INFER WORKING PAPER SERIES

INTERNATIONAL NETWORK FOR ECONOMIC RESEARCH

No. 14 | 2024

Are Immigration Regularization Programs a Pull

Factor? Evidence for OECD Countries

Paúl Elguezabal (University of Goettingen)

Inmaculada Martínez-Zarzoso (University Jaumé I & University of Goettingen)

Website: Contact:
https://infer-research.eu/ publications@infer.info
 Are Immigration Regularization Programs a Pull Factor?
Evidence for OECD Countries

Paúl Elguezabal (University of Goettingen) & Inmaculada Martínez-Zarzoso (University

Jaumé I, Castellón, Spain & University of Goettingen)

Abstract

This paper evaluates the impact of regularisation programs on immigration flows using a newly
collected dataset and panel-data techniques applied to gravity models. The main novelty is
twofold. First to present the dataset with detailed information on regularisation policies in
OECD countries, including those implemented over the period from 1944 to 2023 and
specifying the timeframe of implementation and the origin nationalities targeted. And second,
to estimate the impact with a gravity model of bilateral migration applying a Poisson pseudo
maximum likelihood estimator for an unbalanced panel of 193 origins and 32 OECD
destinations for 199-2022. The main results indicate that the regularisation impact is very
heterogeneous across geographical regions of incoming migrants and across groups of countries
depending on their level of development. In particular, the results indicate that regularisation
programs are a pull factor for lower-income OECD destinations.

Keywords: Migration; Regularisation; Policy evaluation; High Dimensional Fixed Effects (HDFE)
Poisson pseudo maximum likelihood method; Income and regional heterogeneity.

JEL: C23; F22; F68; J15; J18; J61; J68; k37; N30; O15; O17

Corresponding author: Paúl Elguezabal; paul.elguezabal@uni-goettingen.de; Platz der Göttinger

Sieben 3, 37073 Göttingen

1
1. Introduction
Regularisation programs on immigration have been frequently implemented in the past
in many countries and are considered an important immigration policy tool. One of the earliest
regularisation programs was implemented by the end of the Second World War in France.
Moreover, among the most known immigration regularisation-programs is the 1986
Immigration Reform and Control Act (IRCA) of Reagan’s administration (DEMIG, 2015). This
policy tool emerges in the public debate from time to time to deal with of undocumented
migrants, as is the case currently in some European Union countries. For example, it has been
of high relevance in the political discourse in France and Italy and it is in discussion in 2024 in
the Spanish parliament.

Although regularisation is one of the alternatives that governments have when dealing
with a considerable quantity of undocumented migrants, governments fear its implementation
could become a pull factor for future migration, “encouraging “speculative” irregular migration
in anticipation of future regularisation programs” (OECD, 2018). This could be the main
argument against this policy and for this reason other options, such as to expel the clandestine
migrants, disregard them, or tolerate their presence (Sunderhaus S. , 2012), are usually
considered. Since the selection of the most appropriate immigration policy is a highly
controversial debate, which deservers further investigation. The analysis of the socio-economic
consequences of regularisations programs is crucial to provide sound empirical evidence of its
expected effects.

The main aim of this paper is to answer the question whether regularisation act a pull
factor and if so to what extent. To answer this question, we first need a definition and second a
compilation of regularization programs. According to Baldwin-Edwards & Kraler (2009) a
Regularisation program is “…any state procedure by which third country nationals who are
illegally residing, or who are otherwise in breach of national immigration rules, in their current
country of residence are granted a legal status.” Therefore, based on this definition, we focus
on regularisation procedures that comply with three requisites: (1) do not form part of the
regular migration policy framework, (2) run for a limited period of time and (3) target specific
categories of non-nationals in an irregular situation. We also include programs that regularize
migrants on humanitarian grounds as those targeting rejected asylum seekers. Since the datasets
that compile regularization programs are scarce and incompletes, we build a new dataset that
collects all regularization programs covering a large time spam and set of countries. With the

2
data at hand, we will empirically estimate the causal effect of the programs on immigration and
quantify the effect.

The question whether regularisation is as a pull factor, has been previously studied by a
number of authors, but the results are mixed (Finotelli & Arango, 2011). Orrenius & Zavodny
(2003), on USA’s IRCA, found a short-term decrease in illegal migration, which soon returned
to normal levels. Differently, Wong & Kosnac (2017) and Larramona & Sanso‐Navarro (2016),
found that specific regularisations in US and Spain had no impact on subsequent migration. In
contrast, Wehinger (2014), found a positive “pull factor” effect for European countries. This
last outcome is what the literature and policymakers reasonably expect. Concluding, the
migration literature has not broadly studied regularization policies as a pull factor, and some
studies do not even include it as a migration driver (Czaika & Reinprecht, 2022) & (Docquier,
Peri, & Ruyssen, 2014).

We aim to fill this gap by assembling a comprehensive dataset on regularization policies

and by estimating its effect on migration globally. The core sources used to assemble the dataset
are Baldwin-Edwards & Kraler (2009), DEMIG (2015) and the OECD’s international migration
reports (from OECD (1997) to OECD (2023)). Additionally, a handful of papers helped to
complete the dataset. As a result, this novel dataset depicts the implementation of a
regularisation policy for OECD countries over eight decades, specifying when and to which
origin nationality is it applied over the period from1944 to 2023. At least 13,540,077 individuals
benefited from regularisations during this period in the OECD. In order to understand the
magnitude, it is more than the population in 2024 of 16 member countries. Therefore, it
reaffirms the relevance of understanding the policy’s implications. With this new dataset we
estimate a gravity model of migration flows (Czaika & Parsons, 2017) (Mayda, 2010) to
quantify the impact of the regularisation policy on immigration inflows. We control for the
income ratio (origin/destination), the network of origin nationality at the destination, and a
battery of fixed effects (origin-destination, origin-year, and destination-year) to account for all
time-invariant and time-varying omitted explanatory that could bias the results. This model is
applied to an unbalanced panel-dataset of 193 origins and 32 OECD destinations for the period
1996-2022. An additional contribution of this paper is to analyze more closely the regularisation
impact heterogeneity. It helps to understand previous mixed findings in the existent literature.
Furthermore, it allows policymakers to make decisions based on more complete information
regarding the impacts of regularisation on future migration.

3
 The main empirical contribution of the current paper to the previous literature on the
impact of regularisations on migration is that most of them are based on country cases with
limited external validity (Finotelli & Arango, 2011), (Orrenius & Zavodny, 2003), (Wong &
Kosnac, 2017), and (Larramona & Sanso‐Navarro, 2016). The exceptions were Wehinger
(2014) that studied European countries as destinations and Ortega & Peri (2009) and Mayda
(2010)) that focused on 14 OECD destinations. Above all, our study improves previous research
in several respects. First, the data include multiple destinations (32) and origins (193). Second,
current research is the first on the topic that uses a gravity model with HDFE 1 . Third, this is the
first gravity model paper evaluating regularisation policy as target variable, contrasting with
(Ortega & Peri, 2009) & (Mayda, 2010)) that tested migration regulations in general. Fourth,
the PPML improve robustness of results and this is the first evaluation of regularisation
implementing it.

The main results indicate that regularisations do not seem to be a pull factor for all
OECD countries on average. Although, the impact on migration inflow by the policy is not
different from zero, the effect is heterogeneous across groups of countries. Whereas the results
are similar for all origin countries’ income quartiles and richest destinations, the given policy
is found to be a pull factor for lower-income OECD destinations2 , especially when migrants
come from an above-median-income origin country. Moreover, it is shown to be a pull factor
for some regions of origin (Latin-America and South Asia) but only when migrating to the
poorest OECD destinations. The quality of institutions and enforcement capabilities (such as
border controls) of the richest OECD countries3 could compensate for the regularisation
policy’s pull factor and explain why it does not increase future migration inflow as
policymakers expect.

The structure of the paper is the following. Next section presents an overview of the
research conducted on regularisation policy. Section 3 describes the data, variables,
methodology, and model specification. Section 4 presents the results with its heterogeneity
analysis, while the section 5 present robustness checks with three different strategies to address
relevant concerns. Finally, the findings are discussed and conclusion is stated in section 6.

1 Previous researches were based on simple descriptive statistics migration (Finotelli & Arango, 2011), OLS
(Wong & Kosnac, 2017) & (Wehinger, 2014), instrumental variables (Orrenius & Zavodny, 2003) (Ortega &
Peri, 2009), synthetic control methods (Larramona & Sanso‐Navarro, 2016), propensity score matching (Wong
& Kosnac, 2017), and gravity model (Ortega & Peri, 2009) & (Mayda, 2010)).
2 Chile, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Mexico, Poland, Portugal, Slovakia, and Turkey
3 Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Israel, Italy, Japan, Rep of

Korea, Luxembourg, Netherlands, Norway, Slovenia, Spain, Sweden, Switzerland, United Kingdom and USA
4
2. Literature Review
Concerning previous research, the socio-economic effects of former regularisations
have been widely analyzed by scholars in recent years, but the results remain mixed. The labor-
market effects are the most studied outcomes (Kossoudji & Cobb-Clark (2002); Orrenius &
Zavodny (2015); Méndez, Sepúlveda, & Valdés (2016); Elias, Monras, & Vazquez-Grenno
(2024); Deiana, Giua, & Nisticò (2024); Molinari, Impicciatore, & Ortensi (2023) and Bahar,
Ibáñez, & Rozo (2021)). Authors mostly found positive effects on labor outcomes for migrants.
While other authors focused on the effect on national workers (Bahar, Ibáñez, & Rozo (2021);
Machado (2017) and Borjas & Edo (2023)). The overall findings are that the feared labor market
outcomes deterioration for nationals is small, with some regularisation programs even
generating income improvements for nationals.

