## wpiea2024175-print-pdf

## Source details

**Canonical URL:** [wpiea2024175-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024175-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024175-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024175-print-pdf.pdf.json)

---

### 4.1 Initial Model Specification — context, objective, and main results
- Context and motivation
  - Migration increased from 152 million in 1990 to 280 million in 2020 (United Nations data as cited).
  - Emigration flows from LAC to OECD countries quadrupled between 1995-2020.
  - Remittances to LAC increased almost five times between 1995-2020, with a small dent during the global financial crisis (2007-09).
  - LAC is the only major region where remittances have not declined in the past decade; annual growth rates exceeded the world’s average between 2020-22.
  - Five LAC countries received about or over 20 percent of GDP in remittances in 2023 (World Bank, KNOMAD, 2023).
- Research objective
  - Estimate the net joint effect of emigration and remittances on: real GDP growth and labor force participation (LFP), including cohort aged 15-24.
  - Address endogeneity between emigration and remittances and between these variables and outcomes.
- Econometric approach — main specification
  - Estimation period: 1999-2019 (primary results).
  - Method: two-stage least squares (2SLS).
  - Endogenous regressors: emigration rates and remittances per capita.
  - First-stage instrument: shift-share (Bartik-style) IV using destination-country time-varying pull-factors (income, unemployment rate, unit labor costs in manufacturing) weighted by historical emigrant shares.
  - Second stage: regress real GDP growth or change in LFP on predicted emigration rates and predicted remittances per capita.
- Instrument performance
  - Shift-share instruments reported as highly relevant and statistically significant in the first stage.
- Main quantitative findings (1999-2019)
  - Emigration outflows generate an average loss of about 0.28 percentage points of GDP growth per year.
  - Resulting increase in remittances contributes 0.19 percentage points higher GDP growth per year.
  - Net joint effect: 0.09 percentage points lower GDP growth per year (0.28 negative effect from emigration partially offset by 0.19 positive effect from remittances).
- Heterogeneity
  - Negative net impacts on growth and LFP most significant in the Caribbean compared to other LAC subregions (LA5, CAPDR, other LAC).
  - Negative impact on change in labor force participation is small overall, but for ages 15-24 the effect is twice as large as for the overall labor force.
- Robustness and extensions
  - Results robust across various specifications and data alternatives.
  - Framework expanded to net joint effects across LAC subregions and LFP impacts for the 15-24 cohort.

### Data and descriptive evidence (Section III)
- Sample and scope
  - Main specification: 31 LAC countries over 1999-2019; additional pre-sample data from 1989/1990 used for emigrant stocks and initial conditions; some GDP data 1989-2020.
- Key variables and sources
  - GDP: IMF World Economic Outlook database, 1989-2020.
  - LFP rates: World Bank WDI (overall 15+ and youth 15-24 ILO estimates).
  - Emigration: OECD International Migration Database (primary, bilateral inflows/outflows by nationality, 1999-2019) and UNPD International Migrant Stock (1990-2020) for robustness; WPP net migration also used for robustness.
  - Remittances: World Bank WDI (IMF BOP data) as main source; KNOMAD remittances used as robustness alternative.
  - Controls: IMF WEO employment/unemployment/ULC, World Bank GNI/FDI/ODA/trade openness, EM-DAT disaster damages as percent of GDP.
- Descriptive correlations (1999-2019)
  - CORR: -0.12 (Real GDP Growth vs Emigration to OECD Countries (Percent of Origin Population))
  - CORR: -0.05 (Real GDP Growth vs Remmitances per capita (100$))
  - CORR: 0.01 (Labor Force Participation Growth vs Emigration to OECD Countries (Percent of Origin Population))
  - CORR: -0.01 (Labor Force Participation Growth vs Remmittances per capita (100$))
- Measurement choices
  - Main explanatory variable: remittances per capita (avoids dividing by GDP when outcome is GDP growth).
  - Trade openness = exports + imports of goods and services.

### Empirical strategy (Section IV) — baseline and alternatives
- Baseline identification
  - Panel estimation in changes (dependent variables in changes; independent variables in changes because instruments predict changes).
  - Baseline equation includes emigration rate and remittances per capita, controls (FDI % of GDP, ODA % of GNI, natural disaster damages % of GDP), time and origin-country fixed effects.
- Alternative specification
  - Remove origin fixed effects to preserve variation; use geographic fixed effects (Central America, South America, Caribbean) plus time-invariant country controls (log real GDP, log population, log openness in 1989).
- Endogeneity challenges addressed
  - Emigration and remittances jointly included to separate labor-loss channel (β1) from income channel (β2).
  - Shift-share IVs used to isolate exogenous variation in emigration/remittances.

### 4.2 Instrumental Variables — construction, validity, and diagnostics
- Instrument construction
  - Shares: historical emigrant stocks (emigrant stock in destination d from origin in 1990; UNPD; constant across time).
  - Shifts: destination-country changes in GNIPC, LUR, and ULCM.
  - IVs = weighted sums (shift × initial emigrant-stock shares) across destination countries.
- Exogeneity rationale
  - Destination labor-market pull-factors affect migration incentives but are argued to be exogenous to origin-country growth and LFP.
  - Using weighted sums across many destinations reduces direct impact on origin-country outcomes.
- Tests and diagnostics
  - Permutation tests and appendices (Figure A.1–A.4) emphasize importance of historical networks and identify which pull-factors identify emigration vs remittances.
  - Pre-trend tests using 1995-1998 show IVs not significantly correlated with origin-country growth/LFP before sample period.
  - Over-identifying Sargan-Hansen tests reported in main results support IV validity.
- Selected pre-trend statistics (Table 1, 1995-1998)
  - Pull IV: Income — Real GDP growth: 15.88 (14.533)
  - Pull IV: Unemployment rate — LFP Change (15-24): 157.82** (64.432)
  - Pull IV: Manufacturing ULC — Real GDP growth: 307.51 (216.826)
  - Observations reported: 123, 92, 107, 92, 107, 92 (across columns)
  - F-stat reported: 1.031, 3.247, 2.160, 1.384, 1.920, 1.366 (across columns)
  - Test p-value reported: 0.383, 0.026, 0.100, 0.254, 0.134, 0.259 (across columns)
  - Significance notation: *** p<.01, ** p<.05, * p<.10
- First-stage patterns (Table 2 summary)
  - Income strongly and positively correlated with both emigration rate and remittances per capita.
  - Manufacturing ULC more relevant for remittances per capita.
  - Destination unemployment negatively correlated with remittances per capita.
  - First-stage F-stats reported (e.g., 40.352, 48.825 for emigration; 10.403, 53.682 for remittances).

