes, J. M. (1964). The general theory of employment, interest and money. London: Macmilla.
Lo, M. C., & Zivot, E. (2001). Threshold cointegration and nonlinear adjustment to the law of one
price. Macroeconomic Dynamics, 5(4), 533–576. doi: 10.1017/s1365100501023057.
L€utkepohl, H. (2005). New introduction to multiple time series analysis. Berlin: Springer.
Mazzucato, M. (2013). The entrepreneurial state. London, New York: Anthem Press.
Melo, G., & Rodrigues J�unior, W. (1998). Determinantes do investimento privado no Brasil: 1970-
1995. Texto para discuss~ao No. 605, IPEA.
Miguez, T. (2016). Evoluç~ao da Formaç~ao Bruta de Capital Fixo na Economia Brasileira 2000-2013:
uma an�alise multissetorial a partir das matrizes de Absorç~ao de Investimento (MAIS). Ph.D.
Dissertation, Universidade Federal do Rio de Janeiro, Rio de Janeiro.
Orair, R. O., Siqueira, F. D. F., & Gobetti, S. W. (2016). Pol�ıtica Fiscal e Ciclo Econ^
omico: uma an�alise
baseada em multiplicadores do gasto p� ublico. Secretaria do Tesouro Nacional. (Org.). Finanças
P�
ublicas: XXI Pr^emio Tesouro Nacional, 2016: colet^ anea de monografias. Bras�ılia: Secretaria
do Tesouro Nacional, 1, 1–76.
Palomo, T., Carvalho, L., & Toneto, R. (2022). Marginal propensity to consume heterogeneity and
redistributive policies: The Brazilian case. Working Paper No. 10, Made/FEA-USP.
Perrotti, D., & S�anchez, R. J. (2011). La brecha de infraestructura en Am�erica Latina y el Caribe. Serie
DRNI N8153. Santiago do Chile: United Nations ECLAC.
omicos no Brasil. Economia Aplicada, 18(1), 69–90. doi:
Pires, M. (2014). Pol�ıtica fiscal e ciclos econ^
10.1590/1413-8050/ea350.
Reis, C., Ara�
ujo, E., & Gonzales, E. (2019). Public investment boosted private investment in Brazil
between 1982 and 2013. Journal of Economic Issues, 53(3), 813–840. doi: 10.1080/
00213624.2019.1644931.
Resende, C., & Pires, M. (2021). O impulso de multiplicador fiscal: implementaç~ao e evid^encia para o
Brasil. Estudos Econ^ omicos, 51(2), 213–243. doi: 10.1590/0101-41615121mpcr.
Sanches, M., & Carvalho, L. (2022). A contribuiç~ao da pol�ıtica fiscal para a crise brasileira de 2015-
2016: uma an�alise baseada em multiplicadores de despesas e receitas prim�arias do governo
central no per�ıodo 1997-2018. Nova Economia, 32(1), 7–36.
Sanches, M., & Carvalho, L. (2023). Multiplier effects of social protection: A SVAR approach for
Brazil. International Review of Applied Economics, 23(1), 93–112. doi: 10.1080/
02692171.2022.2117286.
Sanches, N., & Rocha, F. (2010). Investimentos estaduais p�
ublicos e privados: ‘bens’ substitutos ou
complementares?. Economia Aplicada, 14(2), 211–223. doi: 10.1590/s1413-
80502010000200006.
Seccareccia, M. (2012). The role of public investment as principal macroeconomic tool to promote EconomiA
long-term growth: Keynes’s legacy. International Journal of Political Economy, 40(4), 62–82.
doi: 10.2753/ijp0891-1916400403.
Serra, G. P., & Sanches, M. S. (2025). Conditional cash transfers versus universal basic income
programs: A comparative economic analysis using a social accounting matrix for Brazil.
Working Paper No. 25, Made/USP.
Sims, C. (1980). Macroeconomics and reality. Econometrica, 48, 1–48. doi: 10.2307/1912017.
Soave, G. P. (2016). Choques Fiscais e Instabilidade Financeira no Brasil: Uma Abordagem TVAR.
Anais do 44º Encontro Nacional de Economia – Anpec.
Soave, G. P., Gomes, F. A. R., & Sakurai, S. N. (2016). Efeitos do gasto do governo sobre o consumo
privado: evid^encias de pa�ıses desenvolvidos e em desenvolvimento. Estudos Econ^
omicos, 46(1),
5–41. doi: 10.1590/0101-416146151gfs.
Solt, F. (2020). Measuring income inequality across countries and over time: the Standardized World
Income Inequality Database. Social Science Quarterly, 101(3), 1183–1199. doi:10.1111/
ssqu.12795.
Sonaglio, C., Braga, M., & Campos, A. (2010). Investimento p�
ublico no Brasil: evid^encias dos efeitos
crowding-in e crowding-out no per�ıodo 1995-2006. EconomiA, 11(2), 383–401.
Stock, J., & Watson, M. (2001). Vector autoregressions. Journal of Economic Perspectives, 15(4),
101–115. doi: 10.1257/jep.15.4.101.
Tadeu, H., & Silva, J. (2013). The determinants of the long term private investment in Brazil: An
empirical analysis using cross-section and a Monte Carlo simulation. Journal of Economics,
Finance and Administrative Science. Special Issue, 18, 11–17. doi: 10.1016/s2077-1886(13)
70025-8.
Taioka, T., Sanches, M., Brenck, C., & Oliveira, R. (2023). O Novo Programa de Aceleraç~ao do
Crescimento (PAC): comparaç~ao com os PACs anteriores e estimaç~ao de seu efeito no PIB. Nota
de Pol�ıtica Econ^
omica, No 45, Centro de Pesquisa em Macroeconomia das Desigualdades,
Made/USP.
Tavani, D., & Zamparelli, L. (2017). Government spending composition, aggregate demand, growth,
and distribution. Review of Keynesian Economics, 5(2), 239–258. doi:10.4337/roke.2017.02.06.
Tsay, R. S. (1998). Testing and modeling multivariate threshold models. Journal of the American
Statistical Association, 93(443), 1188–202. doi: 10.2307/2669861.

Further reading
Brenck, C. (2021). Wage inequality and employment composition in Brazil: A VEC estimation for the
period 2004-2019. Nova Economia, 31(2), 345–380. doi:10.1590/0103-6351/6113.

Supplementary material
The supplementary material for this article can be found online.
About the authors
Marina da Silva Sanches holds a Ph.D. from the University of S~ao Paulo (USP) and is a researcher at the
Research Center of Macroeconomics of Inequalities (Made/USP). Marina da Silva Sanches is the
corresponding author and can be contacted at: marinasanchess.unb@gmail.com
Tainari Taioka is a Ph.D. candidate at the University of S~ao Paulo (USP) and is a researcher at the
Institute for Applied Economic Research (Ipea).
Clara Zanon Brenck holds a Ph.D. from The New School for Social Research and is a professor at the
Federal University of Minas Gerais (UFMG).
Gustavo Pereira Serra holds a Ph.D. from The New School for Social Research and is a professor at the
S~ao Paulo State University (Unesp).

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com
View publication stats