Regarding the impact on macroeconomic outcomes, Borjas & Edo (2023) reported a 1%
GDP increase due to the policy in France. Along those lines, Alvarez, et al (2022), project a
positive impact of 2.6 to 4.4% on GDP to most prominent Latin American destinations as a
result from Venezuelan migration. They argue that expanding access to formal labor (as
regularisation does) accelerates this impact. Moreover, it was estimated by Elias, Monras, &
Vazquez-Grenno (2024) that the 2004 Spanish regularisation of 600,000 undocumented
migrants did increase tax revenues and did not increase social expenditures. Moreover, it has
been shown that the regularisation of migrants reduces crime when evaluating the Italian
regularisation mechanism (Pinotti, 2017). Also, Fasani (2018) using instrumental variables
found that after each Italian regularisation, there was crime reduction among non-EU migrants
in the following year but it was small and no persistent.

The question whether regularisation is as a pull factor, has been studied with no clear
empirical answer (Finotelli & Arango, 2011). Orrenius & Zavodny (2003), on USA’s IRCA,
found a short-term decrease in illegal migration, which soon returned to normal levels.
Differently, Wong & Kosnac (2017) concluded that “knowledge regarding a prospective
legalization program in the US does not increase the intent to migrate among prospective
migrants”. Also, Larramona & Sanso‐Navarro (2016), using synthetic control methods, found
that Spanish 2005 regularisation had no impact on subsequent migration. In contrast, Wehinger
(2014), focusing on Europe and using national apprehension data, found a positive “pull factor”
effect. This last outcome is what the literature and policymakers reasonably expect. In fact,
Ortega & Peri (2009) concluded that migration increases when destination countries adopt more
tolerant immigration laws, including regularisations, but with no specific evaluation of
5
regularisation programs. Following these lines, Mayda (2010), using a gravity model, found
that relaxing migration restrictions (as regularisation does) increases the impact of pull factors
such as GDP per capita at destination and origin characteristics as youth share of the population.
Similarly, relaxing migration restrictions reduces the relevance of push factors such as GDP per
capita at origin and geographical characteristics such as distance. Concluding, the migration
literature has not deeply studied regularization policies as a pull factor, and some studies miss
to include it as a migration driver when enumerating its determinants (Czaika & Reinprecht,
2022) & (Docquier, Peri, & Ruyssen, 2014) .

Since the relevance of regularisation as a pull factor is unclear in the literature, in this
paper we aim to fill this research gap. The existent research on the topic is still scarce and the
findings go in several directions. On the one hand, according to several authors regularisation
could not be considered a pull factor (Finotelli & Arango (2011); Orrenius & Zavodny (2003);
Wong & Kosnac (2017); and Larramona & Sanso‐Navarro (2016)). On the other hand,
Wehinger (2014), considers it as a pull factor, while relaxing migration policies (including not
only regularization, but others as well) are considered pull factors in Ortega & Peri (2009) and
Mayda (2010).

3. Empirical Application
This section describes the data and variables and present some stylized facts in 3.1 and
presents the main model specification and its variations for validation in 3.2.

3.1. Data and variables

The outcome variable is the immigration inflow of each origin country to each OECD
destination country. The source is the international migration database (OECD, 2024). It is
measured in number of people per country and the inflow is used in the model as the ratio of
origin migrants per 100.000 habitants in the destination country. This standardized measure
helps to consider the issue from the destination’s policymaker point of view. This rate reflects
on average, for the full sample, the inflow (from a specific origin at a given year) of 5.9 migrants
for every 100,000 habitants of respective destination (Table 1).

The variable of interest, the regularisation policy, is a dummy variable taking the value
of 1 when an OECD country implemented the policy for a particular origin nationality and for
the year when the undocumented migrant was allowed to register in the program, 0 otherwise.
Not all regularisation programs are implemented for all origins countries. In this latter case, the

6
dummy variable takes the value of 1 only for those selected origins for this particular OECD
country and year, non-benefited countries remain as 0 for that destination-year.

Table 1. Descriptive statistics

(1) (2)
Total Without-outliers
mean sd max min mean sd
Migrant inflow rate 5.90 51.70 15509.31 0.00 5.83 32.49
Regularisation policy 0.11 0.31 1.00 0.00 0.12 0.33
GDP, per capita (origin) 12921.5 19947.3 228667.9 122.88 9941.97 14150.4
6 8 4 7
GDP, per capita 35797.2 20668.5 112417.8 3934.9 37150.8 20781.8
(destination) 6 3 8 4 3 6
Network of migrants 122.76 880.87 57707.49 0.00 127.69 911.11
Observations 222849 172705
Source: Authors’ elaboration.

We have assembled the dataset of regularization policies (RegMig) using three main
sources: First, the “Regularisation in Europe” (REGINE) project produced for the European
Commission’s Directorate-General Justice, Freedom and Security (Baldwin-Edwards & Kraler,
2009); second, the University of Oxford’s International Migration Institute data (DEMIG,
2015) and third, the OECD’s international migration reports (from (OECD, 1997) to (OECD,
2023)). Additionally, the RegMig dataset has been complemented using information from
Apap, De Bruycker, & Schmitter (2000), Levinson, (2005), IADB, Acosta, & Harris (2022),
McDonald (2009), Dehm & Vogl (2022) and Sunderhaus (2007). As a result, this novel dataset
allows to identify the implementation of any regularisation policy for each of the 38 OECD
countries. Since only 27 of them have had one regularization over the period considered, those
are the countries included in the empirical application.

RegMig records the year when the policy allowed undocumented migrants to register in
the program and the origin nationalities included (See Table A1). When the number of
beneficiaries is available, it has also been recorded in the dataset. The total number of
beneficiaries for this period is 13.5 million, but the real number is surely larger as for many
programs it was not available. The dataset comprises 80 years (1944-2023). A dummy variable
for the existence of the program is our variable of interest. This policy was present in 11% of
the observations and 12% when the sample is limited to origin countries not being more than
twice as rich as the destination (Table 1).

The control variables have been selected in accordance with the previous research on
migrations as Czaika & Reinprecht (2022) and Docquier, Peri, & Ruyssen (2014), among
others. First, the income ratio is measured as real GDP per capita of the specific origin divided
7
by the one of the destination countries using the World Bank Open Data (World Bank Group,
2024). The data shows that, in average, origin’s income is 36.09% the income of destinations.
Second, the network of migrant is built dividing the stock of migrants (OECD, 2024) by the
population at origin (World Bank Group, 2024). Second, the average network of migrant in a
destination country is 122.76 for every 100,000 habitants of the origin country. This variable
proxies for the likelihood of someone in origin having a familiar or friendship connection living
in a specific destination, and we expect that larger the diaspora is, greater will be the odds and
the pulling effect. Despite having data for regularisation since 1944 and GDP since 1960; the
yearly bilateral migration data availability limits the sample to the period 1996-2022.

Figure 1. Number of years a regularisation program was implemented in OECD countries

(1996-2022)

Source: RegMig Dataset, Authors’ elaboration.

Figure 1 shows how often OECD countries implemented the policy based on our
regularisation policy dataset (Table A1). It includes years in which a regularization program
was applied to one or several origin nationalities, to few beneficiaries, to rejected asylum
seekers or massive regularisations. It excludes regularisation mechanisms as they are permanent
legal channels.

3.2. Model Specification

The migration literature has commonly used gravity models to estimate the determinants
of bilateral migration flows, as in some recent works including Czaika & Parsons (2017),
Mayda (2010), Bertoli & Moraga (2013), and Beine, Bertoli, & Fernández‐Huertas Moraga
(2016). This literature explains how gravity models account for multilateral resistance in
migration and avoid biased estimations by incorporating several sets of controls to the models.
In this regard, Anderson & Yotov (2012) and Yotov (2022) specifically address the issue in
8
trade models by using origin-time and destination-time fixed effects to account for time-variant
trade multilateral resistances. Additionally, the Poisson pseudo maximum likelihood (PPML)
estimator is used to to keep zeros in migration in the estimations and deal with heteroscedastic
issues, avoiding biases in the estimated coefficients (Giménez-Gómez, Walle, & Zergawu,
2019) & (Kang, 2021).

𝑀𝑖𝑔𝐼𝑛𝑓𝑙𝑜𝑤𝑖𝑗𝑡 = exp (𝛽1 𝑅𝑒𝑔𝑢𝑙𝑎𝑟𝑖𝑠𝑎𝑡𝑖𝑜𝑛 𝑖𝑗,𝑡−1 + 𝛽2 𝑖𝑛𝑐𝑜𝑚𝑒𝑖𝑗,𝑡−1 + 𝛽3 𝑏𝑖𝑙𝑎𝑡𝑁𝑒𝑡𝑤𝑜𝑟𝑘𝑖𝑗,𝑡−1 +

𝜃𝑖𝑡 + 𝜋𝑗𝑡 + 𝛼𝑖𝑗 ) ∗ 𝜈𝑖𝑗𝑡 (1)

Were 𝑀𝑖𝑔𝐼𝑛𝑓𝑙𝑜𝑤𝑖𝑗𝑡, is the migration inflow rate of origin i to country j in year t is

determined by our variable of interest, 𝑅𝑒𝑔𝑢𝑙𝑎𝑟𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑖𝑗,𝑡−1, a dummy for regularisation policy
being implemented in year t-1. The control variables are: 𝑖𝑛𝑐𝑜𝑚𝑒𝑖𝑗,𝑡−1 , the income ratio between
origin/destination and 𝑏𝑖𝑙𝑎𝑡𝑁𝑒𝑡𝑤𝑜𝑟𝑘𝑖𝑗,𝑡−1 the network of nationals of origin i in country j, both
variables lagged one year. Additionally, the specification includes High Dimensional Fixed
Effects ( 𝜃𝑖𝑡 , 𝜋𝑗𝑡 , 𝛼𝑖𝑗) that control for multilateral resistance (𝜃𝑖𝑡 , 𝜋𝑗𝑡 ), omitted time-varying
(𝜃𝑖𝑡 , 𝜋𝑗𝑡) and time-invariant variables (𝛼𝑖𝑗 ), such as the usual gravity controls, namely, distance,
common language, a past colonial relationship. Equation 1 is the main model specification,
which is estimated with PPML. As robustness, a linear gravity model has also been estimated
(results available on request).