### 5.2 Economic Growth — 2SLS estimates, net joint effects, and heterogeneity
- Estimation setup
  - Dependent variable: percentage change in annual real GDP.
  - Key regressors: OECD emigration rate (emigrants per year to all OECD countries as % of population) and remittances per capita (personal remittances in thousands of current USD per population).
  - Instruments: Income, Unemployment rate, Manufacturing ULC (shift-share IVs).
  - Controls: year fixed effects; FDI % GDP, ODA % GNI, natural disaster damages % GDP; country or geographic fixed effects.
- First-stage selected coefficients (country FE / geo FE; robust SEs)
  - Income → OECD emigration rate: 3.60*** (0.332) / 2.43*** (0.243)
  - Income → Remittances per capita: 1.25*** (0.447) / 1.82*** (0.151)
  - Unemployment → Remittances per capita: -0.58*** (0.168) / -0.92 (0.588)
  - Manufacturing ULC → Remittances per capita: 6.68*** (1.657) / 10.53** (4.988)
  - Observations: 612 (Country FE), 461 (Geo FE)
  - First-stage F-statistics reported: 40.352, 48.825 (emigration) and 10.403, 53.682 (remittances)
- 2SLS joint estimates (Table 3; country FE and geo FE)
  - Country FE — 2SLS:
    - OECD emigration rate: -10.28** (4.326)
    - Remittances per capita: 19.93** (9.344)
    - Underidentification test: 12.849; Endogeneity test: 9.022; Sargan-Hansen test: 0.123
  - Geo FE — 2SLS:
    - OECD emigration rate: -15.32* (8.679)
    - Remittances per capita: 10.68 (10.967)
    - Underidentification test: 7.589; Endogeneity test: 29.422; Sargan-Hansen test: 0.021
- Net joint effect under a hypothetical 10 percent destination income shock (first-stage implied responses)
  - First-stage implied average responses:
    - Increase in emigration rate toward OECD by 0.028 percentage point on average (calculation as 0.1 * 3.597 * 0.077 in text).
    - Increase in average remittances per capita by $10.
  - Net joint impacts on real GDP growth (using 2SLS coefficients and predicted changes):
    - Aggregate LAC net joint impact: -0.09 percent.
    - Caribbean net joint impact: -0.19 percent.
    - CAPDR net joint impact: -0.05 percent.
    - LA5 and Other LAC: net joint impact ≈ almost zero.
  - Example calculation note: -0.093 = 0.028 * (-10.282) + 0.010 * 19.933.
- LFP contemporaneous impacts (Table 4; OLS vs 2SLS)
  - LFP Change (15+)
    - OLS: OECD emigration rate 0.63 (0.404); Remittances per capita 0.99* (0.600)
    - 2SLS: OECD emigration rate -7.63*** (1.988); Remittances per capita 3.94 (3.148)
  - LFP Change (15-24)
    - OLS: OECD emigration rate 1.11* (0.595); Remittances per capita 1.579* (0.816)
    - 2SLS: OECD emigration rate -14.64*** (3.196); Remittances per capita 3.27 (5.061)
  - Magnitude interpretation:
    - A 0.1 percent increase in the emigration rate to OECD countries → drop of 0.76 percent of overall LFP (0.1 * 7.63) and drop of 1.46 percent of LFP for ages 15-24 (0.1 * 14.64).
  - Remittances per capita insignificant on LFP in 2SLS.
- Regional heterogeneity (Table 5)
  - Subregional interactions show largest negative net effects in Caribbean:
    - Aggregate net effect estimated at 0.1 percent decrease of economic growth across LAC.
    - Caribbean region: estimated net -0.4 percent decrease of economic growth (text summary).
  - Diagnostics robust to excluding Venezuela.
- Gender heterogeneity (Table 6; 2SLS selected coefficients)
  - LFP Change (15+)
    - Female 2SLS: OECD emigration rate -9.272*** (2.445); Remittances per capita 4.310 (3.871)
    - Male 2SLS: OECD emigration rate -6.100*** (1.820); Remittances per capita 3.573 (2.882)
  - LFP Change (15-24)
    - Female 2SLS: OECD emigration rate -14.370*** (3.272)
    - Male 2SLS: OECD emigration rate -15.078*** (3.522)
  - Interpretation: Emigration associated with declines in LFP for both genders; negative effect especially strong for overall female labor force and large for youth for both genders.

### 6.2 Dynamic Effects — lags, persistence, and longer horizons
- Longer-horizon baseline (1995–2019) — Table 7 (OLS and 2SLS; remittances per capita in US$1000)
  - OECD emigration rate on Real GDP growth:
    - OLS: 0.12 (0.76)
    - 2SLS: -8.71*** (3.21)
  - Remittances per capita on Real GDP growth:
    - OLS: 1.47 (1.215)
    - 2SLS: 19.29** (8.305)
  - OECD emigration rate on LFP Change (15+):
    - OLS: 0.54 (0.554)
    - 2SLS: -3.17*** (1.045)
  - OECD emigration rate on LFP Change (15-24):
    - OLS: 1.16 (1.149)
    - 2SLS: -5.01*** (1.512)
  - Observations: Real GDP growth 724; LFP series 630.
  - Diagnostics: Underidentification test 23.937; Endogeneity test 11.405; Sargan-Hansen 0.416 (where shown).
  - Interpretation: Emigration negative for growth and LFP; remittances partially mitigate growth losses.
- Current vs past effects (Table 8)
  - One-year lagged endogenous variables instrumented by lagged IVs.
  - 2SLS coefficients (selected)
    - Current (baseline): OECD Emigration rate -10.28** (4.326); Remittances per capita 19.93** (9.344)
    - Lagged: Lag OECD Emigration rate -15.96*** (5.497); Lag Remittances per capita 25.50* (13.523)
    - Current and Past included: current coefficients become smaller/insignificant; lagged emigration remains negative and significant (Lag OECD Emigration rate -14.02*** (4.724))
  - Interpretation: Past emigration and past remittances often show larger effects; past emigration the most persistent negative effect on growth.
- Five-year effects (Table 9; 1995–2020, five-year changes)
  - Emigration rate on Real GDP growth:
    - OLS: -3.18*** (0.935)
    - 2SLS: -3.37* (1.723)
  - Remittances per capita on Real GDP growth:
    - OLS: 10.83 (6.872)
    - 2SLS: 26.91 (26.322)
  - Observations: 120 across specifications.
  - Interpretation: Negative emigration impact on growth persists over five-year aggregates; remittances point estimates larger but less statistically significant at five-year horizon.
- Dynamic summary
  - Short run: emigration negative contemporaneous effects on growth and LFP; remittances mitigate growth losses.
  - Medium run (one-year lag): larger and more significant impacts from past emigration and remittances; lagged emigration highly persistent negative.
  - Long run (five-year): negative effect of emigration on growth persists; remittances larger in point estimates but statistical significance weaker.
- Robustness checks (section 6.3 and Annex)
  - Alternative IV (leave-one-out inflow) yields similar conclusions.
  - Base years 1995/1994 produce consistent findings.
  - Alternative measures (net emigration, remittances in constant 2015 USD, KNOMAD remittances, real GDP per capita growth) show qualitative robustness.
  - Lag population denominator specifications yield similar coefficients.
  - Levels specifications: emigration negatively impacts contemporaneous GDP level; remittances positively impact contemporaneous GDP level in OLS, not significant in 2SLS.

### Annex II — additional tables, permutation exercises, and robustness aggregates
- Main regression summaries (Table A.4; 1995–2020 baseline)
  - OECD Emigration rate (OLS / 2SLS) on Real GDP growth: -0.77 / -6.07
  - Remittances per capita (OLS / 2SLS): 1.88 / 8.00
  - Observations: 749 (OLS); First-stage Obs. 749 (2SLS)
- Alternative IVs and robustness (Table A.5–A.13 highlights)
  - Alternative IV (1999–2019) 2SLS for Real GDP growth: OECD Emigration rate -20.29**; Remittances per capita 15.19.
  - Using 1995 base year (Table A.6) 2SLS country FE: OECD Emigration rate -15.570**; Remittances per capita 15.874*.
  - Net emigration rate (Table A.9) 2SLS: OECD Emigration rate -9.07** on Real GDP growth.
  - Remittances in 2015 constant USD (Table A.10) 2SLS: Remittances per capita 29.84** on Real GDP growth.
  - KNOMAD remittances (Table A.11) similar qualitative patterns.
  - Real GDP per capita growth (Table A.12) 2SLS country FE: OECD Emigration rate -9.43**; Remittances per capita 16.20*.
- Permutation and shift-share robustness (Figures A.1–A.4)
  - Permutation exercises (500 replications) show average emigration and remittances effects robustly negative and positive respectively in many exercises:
    - Permutation of shifts (per capita income) average emigration effect -10.18 (320/500 significant at 10%); average remittances effect 19.34 (451/500 significant at 10%).
    - Permutation of shifts (ULCM) average emigration effect -9.69 (493/500 significant at 10%); average remittances effect 13.93 (106/500 significant at 10%).
- Additional methodological notes
  - Remittances per capita typically measured in US$1000 (some tables note 2015 constant US$1000).
  - Standard errors: robust SEs used in 2SLS estimations.
  - Significance: *** p<.01, ** p<.05, * p<.10.
  - Fixed effects: time (t) and country (o) or geographic (g) used across specifications.
  - Controls: FDI % GDP, ODA % GNI, natural disaster damages % GDP; time-invariant controls often include log real GDP, log population, trade openness in 1989 (or 1994/1995 where applicable).

*Italic: Source — IMF Working Paper: "The Joint Effect of Emigration and Remittances on Economic Growth and Labor Force Participation in Latin America and the Caribbean" (sections 2–6 and Annex II as provided).*

### 4.1 Initial Model Specification ........................................................................................

### 4.1 Initial Model Specification

### Context and motivation
- Migration increased worldwide from 152 million in 1990 to 280 million in 2020 (United Nations data as cited).
- Emigration flows from LAC to OECD countries quadrupled between 1995-2020.
- Remittances to LAC increased almost five times between 1995-2020, with a small dent during the global financial crisis (2007-09).
- LAC is the only major region where remittances have not declined in the past decade, and annual growth rates exceeded the world’s average between 2020-22.
- Five LAC countries received about or over 20 percent of GDP in remittances in 2023 (World Bank, KNOMAD, 2023).

### Research objective
- Estimate the net joint effect of emigration and remittances on:
  - Economic growth (real GDP growth).
  - Labor force participation (LFP), including the youngest cohort (15-24).
- Address endogeneity between emigration and remittances and between these variables and outcomes (reverse causality, omitted variable bias, measurement error).

### Econometric approach — main specification
- Estimation period: 1999-2019 (primary results).
- Estimation method: two-stage least squares (2SLS).
- Endogenous regressors: emigration rates and remittances per capita.
- First-stage instrument: a shift-share (Bartik-style) instrumental variable in the spirit of Anelli and others (2023).
  - Shifts: destination-country time-varying pull-factor variables:
    - income,
    - unemployment rate,
    - unit labor costs in manufacturing.
  - Shares: historical shares of emigration from each origin country to destination countries.
  - Rationale: the three pull-factors are likely exogenous for origin-country GDP growth and LFP and are weighted by historical emigration shares to strengthen exogeneity.
- Outcomes regressed (second stage): real GDP growth or change in labor force participation on the predicted values of emigration rates and remittances per capita.

### Instrument performance
- The shift-share instruments are reported as highly relevant and statistically significant in the first stage (see Table 2 and First Stage discussion).

### Main quantitative findings (1999-2019)
- Emigration outflows generate an average loss of about 0.28 percentage points of GDP growth per year.
- The resulting increase in remittances contributes to a 0.19 percentage points higher GDP growth per year.
- Net joint effect: 0.09 percentage points lower GDP growth per year (0.28 negative effect from emigration partially offset by 0.19 positive effect from remittances).

### Heterogeneity and subgroup findings
- The negative net impacts of emigration and remittances on growth and LFP are most significant in the Caribbean countries compared to other LAC subregions (LA5, CAPDR, and other LAC). (Country groupings are in Table A.1.)
- The negative impact on the change in labor force participation is small for the overall population, but for the young population aged 15-24 the effect is twice as large as that on the overall labor force.

### Robustness and extensions (summary)
- Results are robust across various specifications and data alternatives (sections on IVs, model estimation, empirical results, and robustness explore heterogeneous effects, dynamic effects, alternative IVs, measurements and specifications).
- The framework is expanded to:
  - Explore net joint effects across LAC subregions.
  - Distinguish net joint impacts on LFP for the 15-24 cohort.

*IMF Working Paper — The Joint Effect of Emigration and Remittances on Economic Growth and Labor Force Participation in Latin America and the Caribbean*

### section 3 describes the data used in the study. Section 4 shows the empirical strategy and estimation. The

### wpiea2024175-print-pdf — sections 2–4 (Analytical background, Data, Empirical strategy)

### Analytical Background
- Purpose: Summarizes channels through which emigration and remittances affect real GDP growth and labor force participation.
- Main channels and mechanisms discussed:
  - Labor supply (permanent emigration reduces working-age population → reduces labor force, employment, human capital, entrepreneurship, and hence GDP growth from the supply side), but net impact is context-dependent (labor force participation, chronic unemployment/underemployment, skill mix, labor demand).
  - Wages and productivity effects:
    - Emigration of high-skilled labor can cause “brain drain” and reduced productivity when skills are not substitutable → expected negative impact on growth.
    - Positive network externalities (trade, investment, technology transfer, capital market development) from emigrant networks can offset negatives contingent on political stability and good governance.
  - Remittances:
    - Income effect: can decrease labor force participation (substitution of labor income with remittances; raise reservation wages) (Chami and others, 2005, 2008).
    - Alternative effect: remittances may increase labor force participation and hours worked if substitution effect of foregone labor is dominated by shortage of household labor for non-domestic activities (Acosta, 2020).
    - Sectoral and formal/informal differences: emigration and remittances may have different impacts across formal vs. informal employment and rural vs. urban areas (Amuedo-Dorantes and Pozo, 2006; Chami and others, 2018).
    - Aggregate demand and investment channels: remittances can boost household income, be countercyclical, alleviate liquidity constraints and disaster/conflict shocks (Beaton and others, 2017); if used for investment or human capital, remittances can promote long-term growth (Bird and Choi, 2020; Aggarwal and others, 2011; Fromentin, 2018).
    - Negative long-run effects: moral hazard reducing political pressure for reforms (Abdih and others, 2012); support to private consumption may raise non-tradable prices and appreciate the real exchange rate, weakening export competitiveness (Chami and others, 2008; Acosta and others, 2009).
- Empirical literature: macroeconomic impact of remittances on growth is inconclusive (Clemens and McKenzie, 2018; Cazachevici and others, 2020); studies focusing on LAC show mixed results (Acosta and others, 2006; Feeny and others, 2014; Lim and Simmons, 2015).
- Household-level joint decision models (e.g., Lim, Morshed and Turnovsky, 2023) find unclear net effects: positive impact on steady-state output but transition losses from reduced consumption often dominate.