4. Empirical Results
This section presents the main result in sub-section 4.1 and investigates the existence
of heterogeneities in subsections 4.2 and 4.3.

4.1 Main results

The results comparing model alternatives is shown in Table 2. Column 1 shows a
traditional gravity model. Column 2 results are obtained from a gravity model specification
with dyadic fixed effects, while Columns 3 and 4 follow our preferred HDFE model with
different regularisation lags, following the recommended practices for gravity estimations of
Yotov (2022). The variable of interest, regularisation policy, in first row attempts to capture the
impact of a regularisation implemented by year t-1 on migration inflows of years t. In all cases,
the control variables included are the income ratio origin/destination and the network of origin
in destination. In column 1, as dyadic FE are not included, traditional gravity variables are
added as regressors.

9
 Table 2. Gravity model results of regularisation policies and migration.
PPML with different specifications
(1) (2) (3) (4)
Traditional 2-way FE HDFE HDFE
Gravity (3 lags)
Regularisation policy (t-1) 0.360 0.0758 ** -0.0116 -0.0575
(0.250) (0.0321) (0.152) (0.140)
Regularisation policy (t-2) -0.183
(0.121)
Regularisation policy (t-3) 0.130
(0.149)
GDPpc orig-dest (t-1) -0.0529 -1.485*** 0.227 0.225
(0.0444) (0.291) (0.209) (0.210)
Log of Network (t-1) 0.800*** 0.333*** 0.392*** 0.387***
(0.0132) (0.0302) (0.0403) (0.0409)
Contiguity -0.101
(0.0879)
Common official prim lang 0.235***
(0.0880)
Lang spoken both countries 0.0307
(0.0818)
Colonial relationship 0.104
(0.115)
Common colonizer post1945 0.677***
(0.182)
Colonial rel post1945 -0.343***
(0.117)
Were or are same country 0.0967
(0.0944)
Log Distance -0.0160
(0.0361)
RegulTot -0.110
(0.298)
Pseudo Rsquared 0.883 0.874 0.911 0.912
Observations 60128 60631 60249 59655
Dyadic FE No Yes Yes Yes
Year FE No Yes No No
Destination x Year FE Yes No Yes Yes
Origin x Year FE Yes No Yes Yes
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination population.
RegulTot is the marginal effect of the coefficients estimates of the 3 lags of regularisation policy dummy in column
4. Controls: GDPpc ratio origin/destination & Network of origin country at destination.

The 2-way fixed effects specification finds a positive and significant pull effect of
regularisation on migration. This result is in accordance to the expectations in the literature.
However, in the traditional gravity and in the preferred specification, HDFE, the estimated
coefficient of the target variable is not statistically significant. Although, HDFE is preferred to
specification used to obtain results in columns 1 and 2, which might suffer from an omitted

10
variable bias, it could also be that since many programs are common for all origin nationalities
the effect is absorbed by the destination-time fixed effects. It could also be that there is some
heterogeneity in the effects, as will be investigated below.

Nevertheless, since the HDFE specification allows for unbiased estimates of

regularisation policy as it accounts for time-variant and time-invariant omitted variable biases
as push and pull factors as well as multilateral resistance (the influence of “competing” origin
and destinations), we investigate further the results of this specification for specific groups of
countries.

Concerning the control variables, a growing network (in t-1) is a significant pull factor
in all models, while the income ratio coefficient is only significant in the two-way FE
specification. The coefficients of gravity variables in column (1) indicate that sharing official
language and having a common colonizer, increase the inflow of migrants, while colonial
relationship has a negative impact on migration. Additionally, it is worth to highlight that the
lack of significance of regularisation as a pull factor is not driven by omitting additional lags as
is shown in columns 3 and 4. In the following, HDFE equations are based on column 3 as our
preferred model, when estimating the model with specific coefficients for several groups of
origin and destination countries.

Heterogeneity analysis for the main results is presented in what follows. The preferred
model is used to estimate the effect for specific groups. First, country pairs where origin
countries are twice as richer as destinations, 50% richer or just richer, are excluded in columns
2, 3 and 4, respectively of Table A2, presented in the Appendix. In general, the results are
similar to those obtained for the full sample.

4.2 Heterogeneity by income of origin and destination countries

All origin countries were classified by quartiles according to the average real GDP per
capita of the period under study (1996-2022). As a result, the impact of regularisation policy is
neither significant for all origin countries’ income quartiles as seen in Table A3 in the appendix.

Certainly, destination countries were also classified by above and below median
historical income as for origins. It was confirmed that regularisation policy could not be
considered a pull factor for destination countries below or above the median income as it is
shown in Table A4 in appendix. However, in beforementioned table, regularisation could be
considered a pull factor for Q1 income countries.

11
 Table 3. Regularisation policy as drivers for migration
Heterogeneity by origin-destination income groups
(1)
Migration inflow rate
Regularisation policy (t-1) Poor to Q1 0.257
(0.261)
Regularisation policy (t-1) Poor to Q2 -0.405**
(0.171)
Regularisation policy (t-1) Poor to Q3 -0.305*
(0.180)
Regularisation policy (t-1) Poor to Q4 -0.232***
(0.0717)
Regularisation policy (t-1) Rich to Q1 0.353**
(0.169)
Regularisation policy (t-1) Rich to Q2 -0.412***
(0.153)
Regularisation policy (t-1) Rich to Q3 -0.156
(0.177)
Regularisation policy (t-1) Rich to Q4 0
(.)
Pseudo Rsquared 0.911
Observations 60249
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01. Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination population.
Controls: GDPpc ratio origin/destination & logged network. Poor and Rich denote being below or above median
income for origin countries and Qs stand for income quartile for destination countries. Estimation method: PPML
with three sets of fixed effects: origin-time, destination-time, origin-destination.

Additionally, a dummy was created for every combination, below and above median
income (Poor and Rich) origin migrating to the four groups of destination countries split by
income quartiles. The impact was estimated for each group in Table 3. It confirmed that
regularisation is a pull factor for lower-income (Q1) destinations (Chile, Czech Republic,
Estonia, Hungary, Latvia, Lithuania, Mexico, Poland, Portugal, Slovakia, and Turkey)
especially driven by origin countries with higher income. It was also confirmed by Table A5
that shows that for origin countries of income quartile 1, 3 and 4, regularisation was a pull factor
towards lower-income destinations (Q1).

For instance, it means that, in average, a lower-income (Q1) destination implementing

the policy in year t is expected to experience an increase of 42.33% ([((exp^(0.353))-1)*100])
of the immigration inflow rate in the following year, ceteris paribus, for every above median
income origin country benefited from the regularisation.

To illustrate the results with an example, Chile regularised 210,000 undocumented

migrants in 2019 from all nationalities. The Colombian’ migrant inflow of 2019 was 114.15
Colombian migrants to Chile per every 100,000 of Chilean population, a total of 21,734
12
Colombians. The average impact of the regularisation on Colombians’ inflow is expected to be
a rate of 48.32 (114.15 * 0.4233). It implies that 9,200 additional Colombian migrants moved
to Chile in the year 2020 attributed to regularisation, in average and ceteris paribus. This impact
is similar, in average, for every above-median-income origin country with positive migration
inflow at the time of the implementation.

4.3 Heterogeneity by regions of origin and income of destination countries

The impact of the regularisations by regions4 of origin is mostly not significant (Table
A6 of appendix) as it was for income quartiles of origin countries (Table A3). Similar results
are found for all region when migrating to above or below median income destination (Table
A7).
Table 4. Regularisation policy as drivers for migration.
Heterogeneity by regions of origin and incomes at destination.
Target variable: (1) (2) (3) (4)
RP (Regularisation policy(t-1)) To Q1 To Q2 To Q3 To Q4
RP * east -0.326 0.0323 -0.302* 0.0204
(0.207) (0.209) (0.174) (0.0674)
RP * mena -0.00743 -0.284 -0.323 -0.528***
(0.282) (0.182) (0.201) (0.0936)
RP * latam 0.403 ** -0.260 -0.304 -0.197*
(0.167) (0.182) (0.198) (0.102)
RP * ssahara 0.187 -0.361* -0.331* -0.331***
(0.191) (0.187) (0.193) (0.0652)
RP * easiap -0.658 *** -0.428 *** -0.137 -0.0543
(0.227) (0.151) (0.191) (0.0907)
RP * sasia 0.505** -0.334* -0.301 -0.367***
(0.218) (0.187) (0.200) (0.119)
RP * west 0.231 -0.421** -0.110 0
(0.177) (0.182) (0.187) (.)
Pseudo Rsquared 0.911
Observations 60249
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination population.
All coefficients come from one estimation where the policy was interacted with each possible combination of
region of origin with income quartile of destination. Controls: GDPpc ratio origin/destination & logged Network.
Estimation method: PPML with three sets of fixed effects: origin -time, destination-time, origin-destination.

However, regularisation is a pull factor for countries of Latin America and South Asia
regions when migrating towards quartile-one income destinations, as seen in Table 4. Besides
these two cases, the pull factor effect was not found in any other region of origin toward any
other quartile income destinations.