### Data (Section III)
- Sample and scope:
  - Main specification: 31 LAC countries over 1999-2019.
  - Additional pre-sample data used from 1989/1990 for emigration stocks and economic conditions.
  - For some variables GDP data covers 1989-2020.
- Key variables and data sources:
  - GDP and labor force participation:
    - GDP: IMF World Economic Outlook database, 1989-2020.
    - Labor force participation rates: World Bank World Development Indicators (WDI) — two sets: overall (aged over 15) and young population (aged between 15 and 24) (ILO estimates).
  - Emigration:
    - Main sources: OECD International Migration Database and United Nations Population Division (UNPD) International Migrant Stock database.
    - Main specification relies on OECD data: bilateral inflows/outflows to all OECD countries by nationality; sample period 1999-2019 (most countries have data from 1999); captures regular/legal emigration (derived from population registers or residence permit data).
    - UNPD migration stocks: 1990-2020, based on official foreign-born/foreign population statistics; considered highest quality for global coverage.
    - For robustness also use annual migration flows from UN World Population Prospects (WPP); WPP “net migration” (1950-2022) incorporates flows, refugees, and differences between population growth and natural increase and thus partially captures unofficial emigration.
  - Remittances:
    - Personal remittances received from World Bank WDI (uses IMF Balance of Payments data, aggregated by countries of origin).
    - Robustness alternative: remittance inflows from World Bank KNOMAD (aggregated by migration-remittances corridors).
  - Other explanatory/control variables:
    - Employment, unemployment rate, and unit labor cost in manufacturing: IMF World Economic Outlook database.
    - GNI, FDI, ODA, trade openness (percent of nominal GDP): World Bank WDI.
    - Cost of natural disasters: EM-DAT (total damages summed annually and represented as shares of current GDP).
- Descriptive and graphical evidence:
  - Panel Figure 3 correlations (1999-2019): visual correlations between dependent variables (real GDP growth or change in labor force participation) and emigration/remittances per capita.
    - Left charts suggest on average a weak negative relationship between emigration rate and real GDP growth and almost no correlation with change in labor force participation.
    - Right charts: correlations with remittances per capita weaker, positive in some subregions, around zero for the overall dataset.
  - Reported simple correlation coefficients shown on charts:
    - CORR: -0.12 (Real GDP Growth vs Emigration to OECD Countries (Percent of Origin Population))
    - CORR: -0.05 (Real GDP Growth vs Remmitances per capita (100$))
    - CORR: 0.01 (Labor Force Participation Growth vs Emigration to OECD Countries (Percent of Origin Population))
    - CORR: -0.01 (Labor Force Participation Growth vs Remmittances per capita (100$))
- Notes on variable measurement choices:
  - Use remittances per capita as the main explanatory variable (footnote: remittances per capita used rather than remittances as share of GDP because outcome of interest is GDP growth and using remittances as share of GDP could bias estimates).
  - Trade openness defined as sum of exports and imports of goods and services.
- Additional dataset and appendix references:
  - Table A.1 lists countries, regional/geographical groupings, and number of annual observations for real GDP growth or labor force participation.
  - Tables A.1–A.3 show descriptive statistics and sample used; Table A.1 shows subregional grouping and geographic division according to UN WPP.
  - Historical note: share of international migrant stocks from LAC among the world increased from about 10 percent to 15.3 percent during 1990-2020 (UNPD, 2020). (Footnote present in source.)

### Empirical Strategy (Section IV)
- Identification approach: panel estimation exploiting variation across time and countries to estimate joint net effect of emigration and remittances on real GDP growth and labor force participation, given lack of consistent household surveys across region.
- Baseline model (changes in dependent variable; independent variables not used in levels because instruments predict changes):
  - Equation (1) as presented in the source:
    - 푦푦표표,푡푡 = 훼훼 + 훽훽1 퐸퐸퐸퐸퐸퐸퐸퐸퐸퐸퐸퐸 표표,푡푡 푃푃퐸퐸 푃푃푃푃푃푃퐸퐸퐸퐸퐸퐸퐸퐸   푛푛 표표,푡푡 + 훽훽2 푅푅푅푅퐸퐸퐸퐸퐸퐸퐸퐸퐸퐸푛푛푅푅푅푅 푠푠 표표,푡푡 푃푃퐸퐸푃푃푃푃푃푃  퐸퐸퐸퐸퐸퐸퐸퐸 푛푛 표표,푡푡 + 훾훾훾훾퐸퐸푛푛퐸퐸퐸퐸퐸퐸푃푃    푠푠 표표,푡푡 + 휙휙푡푡 + 휙휙표표 + 휖휖표표푡푡
  - Definitions in-source:
    - 푦푦표표,푡푡: dependent variable (either economic growth, or change in the labor force participation) in country of origin o in year t.
    - 퐸퐸... (emigration term): yearly emigration rate (share of emigration flows over population) in countries of origin in year t.
    - 푅푅... (remittances term): share of received remittances per capita in the country of origin in year t.
  - Control variables: net FDI (percent of GDP), net ODA (percent of GNI), total damage of natural disasters (percent of GDP).
  - Fixed effects: 휙휙푡푡 (time-specific) and 휙휙표표 (origin country-specific) to capture year and country invariant characteristics; country fixed effects absorb variables like unemployment rate and GDP per capita.
  - Main specification uses changes in dependent variables; independent variables measured as changes because instruments predict changes over time.
  - Robustness: results reported with various lags in robustness section.
- Alternative specification removing origin-country fixed effects (equation (2)):
  - Rationale: controlling all origin-country fixed effects would remove much of relevant variation given sample size.
  - Replace origin fixed effects with geographic fixed effects (휙휙E) for three LAC subgroups (Central America, South America, Caribbean) and time-invariant country-specific controls (initial economic conditions).
  - Time-invariant controls used (from 1989, prior to 1990 emigrant stocks): log real GDP, log population, log openness (trade as share of GDP).
  - Equation (2) as presented in the source:
    - 푦푦표표,푡푡 = 훼훼 + 훽훽1 퐸퐸... 표표,푡푡 푃푃퐸퐸 푃푃푃푃푃푃... + 훽훽2 푅푅... 표표,푡푡 푃푃퐸퐸 푃푃푃푃푃푃... + 훾훾훾훾퐸퐸... 푠푠 표표,푡푡 + 휎휎훾훾... 푠푠 표표 + 휙휙푡푡 + 휙휙퐸퐸 + 휖휖표표,푡푡
- Identification challenges emphasized:
  - Endogeneity: emigration and remittances likely correlated with unobserved local economic conditions and with each other → biased estimates of 훽훽1 and 훽훽2.
  - Private household decisions: emigration and remittances are household-level decisions that may share common characteristics affecting expenditure, education, healthcare, etc.
  - Omitted variable bias if emigration and remittances not included simultaneously or if a common factor affects both and the dependent variable.
  - Including both emigration and remittances in regressions allows differentiation between effect through labor loss (훽훽1) and income gain (훽훽2), addressing the challenge that remittances are caused by emigration and emigration can have opposing effects on growth.

*Italic: Source — IMF Working Paper: "The Joint Effect of Emigration and Remittances on Economic Growth and Labor Force Participation in Latin America and the Caribbean" (sections 2–4 as provided).*

### 4.2 Instrumental Variables

### 4.2 Instrumental Variables

### Role and rationale for instruments
- Lagged variables of interest have frequently been used as instruments in the literature but are often criticized.
- A stronger instrument is one that is correlated with emigration or remittances but uncorrelated with the common household characteristic leading to both emigration and remitting and spending patterns, so that the instrument isolates variation uncorrelated with the error term.
- The paper relies on a shift-share (Bartik-style) instrumental variable to address endogeneity by decomposing changes in the share of emigration or remittances per capita into two components: the shift and the share, both assumed exogenous.
- Shift-share instruments here are constructed using historical/initial emigrant stocks as “shares” and destination-country changes in pull-factors as “shifts.”