4 The regions’ dummies follow the classification used by World Bank.

13
5. Robustness checks
5.1 Long term impact
The impact of a regularisation on migration inflow could materialize with some delay,
and hence only be observed after some year after the policy has been implemented. This can be
tested by incorporating to the model a series of policy dummies for years t to t-6 (Equation 2).
The marginal effect is estimated by testing for the significance of the linear combination of all
these dummies. However, it brought similar results as seen in Table A8. The HDFE
specifications with 6 and 8 lags are still reflecting a not significant pull factor effect.

𝑀𝑖𝑔𝐼𝑛𝑓𝑙𝑜𝑤𝑖𝑗𝑡 = exp ((∑𝐾=6

𝑘=0 𝛽1𝑘 𝑅𝑒𝑔𝑢𝑙𝑎𝑟𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑖𝑗,𝑡−𝑘 ) +

𝛽6 𝑖𝑛𝑐𝑜𝑚𝑒𝑖𝑗,𝑡−1 + 𝛽7 𝑏𝑖𝑙𝑎𝑡𝑁𝑒𝑡𝑤𝑜𝑟𝑘𝑖𝑗,𝑡−1 + 𝜃𝑖𝑡 + 𝜋𝑗𝑡 + 𝛼𝑖𝑗 ) ∗ 𝜈𝑖𝑗𝑡 (2)

Table A9 shows that it holds for different country-pair restrictions. Furthermore, it can
be seen the impact of each regularisation lag which is different across time as (Czaika & De
Haas, 2017) found for the impact of visas. Certainly, the short-term impact of regularisation on
migration is unsignificant or negative until lag 4, and it becomes positive and significant for
lags 5 and 6. Overall, the impact is not significant as it is shown with RegulTot that is the
accumulated impact of regularisation and its lags.

The pull effect for origin countries by income groups or regions continues to be not
significant as seen in Tables A10 & A11, similar to results of Tables A3 & A6. On the other
hand, the accumulated effect of regularisation on migration inflow by income of destination
countries is in this case significant for below median income destinations (Table A12) similar
to Table A4 where regularisation was a pull factor in the following year for quartile one of
income for destination countries.

Above all, the result is still similar when grouping origin and destination countries by
income (Table 5). The accumulated impact of regularisation over the long period could be
considered a pull factor for above-median income origins when migrating to below-median
income OECD countries (Chile, Czech Republic, Estonia, France, Greece, Hungary, Italy,
Japan, Republic of Korea, Latvia, Lithuania, Mexico, Poland, Portugal, Slovakia, Slovenia,
Spain and Turkey). It seems to be specially driven by countries of income quartile 25 , Table
A13. In the long term the migration controls of below-median income destination countries
appear to be insufficient to counterbalance the pull factor effect of the policy for above-median

5 France, Greece, Italy, Japan, Republic of Korea, Slovenia and Spain.

14
income origin countries nor for any geographical region of origin (Table A14). Therefore, it
could be discarded that the main results are driven by omitting the long-term effects of the
regularisation policy on immigration inflow rate. In fact, this robustness check increases the
confidence of main results of section 4.

Table 5. Regularisation policy as long-term driver for migration.

Heterogeneity by origin-destination income groups
(1) (2) (3) (4)
Poor to Poor Poor to Rich Rich to poor Rich to Rich
RegulTot 1.012 -0.815 1.294** -0.679
(0.623) (0.914) (0.581) (0.903)
Pseudo Rsquared 0.914
Observations 56440
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination population.
RegulTot is the marginal effect of the coefficients estimates of regularisation policy dummy and its lags interacted
with respective income group. Controls: GDPpc ratio origin/destination & logged network. Each column shown
the marginal effect of each origin-destination income group and were estimated in one equation. Estimation
method: PPML with three sets of fixed effects: origin-time, destination-time, origin-destination. Poor and Rich
denote being below or above median income in the respective origin or destination group of countries.

5.2 Impact of general regularisations

An additional concern arises from the fact that previous results reflect the impact of
regularisations implemented to specific destinations and not general regularisations
implemented to all origins (which are most common in our dataset). Indeed, in our specification
with HDFE the impact of all regularisations implemented by a destination country j in year t to
all origins i is absorbed by the destination-year fixed effects (𝜋𝑗𝑡 ).
A technique to avoid that the destination-year fixed effects absorb the impact of the
policy when it was implemented to all countries with no distinction consists on interacting the
policy dummy with the relative distance (distance / mean distance) of country pair. Though for
a regularisation implemented to all countries this interaction is not going to be equal to (𝜋𝑗𝑡 )
because every country pair takes the value of the relative distance instead of taking the value of
1. The relative distance varies from 0.115 to 2.78 with 50% of the sample varying close to 1
(from 0.506 to 1.354).

The results in Table A15 are very similar to those of section 3 (Table 2). HDFE model
for full sample finds no significant pull effect for regularisation while two-way fixed effect
model does find it. There was no evidence for pull factor in heterogeneity analysis for broad
groups (Tables A16 & A17). However, when examining heterogeneity by income quartile
groups of both destinations and origins (Table 6), the results are similar to Table A5 of our
basic specification developed in previous section. There is found evidence of pull factor for
15
destinations of Q1 when origin was Q1 and Q4 as it was in Table A5. However, in Table 6 it
was not significant for Q3 origins to Q1 destinations. In contrast, now the policy is a pull factor
for migrants from Q4 to Q3 destinations.

Table 6. Regularisation policy (interacted with distance) as driver for migration.

Heterogeneity by origin-destination income groups
Target variable: (1)
RPdist (Regularisation policy(t-1)*Dist) Migrant inflow rate
RPdist * Q1toQ1 0.399***
RPdist * Q1toQ2 -0.269**
RPdist * Q1toQ3 -0.159**
RPdist * Q1toQ4 -0.445***
RPdist * Q2toQ1 -0.274
RPdist * Q2toQ2 -0.0938
RPdist * Q2toQ3 -0.0382
RPdist * Q2toQ4 -0.164
RPdist * Q3toQ1 0.117
RPdist * Q3toQ2 -0.0765
RPdist * Q3toQ3 -0.0380
RPdist * Q3toQ4 -0.170***
RPdist * Q4toQ1 0.412*
RPdist * Q4toQ2 -0.375**
RPdist * Q4toQ3 0.115***
RPdist * Q4toQ4 0.117
Pseudo R squared 0.911
Observations 59482
Notes: Standard errors are clustered at the country pair. * p < 0.1, ** p < 0.05, *** p < 0.01 . Dependent variable:
Rate of migrants' inflow from each origin per 100000 hab in destination population. Controls: GDPpc ratio
origin/destination & logged network. Qs stand for income quartile for origin and destination countries Estimation
method: PPML with three sets of fixed effects: origin-time, destination-time, origin-destination.

Therefore, the results from section 3 holds when the impact of all regularisations are
included in the coefficient estimated. The reaction of origin countries to the pull factor is similar
when it is implemented to all origins than to a specific country or group of countries.

5.3 Addressing endogeneity with Instrumental Variable

Furthermore, there is the concern that the regularisation policy being lagged in the model
could not be enough to address endogeneity of our variable of interest. Therefore, an additional
empirical strategy is implemented. Instrumental Variables is a common method implemented
in economic research. The main obstacle is to find valid instrument.

Following Sunderhaus (2012), the year after election and having a new executive
government is associated with the policy implementation and those are not expected to be a
strong predictor for migration inflow. Furthermore, it could be argued that the executive control

16
of legislative chambers allows them to introduce such a policy with no relation to migration.
Other political variables where tested but these fitted the better.

The political variables used are from Scartascini, Cruz, & Keefer (2021) which was part
of the Political sources recommended by Princeton University (2024). It was incorporated to
the PPML gravity model with HDFE but for the first stage the destination-year fixed effect was
omitted and instead year fixed effects were incorporated. The latest was necessary because
political variables are at destination-year level.

It is relevant to remind that the Stata’s command used in previous sections (ppmlhdfe)
does not support instrumental variables. Therefore, it was implemented “manually”. The
column 1 of Table A18 reflects the first stage with the regularisation policy as dependent
variable, the right-hand side of the equation are the regular controls and the battery of fixed
effects (county-pair, year and origin-year). The instrument the lagged dummy for election year.
The second stage’s results are similar to our basic mod el (Table 2, column 3), there is not pulling
effect for regularisation for OECD countries in general.

This instrument was chosen as it reports the higher F (Table A19) in a different
framework (without PPML) and command (ivreghdfe). Furthermore, the validity of instruments
is a relevant concern when instrumental variables is implemented. Table A19, show the test of
these instruments. For all three specifications, the F statistic is close to the rule of thumb (10),
the underidentification (no correlation of instruments with endogenous variable) is rejected as
well as the weak identification is rejected with a high Cragg-Donald Wald F statistic (16.92 to
452.8).

Nonetheless, Table A20 shows an attempt to probe the instruments fulfil the
assumptions of independence and exclusion. There, in column 1 the residual of the second stage
of results shown in Table A18 was regressed by the predicted regularisation variable, the
controls and the instrument. There, the instrument was not significant. It allows to infer that the
instrument is not related to any omitted variable left in the residuals, which could reduce the
risk of a violation of independence assumption. Furthermore, column 2 shows that the
instrument is not significantly correlated with migration inflow rate but through regulation
policy. Though, it could be said with some confidence that the exclusion assumption appears to
hold.

Assuming the instrument fulfil the requirements, the predicted variable was interacted
with the origin-destination income group dummies. Table 7 confirmed previous heterogeneous

17
results. The exogenous implementation of regularisation policy due to previous year elections
cause a pull factor for migrants toward below-median income destination countries. It is
stronger for migrants from above-median income to Q2 destinations and it exist (but weaker)
for all origin income migrants to Q1 destinations.

Table 7. Regularisation policy as drivers for migration.

Instrumental Variables by origin-destination income groups
(1)
Migration Inflow Rate
Estimated regularisation (t-1) * PoortoQ1 0.807*
(0.365)
Estimated regularisation (t-1) * PoortoQ2 0.666
(0.555)
Estimated regularisation (t-1) * PoortoQ3 -0.0514
(0.293)
Estimated regularisation (t-1) * PoortoQ4 0.00243
(0.258)
Estimated regularisation (t-1) * RichtoQ1 2.406*
(0.945)
Estimated regularisation (t-1) * RichtoQ2 2.065**
(0.772)
Estimated regularisation (t-1) * RichtoQ3 -0.624
(0.432)
Estimated regularisation (t-1) * RichtoQ4 -0.349
(0.439)
Pseudo R squared 0.912
Observations 55488
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination population.
Controls: GDPpc ratio origin/destination & Network of origin country at destination. Estimation method: PPML
with three sets of fixed effects: origin-time, destination-time, origin-destination. Regularisation policy was
instrumented by a dummy for election year.

6. Conclusions
This paper investigates whether regularisation policy is effectively a pull factor. The
question is relevant because an increased future migration is one of the main worries of
policymakers when deciding solutions (as a regularisation policy) for a high presence of
undocumented migrants in the respective country. Consequently, this paper assessed this
question by estimating a gravity model of migration augmented with a regularization dummy
variable, while controlling for income ratios, network effects and the battery of fixed effects.
The model is estimated with a PPML estimator that is able to account for zero migration flows
and deals with other econometric issues. Furthermore, the heterogeneity of the policy was
considered by estimating the main empirical model allowing for heterogeneity in the effects by
groups of origin and destination countries classified by income and geographical regions.
18
 Given that the results obtained by previous research indicate that regularization
programs are a pull factor only in some cases, it is very relevant to reexamine the effects of
those polices. Whereas most of the previous studies dealt with the impact of the policy when
implemented in only one or two countries at most, this paper extends the question to all OECD
receiving countries. On the other hand, it uses an improved empirical strategy compared to
Wehinger (2014) and considered broader spectrum of OECD destinations than Ortega & Peri
(2009) and Mayda (2010).

Generally, our main results indicate that the pull factor of regularization programs
materialize depending on destination countries’ income. It seems to reflects the state capacity
for border control and migration regulation enforcement at the destination country. Richer
destinations are able to control better the migration inflow after a regularisation while those
lower-income (Q1) countries are not good enough. These results are robust when considering
long term impact, including regularisations for all origins and addressing endogeneity concerns
with instrumental variables.

It is acknowledged that current paper contains a number of weaknesses and limitations.

First, the dependent variable is a proxy variable of what it should be; OECD’s migration inflow
data is different from illegal entrees and overstayers. Second, the attempt to address the
endogeneity concern with instrumental variables is always debatable in economic literature.
Therefore, it cannot be claimed to state fully causal estimates. Third, the regularisation policy
is a dummy, though it is equalizing a regularisation of several hundred thousand to another of
few hundreds benefited of the program. Therefore, this could be considered a measurement
error. The number of beneficiaries is reflected in the dataset but it was not possible to collect
this information for all regularisations. Consequently, the magnitude of the regularisation has
not yet been exploited. We leave this for further research.

Despite the abovementioned weaknesses and limitations, these results could be seen as
a worthy contribution to the scarce literature on the evaluation of regularisation policy as a pull
factor. Especially, the current findings are highly relevant for policymakers. They should not
avoid this policy for above-median income destinations worried about regularization policy as
a pull factor. Furthermore, for those below-median income destinations, policymakers should
accompany this policy with real improvement on border control and the migration regulations
dealing with overstayers. All these, assuming that increased migration is not desirable, which
is not always the case and will be more so in the future with ageing population in most
developed countries.
19
References
Alvarez, J., Arena, M. M., Brousseau, A., Faruqee, M. H., Corugedo, E. W., Guajardo, M. J., . . .
Yepez, J. (2022). Regional spillovers from the Venezuelan crisis: Migration flows and their
impact on Latin America and the Caribbean. (Western Hemisphere Department (Series) ed.).
Washington, DC: International Monetary Fund.
Anderson, J., & Yotov, Y. (2012). Gold standard gravity. (N. B. Research, Ed.) NBER working Paper,
17835. doi:DOI 10.3386/w17835
Apap, J., De Bruycker, P., & Schmitter, C. (2000). Regularisation of Illegal Aliens in the European
Union-Summary Report of a Comparative Study. European Journal of Migration and Law,
2(3 & 4), 263-308. Retrieved from
https://heinonline.org/HOL/Page?handle=hein.journals/ejml2&id=273
Bahar, D., Ibáñez, A. M., & Rozo, S. V. (2021). Give me your tired and your poor: Impact of a large-
scale amnesty program for undocumented refugees. Journal of Development Economics, 151.
doi:https://doi.org/10.1016/j.jdeveco.2021.102652.
Baldwin-Edwards, M., & Kraler, A. (2009). REGINE-Regularisations in Europe. Amsterdam
University Press.
Beine, M., Bertoli, S., & Fernández‐Huertas Moraga, J. (2016). A practitioners’ guide to gravity
models of international migration. The World Economy, 39(4), 496-512.
doi:https://doi.org/10.1111/twec.12265
Bertoli, S., & Moraga, J. F. (2013). Multilateral resistance to migration. Journal of development
economics, 102,, 79-100. doi:https://doi.org/10.1016/j.jdeveco.2012.12.001
Borjas, G. J., & Edo, A. (2023). Monopsony, Efficiency, and the Regularization of Undocumented
Immigrants. National Bureau of Economic Research., No. w31457. doi:DOI: 10.3386/w31457
Czaika, M., & De Haas, H. (2017). The effect of visas on migration processes. International Migration
Review,, 51(4), 893-926.
Czaika, M., & Parsons, C. R. (2017). The gravity of high-skilled migration policies. Demography,
54(2), 603-630. doi:https://doi.org/10.1007/s13524-017-0559-1
Czaika, M., & Reinprecht, C. (2022). Migration drivers: Why do people migrate. In P. Scholten,
Introduction to migration studies: An interactive guide to the literatures on migration and
diversity (pp. 49-82). Rotterdam: Springer. doi:https://doi.org/10.1007/978-3-030-92377-8
Dehm, S., & Vogl, A. (2022). Immigration amnesties in Australia: Lessons for law reform from past
campaigns. The Sydney Law Review, 44(3), 381-413.
doi:https://search.informit.org/doi/10.3316/informit.825807618134080
Deiana, C., Giua, L., & Nisticò, R. (. (2024). Legalization and Long-Term Outcomes of Immigrant
Workers. (M. Center for Economic Studies and ifo Institute (CESifo), Ed.) CESifo Working
Paper, No. 11026. doi:https://dx.doi.org/10.2139/ssrn.4783706
DEMIG. (2015). DEMIG POLICY, version 1.3, Online Edition. (U. o. Oxford: International Migration
Institute, Editor) Retrieved August 2024, from https://www.migrationinstitute.org/data/demig-
data/demig-policy-1
Docquier, F., Peri, G., & Ruyssen, I. (2014). The cross‐country determinants of potential and actual
migration. International Migration Review, 48, S37-S99. doi:
https://doi.org/10.1111/imre.12137

20
Elias, F., Monras, J., & Vazquez-Grenno, J. (2024). Understanding the effects of granting work
permits to undocumented immigrants. Journal of Labor Economics.
doi:https://doi.org/10.1086/730122
Fasani, F. (2018). Immigrant crime and legal status: Evidence from repeated amnesty programs.
Journal of Economic Geography, 18(4), 887-914. doi:https://doi.org/10.1093/jeg/lby028
Finotelli, C., & Arango, J. (2011). Regularisation of unauthorised immigrants in Italy and Spain:
determinants and effects. Documents d'anàlisi geogràfica,, 57(3), 495-515. Retrieved from
https://www.raco.cat/index.php/DocumentsAnalisi/article/view/248439
Giménez-Gómez, J.-M., Walle, Y. M., & Zergawu, Y. Z. (2019). Trends in African migration to
Europe: Drivers beyond economic motivations. Journal of Conflict Resolution, 63(8), 1797-
1831. doi:https://doi.org/10.1177/0022002718823907
IADB, Acosta, D., & Harris, J. (2022). Regímenes de política migratoria en América Latina y el
Caribe: inmigración, libre movilidad regional, refugio y nacionalidad. Washington, DC:
IADB. doi:http://dx.doi.org/10.18235/0004362
Kang, Y. D. (2021). Refugee crisis in Europe: determinants of asylum seeking in European countries
from 2008–2014. Journal of European Integration, 43(1), 33-48.
doi:https://doi.org/10.1080/07036337.2020.1718673
Kossoudji, S. A., & Cobb-Clark, D. A. (2002). Coming out of the shadows: Learning about legal
status and wages from the legalized population. Journal of Labor Economics, 20(3), 598-628.
doi:https://doi.org/10.1086/339611
Larramona, G., & Sanso‐Navarro, M. (2016). Do regularization programs for illegal immigrants have
a magnet effect? Evidence from Spain. The Manchester School, 84(2), 296-311.
doi:https://doi.org/10.1111/manc.12099
Levinson, A. (2005). The regularisation of unauthorized migrants: Literature survey and country case
studies. Oxford: Centre on Migration, Policy and Society, University of Oxford.
Machado, J. (2017). Dealing with undocumented immigrants: the welfare effects of amnesties and
deportations. Journal of Demographic Economics, 83(4), 445-492.
doi:https://doi.org/10.1017/dem.2017.14
Mayda, A. M. (2010). International migration: A panel data analysis of the determinants of bilateral
flows. Journal of population economics, 23, 1249-1274. doi: https://doi.org/10.1007/s00148-
009-0251-x
McDonald, J. (2009). Migrant illegality, nation-building, and the politics of regularization in Canada.
Refuge: Canada's Journal on Refugees, 26(2), 65-77. Retrieved from
https://heinonline.org/HOL/Page?handle=hein.journals/rfgcjr26&id=258
Méndez, F., Sepúlveda, F., & Valdés, N. (2016). Legalization and human capital accumulation.
Journal of Population Economics, 29, 721-756. doi:https://doi.org/10.1007/s00148-016-0585-
0
Molinari, R., Impicciatore, R., & Ortensi, L. E. (2023). Traces in the shadow: Occupational outcomes
of previously undocumented migrants in Italy. International Migration, 61(6), 75-101.
doi:https://doi.org/10.1111/imig.13144
OECD. (1997). Trends in International Migration 1997: Continuous Reporting System on Migration,.
Paris: OECD Publishing. doi:https://doi.org/10.1787/migr_outlook -1997-en.