### Construction of the shift-share (pull-factor) instruments
- Shares: historical/initial emigrant stocks in destination countries from each origin country (emigrant stock in country d from origin E in 1990; UNPD data; constant for each origin-destination pair over time).
- Shifts (changes in destination-country indicators): 
  - Income represented by GNI per capita (GNIPC)
  - Labor unemployment rate (LUR)
  - Unit labor costs in manufacturing (ULCM)
- The three pull-factor instrumental variables are the weighted sums (shift × initial emigrant-stock shares) of destination-country changes in GNIPC, LUR, and ULCM (equations (3), (4), (5) in the source).
- Intuition: a higher initial emigrant stock increases migration through social networks and ensures an informative first-stage; destination labor-market changes provide exogenous variation that affects incentives to migrate and remit while having lower correlation with origin-country GDP growth or labor force outcomes.

### Validity and exogeneity arguments
- Pull-factors chosen because they are indicators of destination labor markets, directly related to incentives to migrate and remit, and more likely to be exogenous for origin countries than broader destination indicators used in prior work (e.g., destination economic growth or interest rates).
- Using weighted sums across all destination countries (rather than a single destination such as the US) restricts the magnitude of any direct impact on origin-country outcomes via interaction with migrant shares, supporting instrument exogeneity.
- No reverse causality is expected from origin-country emigration rates to destination-country income changes.

### Tests of instrument relevance and validity
- Permutation tests: results of permutation tests performed by Anelli and others (2023) are shown in Figure A.1 in the Appendix and emphasize the importance of historical origin–destination networks for identification.
- Identification patterns from appendices:
  - Figure A.2: per capita income and unemployment together identify the effect of emigration, but not remittances.
  - Figure A.4: unemployment and unit labor cost in manufacturing together identify the effect of remittances, but not emigration.
- Over-identifying restrictions: Sargan-Hansen tests on instrument validity are performed in 2SLS estimation (three IVs against two endogenous variables); test statistics are reported in main result tables and support IV validity (see section 5).
- Pre-trend tests:
  - Sample used for estimation: 1999-2019.
  - Pre-sample period used for testing independence: 1995-1998.
  - Table 1 (Pre-trend Correlations, 1995-1998) shows the IVs are not significantly correlated with economic growth and labor force participation in origin countries prior to the sample period, suggesting independence of the IVs from preexisting economic trends.
  - Additional appendices:
    - Table A.4: pre-sample-period economic growth and labor force participation in 1995-1998 are not significantly correlated with future instrumental variables using four-year lags.
    - Table A.5: responses to the three IVs across different horizons (two-year lags to two-year leads) confirm contemporaneous variations of the IVs are most relevant to both emigration and remittances.

### Selected statistics from Table 1 (Pre-trend Correlations, 1995-1998)
- Pull IV: Income — correlation coefficients (with robust standard errors below):
  - Real GDP growth: 15.88 (14.533)
  - LFP Change (15+): 1.05 (16.667)
  - LFP Change (15-24): 9.77 (6.497)
  - Additional columns in table include values 9.17 (7.761), 23.51 (14.886), 29.18* (17.095)
- Pull IV: Unemployment rate — correlation coefficients (with robust standard errors below):
  - Real GDP growth: 45.38 (142.000)
  - LFP Change (15+): 287.09 (178.333)
  - LFP Change (15-24): 157.82** (64.432)
  - Additional columns in table include values 142.71* (82.510), 379.08** (165.773), 368.56* (192.788)
- Pull IV: Manufacturing ULC — correlation coefficients (with robust standard errors below):
  - Real GDP growth: 307.51 (216.826)
  - LFP Change (15+): -244.01 (266.973)
  - LFP Change (15-24): 84.9 (119.414)
  - Additional columns in table include values 3.99 (89.675), 269.65 (304.664), 200.73 (250.993)
- Observations reported: 123, 92, 107, 92, 107, 92 (across columns)
- F-stat reported: 1.031, 3.247, 2.160, 1.384, 1.920, 1.366 (across columns)
- Test p-value reported: 0.383, 0.026, 0.100, 0.254, 0.134, 0.259 (across columns)
- Notes in Table 1:
  - Robust standard errors shown below coefficients.
  - *** p<.01, ** p<.05, * p<.10.
  - Pull IVs are constructed using four-year forward shocks from the sample period in the future.
  - Tests conducted under null that all coefficients of included instruments are zero.
  - Time-invariant controls: log real GDP, log population and trade openness in 1989.
  - Other controls: FDI as a share of GDP, ODA as a share of GNI, and natural disaster damages as a share of GDP.
  - Fixed effects: time (t) fixed effects used; either origin country (o) or geographic (g: Central America, Caribbean, and South America) fixed effects included.

### First-stage relevance and patterns (summary)
- First-stage regressions (Table 2 in source) indicate the three IVs are highly informative of emigration and remittances across specifications (F-statistics reported).
- Key first-stage findings:
  - Destination-country income changes are positively and significantly correlated with both the emigration rate and remittances per capita.
  - Unit labor cost in manufacturing is more relevant for remittances per capita.
  - Destination unemployment rate shows the expected negative correlation with received remittances: higher destination unemployment is associated with less remittances received in the origin country.
- The effect of destination unemployment on emigration is less clear: one column suggests a positive relationship under country fixed effects, while geographic fixed effects produce a much smaller and less significant correlation.
- Country-fixed effects are important for the unemployment IV; geographic fixed effects are less relevant for that IV.

### Link to estimation strategy (overview of next steps)
- After instrument identification and validation, the paper proceeds with two-stage estimation:
  - First stage: regress emigration rate and remittances per capita (both as share of population) on the three pull-factor IVs, controls (FDI, ODA, natural disasters), time and country/geographic fixed effects (equations (6) and (7)).
  - Predicted values from first stage are used as instruments in second stage.
  - Second stage: regress dependent variables (changes in real GDP or labor force participation rate) on predicted emigration and predicted remittances, plus controls and fixed effects (equation (8)) to estimate β̂1 and β̂2.
- Treatment of sample differences:
  - Two-sample two-stage least squares (TS2SLS) used when labor force participation rates (y_d,t) contain a non-negligible share of missing values in the second stage; economic growth second-stage uses same sample as first-stage (no missing values for growth).
  - Predicted emigration rates and remittances per capita from the same first-stage sample are used for both second-stage estimations; second-stage LFP estimation is restricted by LFP data availability and therefore uses a smaller sample.

*IMF Working Papers — "The Joint Effect of Emigration and Remittances on Economic Growth and Labor Force Participation in Latin America and the Caribbean" (section 4.2 Instrumental Variables).*

### 5.2 Economic Growth

### 5.2 Economic Growth

### Estimation setup and instruments
- Dependent variable: percentage change in annual real GDP.
- Key explanatory variables: OECD emigration rate (emigrants per year to all OECD countries as a percentage of current-year population of the country of origin) and remittances per capita (personal remittances measured in thousands of current USD divided by total population).
- Instrumental variables (IVs): pull IVs for Income, Unemployment rate, and Manufacturing unit labor cost; two-stage least squares (2SLS) uses predicted values from these instruments.
- Controls: year fixed-effects in all estimations; additional controls of FDI (share of GDP), ODA (share of GNI), and natural disaster damages (share of GDP). Time-invariant controls (when geographic fixed effects are used) include log real GDP, log population and trade openness in 1989.
- Fixed effects: origin country (o) or geographic (g: Central America, Caribbean, and South America) and time (t).
- Robust standard errors reported in country fixed effects sample.