21
OECD. (2018). International Migration Outlook 2018. Paris: OECD Publishing.
doi:https://doi.org/10.1787/migr_outlook-2018-en
OECD. (2023). International Migration Outlook 2023. Paris: OECD Publishing.
doi:https://doi.org/10.1787/b0f40584-en
OECD. (2024, July 24). OECD Data Explorer. International Migration Database. Paris, France.
Retrieved August 2024, from https://data-explorer.oecd.org/
Orrenius, P. M., & Zavodny, M. (2003). Do amnesty programs reduce undocumented immigration?
Evidence from IRCA. Demography, 40, 437-450. doi:https://doi.org/10.1353/dem.2003.0028
Orrenius, P. M., & Zavodny, M. (2015). The impact of temporary protected status on immigrants'
labor market outcomes. American Economic Review, 105(5), 576-580. doi:DOI:
10.1257/aer.p20151109
Ortega, F., & Peri, G. (2009). The causes and effects of international migrations: Evidence from
OECD countries 1980-2005. National Bureau of Economic Research., No. w14833.
Pinotti, P. (2017). Clicking on heaven’s door: The effect of immigrant legalization on crime. American
Economic Review, 107(1), 138-168. doi:DOI: 10.1257/aer.20150355
Princeton University. (2024). “Elections and Voting Data Guide: United States (U.S.) and
International”. Retrieved August 2024, from https://libguides.princeton.edu/elections/foreign
Scartascini, C., Cruz, C., & Keefer, P. (2021). The database of political institutions 2020 (DPI2020).
Washington, DC: Inter-American Development Bank Research Department.
doi:http://dx.doi.org/10.18235/0003049
Sunderhaus, S. (2007). Regularization programs for undocumented migrants. Migration Letters, 4(1),
65-76.
Sunderhaus, S. (2012). Regularization progrmans for undocumented migrants: A global survey on
more than 60 legalizations in all continents. Saarbrücken, Germany: AV Akademikeverlang
Gmbh & Co. KG.
Wehinger, F. (2014). Do amnesties pull in illegal immigrants? An analysis of European apprehension
data. International Journal of Migration and Border Studies, 1 (2), 231-246.
doi:https://doi.org/10.1504/IJMBS.2014.066312
Wong, T. K., & Kosnac, H. (2017). Does the legalization of undocumented immigrants in the US
encourage unauthorized immigration from Mexico? An empirical analysis of the moral hazard
of legalization. International Migration, 55(2), 159-173. doi:
https://doi.org/10.1111/imig.12319
World Bank Group. (2024). World Bank Open Data. Retrieved July 2024, from
https://data.worldbank.org/
Yotov, Y. (2022). Gravity at sixty: The Workhorse Model of Trade. CESifo Working Papers, 9584.
doi:https://dx.doi.org/10.2139/ssrn.4037001

22
A. Appendix
Table A1. RegMig.
List of regularisation programs implemented by OECD countries (1945-2023).
Number of
OECD country Origin countries Year Beneficiaries
Australia All 1974 367
Australia All 1976 7,861
Australia All 1980 14,000
Austria All 1954
Austria All 1990 30,000
Belgium All 1974-1975 7,448
Belgium All 1981
Belgium All 1995-1999 6,137
Belgium All 2000 52,000
Belgium All 2001-2004 5,644
Belgium All 2005-2007 17,070
Belgium All 2009 25,000
Belgium All 2020
Belgium All 2022
Canada China 1962-1972 12,000
Canada All 1973 39,000
Canada Haiti 1981 4,000
Canada All 1983
Canada All 1985-1986 1,000
Canada All 1994-1998 3,000
Canada Algeria 2002 900
Chile All 1998 62,000
Chile All 2003
Chile All 2007-2008 44,000
Chile All 2018-2019 210,000
Chile All 2021-2022 100,000
Colombia All 2008-2009
Colombia Venezuela 2018-2023 2,942,003
Costa Rica All 1990-1994
Costa Rica All 1998
Costa Rica All 2012
Costa Rica All 2020
Costa Rica Venezuela, Nicaragua & Cuba 2021
Costa Rica All 2022 18,000
Costa Rica Venezuela, Nicaragua & Cuba 2023
Czech Republic All Never

23
Denmark All Never
Estonia All Never
Finland All 2020
France All 1945
France All 1948
France All 1956
France Portugal 1968
France All 1973 40,000
France All 1979
France All 1981-1982 121,100
France All 1991 15,000
France Sub-Saharan countries & Algeria 1995
France All 1997-1998 77,800
France All 2006-2009 8,000
France All 2012-2016 101,000
France All 2023
Poland, Hungary, China, Ethiopia, Afghanistan,
Germany Iran, Lebanon, Sri Lanka & Turkey 1991
Germany Angola, Mozambique & Vietnam 1993
Germany Pakistan & Turkey 1994
Germany All 1996-1997 7,856
Germany All 1999-2000 23,000
Germany Yugoslavia & Bosnia and Herzegovina 2001
Germany All 2005-2009 180,000
Germany All 2011
Germany All 2023
Greece All 1998-1999 147,700
Greece All 2001-2002 351,000
Greece All 2005 142,000
Greece All 2007 19,979
Greece All 2011
Greece All 2018-2020
Hungary All 1994
Hungary All 2004
Iceland All Never
Ireland All 2009
Ireland All 2018-2019
Ireland All 2022 17,000
Israel All Never
Italy All 1982 5,000
Italy All 1986-1987 105,000
Italy All 1990 217,626
Italy All 1995-1996 244,500

24
Italy All 1998-1999 217,100
Italy All 2002-2003 634,700
Italy All 2009 233,000
Italy All 2012 115,988
Italy All 2020-2022 207,000
Japan All 2000
Korea All 1992-1994
Korea China 2002
Korea All 2004
Korea All 2021-2022
Latvia All Never
Lithuania All 1996 51
Lithuania All 1999 157
Lithuania All 2004 77
Luxembourg Spain & Portugal 1986 1,100
Luxembourg Bosnia and Herzegovina 1994 470
Luxembourg Yugoslavia 1995 996
Luxembourg All 1996 1,500
Luxembourg All 2001 1,839
Luxembourg All 2013
Mexico Guatemala 1999 150,000
Mexico All 2000-2001 2,600
Mexico All 2004-2008 4,000
Mexico All 2011 1,800
Mexico All 2016
Mexico All 2018
Netherlands All 1964
Netherlands All 1975 15,000
Netherlands All 1978
Netherlands All 1980 1,800
Netherlands All 1991-1995 2,106
Netherlands All 1999 2,200
Netherlands All 2004 2,300
Netherlands All 2007 26,000
Netherlands All 2019
New Zealand Fiji, Tonga & Samoa 1976
New Zealand All 2000-2001
Norway All 2007
Norway All 2021
Poland All 2003 2,747
Poland All 2007 177
Poland All 2012 4,500
Portugal All 1992-1993 38,400
25
Portugal All 1996 31,000
Portugal All 2001 123,700
Portugal Brazil 2003 19,408
Portugal All 2004 19,261
Portugal Brazil 2005
Portugal All 2019-2020
Slovak Republic All Never
Slovenia All Never
Spain All 1985 43,800
Spain All 1991 110,100
Spain All 1994
Spain All 1996 21,300
Spain All 2000-2001 398,500
Spain All 2004-2006 568,700
Sweden All 2006 17,000
Sweden All 2020
Switzerland All 2000-2001 17,323
Turkey All Never
United Kingdom Common Wealth & Pakistan 1974-1978 2,274
United Kingdom All 1987
United Kingdom All 1989 3,750
United Kingdom All 1998-2000 5,900
United Kingdom All 2004
United States All 1952
United States All 1987-1988 2,684,900
United States All 1995-1998 500,000
El Salvador, Guatemala, Haiti, Nicaragua,
Belarus, Bulgaria, Czech Republic, Hungary,
Moldova, Poland, Romania, Slovakia, Russian
United States Federation, Ukraine & Cuba 1997-1998 405,000
United States Cuba, Haiti & Nicaragua 1999-2000 1,075,000
United States All 2000-2002 400,000
El Salvador, Haiti, Nicaragua, Ukraine, South
Sudan, Nepal, Sudan, Afghanistan, Cameroon,
Syria, Venezuela, Myanmar, Ethiopia, Yemen,
United States Somalia & Honduras 2022-2023

26
 Table A2. Regularisation policy as driver for migration.
PPML-HDFE with different income ratios restrictions
(1) (2) (3) (4)
Full sample Inc_o_d <2 Inc_o_d <1.5 Inc_o_d <1
Regularisation policy(t- -0.0116 0.0324 -0.0116 -0.0244
1)
(0.152) (0.143) (0.140) (0.144)
GDPpc orig-dest (t-1) 0.227 0.664** 0.259 -0.399
(0.209) (0.333) (0.525) (0.742)
Log of Network (t-1) 0.392*** 0.394*** 0.399*** 0.404***
(0.0403) (0.0381) (0.0386) (0.0379)
Pseudo Rsquared 0.911 0.913 0.915 0.918
Observations 60249 57195 55605 52096
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. Controls: GDPpc ratio origin/destination & network. Column 2 restrict the sample to those
observations where per capita income of origin country is not more than double destination country. Column 3
exclude observations where origin is 50% richer and column 4 exclude those where origin in come is larger than
destination.

Table A3. Regularisation policy as drivers for migration.

Heterogeneity by origin incomes
(1) (2)
Migrant inflow rate Migrant inflow rate
Regularisation to origins Q1 income -0.147
(0.159)
Regularisation to origins Q2 income -0.0474
(0.157)
Regularisation to origins Q3 income -0.00734
(0.153)
Regularisation to origins Q4 income 0.0453
(0.152)
Regularisation to origins below median income -0.0978
(0.156)
Regularisation to origins above median income 0.00211
(0.153)
Pseudo Rsquared 0.911 0.911
Observations 60249 60249
Notes: Standard errors are clustered at the country-pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. Controls: GDPpc ratio origin/destination & logged network. Dyadic, origin-year and destination-year
fixed effects are used.

27
 Table A4. Regularisation policy as drivers for migration
Heterogeneity by destination income
(1) (2)
col_1 col_2
Regularisation destinations Q1 income 0.345**
(0.172)
Regularisation destinations Q2 income -0.409***
(0.153)
Regularisation destinations Q3 income -0.271
(0.178)
Regularisation destinations Q4 income 0
(.)
Regularisation destinations below median income 0.00701
(0.162)
Regularisation destinations above median income -0.270
(0.178)
Pseudo Rsquared 0.911 0.911
Observations 60249 60249
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. Controls: GDPpc ratio origin/destination & logged network. Dyadic, origin-year and destination-year
fixed effects are used.

28
 Table A5. Regularisation policy as drivers for migration.
Heterogeneity by origin-destination income groups
(1)
col_1
LagRegulQ1toQ1 0.677***
(0.250)
LagRegulQ1toQ2 -0.622***
(0.182)
LagRegulQ1toQ3 -0.380**
(0.192)
LagRegulQ1toQ4 -0.384***
(0.0830)
LagRegulQ2toQ1 -0.0199
(0.241)
LagRegulQ2toQ2 -0.419**
(0.186)
LagRegulQ2toQ3 -0.297*
(0.179)
LagRegulQ2toQ4 -0.116
(0.0910)
LagRegulQ3toQ1 0.345**
(0.170)
LagRegulQ3toQ2 -0.408***
(0.152)
LagRegulQ3toQ3 -0.261
(0.178)
LagRegulQ3toQ4 -0.0137
(0.0693)
LagRegulQ4toQ1 0.464**
(0.217)
LagRegulQ4toQ2 -0.670***
(0.199)
LagRegulQ4toQ3 -0.106
(0.183)
LagRegulQ4toQ4 0
(.)
Pseudo Rsquared 0.911
Observations 60249
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. Controls: GDPpc ratio origin/destination & logged network. Estimation method: PPML with three
sets of fixed effects: origin-time, destination-time, origin-destination.

29
 Table A6. Regularisation policy as drivers for migration.
Heterogeneity by regions of origin.
(1)
Migrant inflow rate
Regularisation policy(t-1) * east 0.0236
(0.157)
Regularisation policy(t-1) * mena -0.108
(0.155)
Regularisation policy(t-1) * latam 0.0190
(0.153)
Regularisation policy(t-1) * ssahara -0.130
(0.148)
Regularisation policy(t-1) * easiap -0.0519
(0.150)
Regularisation policy(t-1) * sasia -0.0659
(0.158)
Regularisation policy(t-1) * west 0.0601
(0.146)
Pseudo Rsquared 0.911
Observations 60249
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. RegulTot is the linear combination of the coefficients estimates of regularisation policy dummy, its 6
lags interacted with each regional dummy. Controls: GDPpc ratio origin/destination & logged Network.
Estimation method: PPML with three sets of fixed effects: origin -time, destination-time, origin-destination.

30
Table A7. Regularisation policy as drivers for migration. PPML-HDFE. Heterogeneity by
regions of origin and incomes at destination.
(1)
Migrant inflow rate
Regularisation policy (t-1) east to Poor 0.277
(0.193)
Regularisation policy (t-1) mena to Poor 0.0502
(0.156)
Regularisation policy (t-1) latam to Poor 0.148
(0.144)
Regularisation policy (t-1) ssahara to Poor -0.00764
(0.157)
Regularisation policy (t-1) easiap to Poor -0.183
(0.148)
Regularisation policy (t-1) sasia to Poor 0.116
(0.174)
Regularisation policy (t-1) west to Poor 0.0112
(0.146)
Regularisation policy (t-1) east to Rich -0.283
(0.176)
Regularisation policy (t-1) mena to Rich -0.380*
(0.198)
Regularisation policy (t-1) latam to Rich -0.327*
(0.196)
Regularisation policy (t-1) ssahara to Rich -0.378**
(0.191)
Regularisation policy (t-1) easiap to Rich -0.169
(0.190)
Regularisation policy (t-1) sasia to Rich -0.351*
(0.198)
Regularisation policy (t-1) west to Rich -0.143
(0.186)
Pseudo Rsquared 0.911
Observations 60249
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. All coefficients come from one estimation where the policy was interacted with each possible
combination of region of origin with income of destination. Controls: GDPpc ratio origin/destination & logged
Network. Estimation method: PPML with three sets of fixed effects: origin -time, destination-time, origin-
destination.

31
 Table A8. Regularisation policy as drivers for migration.
PPML with different specifications
(1) (2) (3) (4)
Traditiona 2-way FE HDFE HDFE
l Gravity (6 lags) (8 lags)
Marginal Effect
Regularisation Policy 1.765** 0.250** 0.723 1.152
(0.695) (0.101) (0.523) (0.754)
L.GDP pc of origin relative to -0.0331 -1.640*** 0.305 0.220
destination
(0.0442) (0.313) (0.223) (0.242)
L.Log of Network relative to origin 0.806*** 0.297*** 0.403** 0.411***
pop *

(0.0130) (0.0402) (0.0416 (0.0433)

)
Contiguity -0.0922
(0.0881)
Common official of primary language 0.235***
(0.0864)
Language spoken by both countries 0.0262
(0.0812)
Colonial relationship 0.103
(0.115)
Common colonizer post 1945 0.670***
(0.183)
Colonial relationship post 1945 -0.344***
(0.119)
Were or are the same country 0.0749
(0.0951)
Log Distance -0.00954
(0.0366)
Pseudo Rsquared 0.886 0.879 0.914 0.916
Observations 56367 56849 56440 53590
Dyadic FE No Yes Yes Yes
Year FE No Yes No No
Destination x Year FE Yes No Yes Yes
Origin x Year FE Yes No Yes Yes
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. Regularisation Policy is the linear combination of the coefficients estimates of regularisation policy
dummy and 6 lags (8 for column 4). Controls: GDPpc ratio origin/destination & Network of origin country at
destination. Estimation method: PPML with three sets of fixed effects: origin -time, destination-time, origin-
destination.

32
 Table A9. Regularisation policy as drivers for migration.
PPML-HDFE with different income restrictions for origin country
(1) (2) (3) (4)
Full sample Inc_o_d <2 Inc_o_d <1.5 Inc_o_d <1
Regul marginal effect 0.723 0.805 0.632 0.530
(0.523) (0.585) (0.608) (0.651)

Regularisation policy (t) 0.108 0.158 0.139 0.142

(0.155) (0.161) (0.176) (0.186)
Regularisation policy (t-1) 0.0336 0.0697 0.0290 0.0207
(0.157) (0.154) (0.154) (0.160)
Regularisation policy (t-2) -0.265** -0.232** -0.201* -0.205*
(0.113) (0.112) (0.120) (0.123)
Regularisation policy(t-3) -0.0322 -0.0472 -0.0792 -0.103
(0.147) (0.149) (0.147) (0.149)
Regularisation policy(t-4) 0.112 0.107 0.0823 0.0814
(0.102) (0.112) (0.118) (0.120)
Regularisation policy(t-5) 0.480*** 0.465*** 0.412*** 0.380***
(0.124) (0.128) (0.129) (0.140)
Regularisation policy(t-6) 0.286*** 0.284** 0.250** 0.213*
(0.110) (0.116) (0.120) (0.129)
GDPpc orig-dest (t-1) 0.305 0.787** 0.421 -0.479
(0.223) (0.371) (0.561) (0.784)
Log of Network (t-1) 0.403*** 0.407*** 0.410*** 0.414***
(0.0416) (0.0390) (0.0392) (0.0378)
Pseudo R-squared 0.914 0.916 0.918 0.921
Observations 56440 53628 52168 48920
Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, *** p <
0.01. Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination population.
RegulTot is the linear combination of the coefficients estimates of regularisation policy dummy and its lags.
Controls: GDPpc ratio origin/destination & network. Column 2 restrict the sample to those observations where
per capita income of origin country is not more than double destination country. Column 3 exclude observations
where origin is 50% richer and column 4 exclude those where origin income is larger than destination.

33
 Table A10. Regularisation policy as long-term driver for migration.
Heterogeneity by origin incomes quartiles
(1) (2) (3) (4)
IncQtl1 IncQtl2 IncQtl3 IncQtl4
Marginal effect
RegulTot 0.542 0.663 0.819 0.679
(0.521) (0.527) (0.519) (0.504)
Pseudo R-squared 0.914
Observations 56440
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. RegulTot is the linear combination of the coefficients estimates of regularisation policy dummy and
its lags interacted with respective income group. Controls: GDPpc ratio origin/destination & logged network.
Each column shown the marginal effect of each origin's income quartile. Estimation method: PPML with three
sets of fixed effects: origin-time, destination-time, origin-destination.

Table A11. Regularisation policy as drivers for migration.