### First-stage diagnostics (selected coefficients and statistics)
- Pull IV: Income on OECD emigration rate: 3.60*** (0.332) with Country FE; 2.43*** (0.243) with Geo FE.
- Pull IV: Income on Remittances per capita: 1.25*** (0.447) with Country FE; 1.82*** (0.151) with Geo FE.
- Pull IV: Unemployment rate on OECD emigration rate: 1.25*** (0.377) Country FE; 0.06 (1.065) Geo FE.
- Pull IV: Unemployment rate on Remittances per capita: -0.58*** (0.168) Country FE; -0.92 (0.588) Geo FE.
- Pull IV: Manufacturing unit labor cost on Remittances per capita: 6.68*** (1.657) Country FE; 10.53** (4.988) Geo FE.
- Observations: 612 (Country FE) and 461 (Geo FE).
- R-squared: 0.877 and 0.615 (OECD emigration rate first-stage); 0.900 and 0.660 (Remittances per capita first-stage).
- First-stage F-stat: 40.352, 48.825 (emigration) and 10.403, 53.682 (remittances) as reported.

### 2SLS joint estimates: emigration, remittances, and contemporaneous GDP growth (Table 3)
- Country fixed effects (Country FE)
  - OLS:
    - OECD emigration rate: -0.08 (0.965)
    - Remittances per capita: 2.18 (1.524)
  - 2SLS:
    - OECD emigration rate: -10.28** (4.326)
    - Remittances per capita: 19.93** (9.344)
  - Instruments used: Income, unemployment rate, unit labor costs in manufacturing.
  - Underidentification test: 12.849
  - Endogeneity test: 9.022
  - Sargan-Hansen test: 0.123
- Geographic fixed effects (Geo FE)
  - OLS:
    - OECD emigration rate: 0.39 (0.901)
    - Remittances per capita: -3.07** (1.222)
  - 2SLS:
    - OECD emigration rate: -15.32* (8.679)
    - Remittances per capita: 10.68 (10.967)
  - Underidentification test: 7.589
  - Endogeneity test: 29.422
  - Sargan-Hansen test: 0.021
- Key interpretation:
  - 2SLS estimates show a significant contemporaneous negative effect of emigration rate on real GDP growth, and a significant contemporaneous positive effect of remittances per capita on real GDP growth (in country FE 2SLS).
  - OLS estimates do not show significant links between emigration or remittances and economic growth.

### Net joint effect under a pull shock scenario (10 percent increase in destination income)
- Shock: hypothetical 10 percent increase in income growth in all destination countries used to compute joint effect through predicted changes in emigration and remittances.
- First-stage implied average responses across all LAC countries:
  - Increase in emigration rate towards OECD countries by 0.028 percentage point on average (result of 0.1 * 3.597 * 0.077 as explained).
  - Increase in average remittances per capita by $10.
- Net joint impacts on real GDP growth (using 2SLS coefficients and predicted changes):
  - Aggregate LAC: net joint impact on GDP growth = -0.09 percent.
  - Caribbean: net joint impact = -0.19 percent.
  - CAPDR: net joint impact = -0.05 percent.
  - LA5: net joint impact ≈ almost zero.
  - Other LAC: net joint impact ≈ almost zero.
- Example calculation note given in text: -0.093 = 0.028 * (-10.282) + 0.010 * 19.933 (methodology for computing net joint effect).

### Interpretation of economic mechanisms (as stated)
- Emigration rate → negative contemporaneous effect on GDP growth consistent with labor loss in origin countries associated with weaker economic activity.
- Remittances per capita → positive contemporaneous effect on GDP growth consistent with monetary compensation to households that helps raise economic activity.
- Net joint effect not a simple sum of two coefficients because emigration and remittances do not move one-to-one; external destination shocks are used to trace joint impacts.

### Change in labor force participation (LFP) — contemporaneous impacts (Table 4 and narrative)
- Dependent variables: change in overall labor force participation (15+) and change in labor participation for ages 15-24.
- OLS vs 2SLS: OLS shows no clear relationship; IV (2SLS / two-sample 2SLS) shows stronger, statistically significant relationships indicating endogeneity matters.
- Table 4 coefficients (selected)
  - LFP Change (15+)
    - OLS: OECD emigration rate 0.63 (0.404); Remittances per capita 0.99* (0.600)
    - 2SLS: OECD emigration rate -7.63*** (1.988); Remittances per capita 3.94 (3.148)
  - LFP Change (15-24)
    - OLS: OECD emigration rate 1.11* (0.595); Remittances per capita 1.579* (0.816)
    - 2SLS: OECD emigration rate -14.64*** (3.196); Remittances per capita 3.27 (5.061)
- Magnitude interpretation given in text:
  - A 0.1 percent increase in the emigration rate to OECD countries is associated with:
    - a drop of 0.76 percent of the overall labor force participation rate in the origin countries (0.1 * 7.63 = 0.763).
    - a drop of 1.46 percent of labor force participation for the youth cohort aged 15-24 (0.1 * 14.64 = 1.464).
- Remittances per capita: insignificant effect on LFP changes in 2SLS estimations; literature-consistent ambiguity.

### Net joint effect on LFP under 10 percent destination income shock
- Net joint effect of emigration and remittances is negative for LFP changes across all LAC countries.
- The reduction is much more severe for the younger generation (15-24), with especially strong impacts in the Caribbean region (figures referenced in text).

### Heterogeneous regional effects (summary of findings and Table 5)
- Subregional specification interacts emigration and remittances with subregion dummies: LA5, CAPDR, Caribbean, Other LAC.
- Table 5 highlights:
  - LA5 # OECD emigration rate: OLS 15.86*** (4.158); 2SLS -9.96 (21.174)
  - CAPDR # OECD emigration rate: OLS 1.21 (1.986); 2SLS -5.93 (7.618)
  - Caribbean # OECD emigration rate: OLS 0.02 (1.010); 2SLS -11.37* (6.045)
  - Other LAC # OECD emigration rate: OLS 1.61 (1.574); 2SLS -4.05 (7.563)
  - LA5 # Remittances per capita: OLS -14.58*** (5.148); 2SLS 12.84 (19.753)
  - Caribbean # Remittances per capita: OLS -5.73*** (1.561); 2SLS 6.38 (8.352)
- Observations: 461 (main), 445 (excluding Venezuela).
- Diagnostics when interacting by subregion: Underidentification test 13.659 (main) and 12.904 (excluding Venezuela); Endogeneity test 17.499 (main) and 30.022 (excluding Venezuela); Sargan-Hansen test 3.626 (p-value 0.459) and 4.172 (p-value 0.383).
- Aggregate heterogeneity summary:
  - Overall net effect estimated at 0.1 percent decrease of economic growth across LAC countries.
  - Caribbean region: estimated net -0.4 percent decrease of economic growth.
- Robustness: excluding Venezuela (treated as outlier due to refugee crisis) leaves main results consistent.

### Heterogeneous effects by gender (Table 6 summary)
- Female vs Male LFP change (selected 2SLS coefficients):
  - LFP Change (15+)
    - Female 2SLS: OECD emigration rate -9.272*** (2.445); Remittances per capita 4.310 (3.871)
    - Male 2SLS: OECD emigration rate -6.100*** (1.820); Remittances per capita 3.573 (2.882)
  - LFP Change (15-24)
    - Female 2SLS: OECD emigration rate -14.370*** (3.272); Remittances per capita 2.750 (5.181)
    - Male 2SLS: OECD emigration rate -15.078*** (3.522); Remittances per capita 3.712 (5.577)
- Interpretation:
  - Emigration is associated with declines in labor force participation for both genders.
  - The negative effect from emigration is especially strong for the overall female labor force.
  - Emigration effects for the younger cohort are large for both genders and similar in scale.