Heterogeneity by regions of origin.
(1) (2) (3) (4) (5) (6) (7)
east mena latam ssahara easiap sasia west
Marginal effect
RegulTot 0.969* 0.624 0.686 0.641 0.621 0.773 0.776
(0.547) (0.552) (0.532) (0.535) (0.528) (0.538) (0.520)
Pseudo R-squared 0.914
Observations 56440
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. RegulTot is the linear combination of the coefficients estimates of regularisation policy dummy, its 6
lags interacted with each regional dummy. Controls: GDPpc ratio origin/destination & logged Network.
Estimation method: PPML with three sets of fixed effects: origin -time, destination-time, origin-destination.

Table A12. Regularisation policy as drivers for migration.

Heterogeneity by destination income.
(1) (2)
Poorest destinations Richest destinations
Marginal Effect
RegulTot 1.272** -0.909
(0.583) (0.914)
Pseudo R-squared 0.914
Observations 56440
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. RegulTot is the linear combination of the coefficients estimates of regularisation policy dummy and
its lags interacted with respective income group. Controls: GDPpc ratio origin/destination & logged network.
Each column shown the marginal effect of each destination's income group. Estima tion method: PPML with
three sets of fixed effects: origin-time, destination-time, origin-destination.

34
 Table A13. Regularisation policy as long-term driver for migration.
Heterogeneity by origin-destination income groups
(8)
Migration inflow rate

RegulTotPoortoQ1 1.029
(1.137)
RegulTotPoortoQ2 1.456
(0.997)
RegulTotPoortoQ3 -0.737
(0.921)
RegulTotPoortoQ4 -1.091***
(0.311)
RegulTotRichtoQ1 1.283
(0.815)
RegulTotRichtoQ2 1.714*
(0.966)
RegulTotRichtoQ3 -0.675
(0.909)
RegulTotRichtoQ4 0
(.)
Pseudo Rsquared 0.914
Observations 56440
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. RegulTot is the linear combination of the coefficients estimates of regularisation policy dummy and
its lags interacted with respective income group. Controls: GDPpc ratio origin/destination & logged network.
Each coefficient is the marginal effect of each origin-destination income group and were estimated in one
equation. Estimation method: PPML with three sets of fixed effects: origin -time, destination-time, origin-
destination. Poor and Rich denote being below or above median income for origin countries and Qs stand for
income quartile for destination countries

Table A14. Regularisation policy as long-term driver for migration.

Heterogeneity by regions of origin and incomes at destination
(1) (2) (3) (4) (5) (6) (7)
east mena latam ssahara easiap sasia west
Marginal effect
Regul Region to Rich -0.760 -0.851 -0.654 -0.884 -0.745 -0.778 -0.622
(0.859) (0.906) (0.908) (0.903) (0.904) (0.902) (0.883)
Regul Region to Poor 2.940*** 1.368** 1.459*** 1.514** 1.170* 1.756*** 1.166*
(0.706) (0.631) (0.561) (0.661) (0.626) (0.629) (0.615)
Pseudo Rsquared 0.915
Observations 56440
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. RegulTot is the linear combination of the coefficients estimates of regularisation policy dummy, its 6
lags interacted with each regional dummy. Controls: GDPpc ratio origin/destination & logged Network.
Estimation method: PPML with three sets of fixed effects: origin -time, destination-time, origin-destination

35
 Table A15. Regularisation policy (interacted with distance) as driver for migration.
PPML with different specifications
(1) (2) (3) (4)
Traditional 2-way FE HDFE HDFE
Gravity
Regularisation (t-1) * Dist 0.0108 0.140*** -0.0414 -0.0395
(0.0488) (0.0458) (0.0298) (0.0288)
Regularisation (t-2) * Dist -0.0207
(0.0205)
Regularisation (t-3) * Dist -0.0280
(0.0268)
GDPpc orig-dest (t-1) -0.0529 -1.450*** 0.195 0.194
(0.0443) (0.287) (0.211) (0.214)
Log of Network (t-1) 0.800*** 0.323*** 0.393*** 0.387***
(0.0132) (0.0304) (0.0402) (0.0408)
Contiguity -0.102
(0.0882)
Common official prim lang 0.235***
(0.0880)
Lang spoken both countries 0.0312
(0.0819)
Colonial relationship 0.105
(0.116)
Common colonizer post1945 0.676***
(0.182)
Colonial rel post1945 -0.345***
(0.118)
Were or are same country 0.0979
(0.0943)
Log Distance -0.0171
(0.0367)
Marginal effect
RegulTot -0.0883
(0.0561)
Pseudo Rsquared 0.883 0.874 0.911 0.912
Observations 60128 59859 59482 58915
Dyadic FE No Yes Yes Yes
Year FE No Yes No No
Destination x Year FE Yes No Yes Yes
Origin x Year FE Yes No Yes Yes
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. RegulTot is the linear combination of the coefficients estimates of the 3 lags of regularisation policy
dummy. Controls: GDPpc ratio origin/destination & Network of origin country at destination.

36
 Table A16. Regularisation policy (interacted with distance) as drivers for migration.
Heterogeneity by origin-destination income groups
(1)
Migrant inflow rate
Regularisation (t-1) * Dist PoortoQ1 0.248
(0.167)
Regularisation (t-1) * Dist PoortoQ2 -0.158*
(0.0870)
Regularisation (t-1) * Dist PoortoQ3 -0.0971*
(0.0549)
Regularisation (t-1) * Dist PoortoQ4 -0.298***
(0.0945)
Regularisation (t-1) * Dist RichtoQ1 0.138
(0.134)
Regularisation (t-1) * Dist RichtoQ2 -0.110*
(0.0629)
Regularisation (t-1) * Dist RichtoQ3 0.0399
(0.0352)
Regularisation (t-1) * Dist RichtoQ4 -0.0333
(0.0720)
Pseudo Rsquared 0.911
Observations 59482
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. Controls: GDPpc ratio origin/destination & logged network. Estimation method: PPML with three
sets of fixed effects: origin-time, destination-time, origin-destination

Table A17. Regularisation policy (interacted with distance) as driver for migration.
Heterogeneity by origin-destination income groups
(1)
Migrant inflow rate
Regularisation (t-1) * Dist PoortoPoor -0.0682
(0.0840)
Regularisation (t-1) * Dist PoortoRich -0.135***
(0.0497)
Regularisation (t-1) * Dist RichtoPoor -0.0581
(0.0583)
Regularisation (t-1) * Dist RichtoRich 0.0238
(0.0312)
Pseudo Rsquared 0.911
Observations 59482
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. RegulTot is the linear combination of the coefficients estimates of regularisation policy dummy and
its lags interacted with respective income group. Controls: GDPpc ratio origin/destination & logged network.
Poor and Rich denote being below or above median income in the respective origin or destination group of
countries. Estimation method: PPML with three sets of fixed effects: origin -time, destination-time, origin-
destination.

37
 Table A18. Regularisation policy as drivers for migration.
Instrumental Variables Approach
(1) (2)
1st stage 2nd Stage
Regularisation (t-1) 0.208
(0.172)
GDPpc orig-dest (t-1) -1.148*** 0.241
(0.261) (0.209)
Log of Network (t-1) -0.149*** 0.406***
(0.0277) (0.0447)
Election year (t-2) 0.239***
(0.0192)
Pseudo Rsquared 0.149 0.912
Observations 34825 55488
Dyadic FE Yes Yes
Year FE Yes No
Destination x Year FE No Yes
Origin x Year FE Yes Yes
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. Controls: GDPpc ratio origin/destination & Network of origin country at destination. Estimation
method: PPML with fixed effects. Regularisation policy was instrumented with the lagged dummy for election
year.

Table A19. Regularisation policy as drivers for migration.

Alternative Instruments
(1) (2) (3)
Migration Inflow rate
Regularisation (t-1) -1.789 3.874 0.0811
(8.223) (6.837) (1.845)
GDPpc orig-dest (t-1) 0.184 0.733 0.464
(1.446) (1.356) (1.177)
Log of Network (t-1) 1.898*** 1.925*** 1.826***
(0.362) (0.364) (0.356)
F statistic 9.591 9.472 8.874
p-value of underident K-P 0.0000225 2.88e-22 1.48e-91
Weak ident C-D Wald 16.92 96.85 452.8
Observations 60726 59351 59207
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable: Rate of migrants' inflow from each origin per 100000 hab in destination
population. Controls: GDPpc ratio origin/destination & Network of origin country at destination. A set of fixed
effects was included: country-pair, year and origin-year. Regularisation policy was instrumented by a set of
political variables: dummies for election year in column 1, having control of b oth legislative chambers in column
2, and a having a new executive government in column 3, all of them in lags.

38
 Table A20. Instrumental Variables.
Independence and restriction assumptions
(1) (2)
Assumption Independence. Exclusion.
Dependent Variable Residual Migration Inflow
Estimated regularisation (t-1) 0.234 -0.0233
(0.167) (0.0965)
Election year (t-2) -0.00416 0.00908
(0.0103) (0.00746)
GDPpc orig-dest (t-1) 0.526*** -1.012***
(0.0318) (0.264)
Log of Network (t-1) 0.701*** 0.277***
(0.0318) (0.0371)
Pseudo Rsquared 0.986 0.899
Observations 56003 55507
Notes: Standard errors are clustered at the country pair and are shown in parentheses. * p < 0.1, ** p < 0.05, ***
p < 0.01 . Dependent variable in Column 1 is the Square residual of Table A18's second stage and for column 2
is the Rate of migrants' inflow from each origin per 100000 hab in destination population. Controls: GDPpc ratio
origin/destination & Network of origin country at destination. Estimation method: PPML with three sets of fixed
effects: origin-time, time, origin-destination. Regularisation policy is the estimated one, it was instrumented by
the lagged dummy for election year.

39