*Source: IMF Working Paper chapter 5.2 "Economic Growth" (results and tables as presented).*

### 6.2 Dynamic Effects

### 6.2 Dynamic Effects

### Longer horizon and baseline robustness
- Baseline sample period extended from 1998–2019 to 1995–2019 to assess longer-run impacts; UNPD emigration data used to focus on long-run effects.
- Longer-horizon estimates treated as robustness checks because emigration data are not available for all destination countries before 1999.

### Joint estimates (1995–2019) — Table 7 (Real GDP growth; LFP Change 15+; LFP Change 15–24)
- Estimation methods: OLS and 2SLS with the same instrumental variables and procedure as baseline; remittances per capita measured in US$1000.
- OECD emigration rate on Real GDP growth:
  - OLS: 0.12 (standard error (0.76))
  - 2SLS: -8.71*** (standard error (3.21))
- Remittances per capita on Real GDP growth:
  - OLS: 1.47 (standard error (1.215))
  - 2SLS: 19.29** (standard error (8.305))
- OECD emigration rate on LFP Change (15+):
  - OLS: 0.54 (standard error (0.554))
  - 2SLS: -3.17*** (standard error (1.045))
- Remittances per capita on LFP Change (15+):
  - OLS: 0.83 (standard error (0.988))
  - 2SLS: 4.73* (standard error (2.492))
- OECD emigration rate on LFP Change (15–24):
  - OLS: 1.16 (standard error (1.149))
  - 2SLS: -5.01*** (standard error (1.512))
- Remittances per capita on LFP Change (15–24):
  - OLS: 2.15 (standard error (2.082))
  - 2SLS: 5.43 (standard error (3.604))
- Sample sizes and model diagnostics:
  - Observations: Real GDP growth 724 (OLS and 2SLS), LFP Change series 630 (OLS and 2SLS).
  - First-stage observations reported as 724.
  - Controls: Yes. Fixed effects: country (o) and time (t).
  - Instruments: Income, LUR, ULCM.
  - Underidentification test statistic reported as 23.937 (where shown).
  - Endogeneity test statistic reported as 11.405 (where shown).
  - Sargan-Hansen test statistic reported as 0.416 (where shown).
- Interpretation: Emigration associated with negative impact on growth and decreases in labor force participation, especially among the younger cohort (15–24). Remittances provide some mitigation but not necessarily for the younger cohort.

### Current and past effects — methodology and Table 8 (Joint 2SLS)
- Identification strategy: one-year lagged endogenous variables instrumented by one-year lagged shift-share instruments; also estimate model including both current and one-year lagged endogenous variables with corresponding instruments (Jaeger, Ruist, and Stuhler (2018) approach).
- Over-identification: Adding lags raises Hansen test statistic; last specification reports p-value 0.161 (does not reject null at 0.15 significance level).
- Coefficients (2SLS only; Robust standard errors shown below coefficients):
  - Baseline (current):
    - OECD Emigration rate: -10.28** (standard error (4.326))
    - Remittances per capita: 19.93** (standard error (9.344))
  - Past (one-year lagged):
    - Lag OECD Emigration rate: -15.96*** (standard error (5.497))
    - Lag Remittances per capita: 25.50* (standard error (13.523))
  - Current and Past (both included):
    - OECD Emigration rate (current): 3.74 (standard error (4.225))
    - Remittances per capita (current): -42.04 (standard error (37.190))
    - Lag OECD Emigration rate: -14.02*** (standard error (4.724))
    - Lag Remittances per capita: 57.82 (standard error (42.848))
- Sample sizes and diagnostics:
  - Observations: Baseline 612, Past 608, Current and Past 608.
  - Controls: Yes (Past model uses one lag).
  - Fixed effects: country (o) and time (t).
  - Instruments: Income, LUR, ULCM (and their one-period lags where applicable).
  - Underidentification test statistics: 12.849 (Baseline), 12.666 (Past), 4.637 (Current and Past).
  - Endogeneity test statistics: 9.022 (Baseline), 9.485 (Past), 11.761 (Current and Past).
  - Sargan-Hansen test statistics: 0.123 (Baseline), 1.249 (Past), 3.656 (Current and Past).
- Interpretation:
  - Past emigration and past received remittances have larger and more significant impacts on economic growth than current effects.
  - Past emigration is the most persistent negative effect on economic growth.

### Five-year effects — Table 9 (1995–2020 five-year changes)
- Five-year variables constructed: emigration rate is change in 5-year emigrant stock over lagged 5-year population; remittances per capita is 5-year average over lagged 5-year population (2015 constant USD in thousands).
- Instruments adjusted to five-year growth/averages: income and manufacturing unit labor cost changes as five-year growth; unemployment rate as five-year average.
- Fixed effects: geographic fixed effects (g) and time (t); time-invariant characteristics controlled for instead of country fixed effects due to reduced sample size.
- Estimation results (OLS and 2SLS; robust standard errors shown below coefficients):
  - Emigration rate on Real GDP growth:
    - OLS: -3.18*** (standard error (0.935))
    - 2SLS: -3.37* (standard error (1.723))
  - Remittances per capita on Real GDP growth:
    - OLS: 10.83 (standard error (6.872))
    - 2SLS: 26.91 (standard error (26.322))
  - Emigration rate on LFP Change (15+):
    - OLS: -0.13 (standard error (0.125))
    - 2SLS: -0.38 (standard error (0.547))
  - Remittances per capita on LFP Change (15+):
    - OLS: 0.53 (standard error (1.396))
    - 2SLS: 12.83 (standard error (8.066))
  - Emigration rate on LFP Change (15–24):
    - OLS: 0.02 (standard error (0.139))
    - 2SLS: -0.52 (standard error (0.848))
  - Remittances per capita on LFP Change (15–24):
    - OLS: 2.48 (standard error (1.818))
    - 2SLS: 17.10 (standard error (10.987))
- Sample size and diagnostics:
  - Observations: 120 across all specifications.
  - Time-invariant controls: Yes (log real GDP, log population, trade openness in 1989).
  - Other controls: Yes (five-year averages of FDI as share of GDP, ODA as share of GNI, total natural disaster damages over five years as share of five-year lagged GDP).
  - Instruments: Income, Average LUR, ULCM (five-year representations).
  - Underidentification test: 6.029 (reported across estimations).
  - Endogeneity test statistics: 0.452; 5.320; 4.311 (reported across estimations).
  - Sargan-Hansen test statistics: 1.518; 0.018; 0.008 (reported across estimations).
- Interpretation:
  - Five-year data show similar relationship: emigration has a negative impact on growth (2SLS coefficient -3.37*).
  - Correlation between five-year average OECD emigration rate and five-year UNPD emigration rate is around 0.31.
  - Impact through remittances appears larger in scale but less significant in the five-year horizon; lagged remittances appear to have larger effects on growth.
  - Five-year effects on labor force participation are less significant, suggesting lower LFP may be primarily a short-run negative effect on economic growth.

### Summary of dynamic patterns
- Short run (current-period): Emigration associated with negative contemporaneous effects on growth and LFP; remittances can mitigate some negative effects.
- Medium run (one-year lag): Past emigration and past remittances often show larger and more significant effects on growth than current measures; lagged emigration displays persistent negative effects.
- Longer run (five-year): Negative impact of emigration on growth persists though coefficients differ due to different emigration measurements and smaller sample; remittances show larger point estimates but with less statistical significance for five-year aggregates.
- Labor force participation effects are strongest in the short run and concentrated among younger cohorts (15–24); five-year results for LFP are less significant.

### Robustness checks and interpretations referenced in section 6.3
- Alternative IV: A leave-one-out total immigrant inflow shift-share IV (subtracting own bilateral emigration flow) used to replace GNI per capita as a pull factor; results reported in Table A.5 yield similar estimates for economic growth (emigration negative; remittances mitigating) and qualitatively similar LFP results.
- Alternative base year: IVs constructed using historical migration stock in 1995 (with initial conditions from 1994) produce consistent findings (Table A.6).
- Alternative measures: Results robust to different variable measures — net emigration rate from the World Bank (Table A.9), remittances in constant 2015 USD (Table A.10), KNOMAD remittance data (Table A.11), and effects on real GDP growth per capita (Table A.12).
- Population denominator concern: Re-calculating endogenous variables using one-year lagged population yields coefficients almost the same as baseline (Table A.13), indicating population change in the denominator is not driving results.
- Levels specification: Using level of GDP instead of changes shows emigration negatively impacts contemporaneous GDP level; remittances positively impact contemporaneous GDP level but not significantly in 2SLS (Table A.14).

_Italicized source: IMF Working Paper — "The Joint Effect of Emigration and Remittances on Economic Growth and Labor Force Participation in Latin America and the Caribbean" (section 6.2 Dynamic Effects)._

### Annex II. Additional Tables and Figures

### Annex II. Additional Tables and Figures

### Main regression results: Real GDP growth and labor force participation (1995–2020; baseline)
- Table A.4 — Real GDP growth (OLS / 2SLS)
  - OECD Emigration rate: -0.77 / -6.07
  - Remittances per capita: 1.88 / 8.00
  - Observations: 749 (OLS), First-stage Obs.: 749 (2SLS)
  - Controls: Yes; Fixed effects: o,t; Instruments: Income, LUR, ULCM
  - Underidentification test: 26.405; Endogeneity test: 8.714; Sargan-Hansen test: 0.900
- Table A.4 — LFP Change (age 15+; OLS / 2SLS)
  - OECD Emigration rate: 0.14 / -3.35***
  - Remittances per capita: 1.94* / 5.37***
  - Observations: 652
- Table A.4 — LFP Change (age 15–24; OLS / 2SLS)
  - OECD Emigration rate: 0.95 / -3.44***
  - Remittances per capita: 3.48* / 2.42
  - Notes: Standard errors shown below coefficients. Remittances per capita is measured in US$1000.

### Alternative IVs and time-horizon robustness
- Table A.5 — Main results with alternative IV (1999–2019)
  - Real GDP growth (OLS / 2SLS): OECD Emigration rate -0.08 / -20.29**; Remittances per capita 2.18 / 15.19
  - LFP Change (15+): OECD Emigration rate 1.17* / -1.70; Remittances per capita 1.64 / 4.76**
  - LFP Change (15–24): OECD Emigration rate 2.80* / -2.21; Remittances per capita 4.24* / 5.54*
  - Observations: 612 (real GDP); 534 (LFP regressions)
  - Instruments: Immigration, LUR, ULCM; Underidentification test: 13.219; Endogeneity test: 17.752; Sargan-Hansen test: 1.062
- Table A.6 — Using 1995 as base year (real GDP growth regressions)
  - Country FE, 2SLS: OECD Emigration rate -15.570**; Remittances per capita 15.874*
  - Geographic FE, 2SLS: OECD Emigration rate -22.991**; Remittances per capita 22.060*
  - Observations: 612 (country FE), 482 (geographic FE)
  - Instruments: Income, LUR, ULCM; Underidentification tests: 9.900 / 8.572; Endogeneity tests: 10.668 / 36.923; Sargan-Hansen tests: 0.061 / 0.126

### Pre-trends and IV validity checks
- Table A.7 — Pre-trends with four-year lagged IV (1995–1998)
  - Pull IV: Income — Real GDP growth (Country FE / Geo FE): -7.97 / -16.29**
  - Pull IV: Unemployment rate — Real GDP growth (Country FE / Geo FE): -9.43 / -9.92
  - Pull IV: Manufact. ULC — Real GDP growth (Country FE / Geo FE): -192.54 / 42.42
  - Test F stat: 1.064 / 1.789 / 3.490 / 1.110 / 3.859 / 1.412; Test p-value: 0.369 / 0.156 / 0.020 / 0.350 / 0.013 / 0.246
  - Observations range: 92–123 depending on specification
  - Notes: Pull IVs constructed using four-year forward shocks. Tests under null that all instrument coefficients are zero.

### First-stage responses of endogenous variables to shift-share instruments across horizons
- Table A.8 — Responses to IVs across horizons (t = -2, -1, 0, 1, 2)
  - OECD Emigration Rate responses (GNIPC, LUR, ULCM)
    - t = -2: 1.058, -0.216, -3.436
    - t = -1: 0.562, 0.054, -5.229*
    - t = 0: 2.996**, 0.693, 0.776
    - t = 1: -0.627, -0.131, 1.342
    - t = 2: -0.235, -0.228, 2.715
    - Observations: 680 per column
  - Remittances per capita responses (GNIPC, LUR, ULCM)
    - t = -2: -0.090, -0.161, 4.471***
    - t = -1: 0.480*, 0.304*, 4.820***
    - t = 0: 0.607**, -0.669**, 6.394***
    - t = 1: 0.343, -0.192, 2.909*
    - t = 2: -0.002, 0.025, 3.043
    - Observations: 669 per column
  - Notes: Robust standard errors shown; first-stage F tests conducted under null that all coefficients of included instruments are zero. Other controls and time and country fixed effects included.

### Permutation exercises (shift-share robustness)
- Figure A.1 — Permutation of Shares (emigration and remittances)
  - Exercise: 500 replications; visualization reports 100 permutations.
- Figure A.2 — Permutation of Shifts (per capita income)
  - Average emigration effect: -10.18 (320 out of 500 times significant at the 10 percent level)
  - Average remittances effect: 19.34 (451 out of 500 times significant at the 10 percent level)
  - Exercise: 500 replications; visualization reports 100 permutations.
- Figure A.3 — Permutation of Shifts (LUR)
  - Average emigration effect: -10.93 (176 out of 500 times significant at the 10 percent level)
  - Average remittances effect: 19.50 (157 out of 500 times significant at the 10 percent level)
- Figure A.4 — Permutation of Shifts (ULCM)
  - Average emigration effect: -9.69 (493 out of 500 times significant at the 10 percent level)
  - Average remittances effect: 13.93 (106 out of 500 times significant at the 10 percent level)
- Notes across figures: Each exercise based on 500 replications; graphs report 100 permutations for visualization clarity.

### Heterogeneity and regional patterns
- Figure A.5 — Heterogeneous sub-regional individual and joint effects (excluding Venezuela)
  - Panels: (a) Effects on emigration and remittances; (b) Joint effect on economic growth
  - (Visuals presented in source; numerical details not tabulated in provided text excerpt)

### Robustness checks: alternative definitions and samples
- Table A.9 — Using Net Emigration Rate
  - Real GDP growth (OLS / 2SLS): Net Emigration/Population -0.06 / -9.07**
  - LFP Change (15+): 0.09 / -12.43***
  - LFP Change (15–24): 0.41 / -21.78***
  - Remittances per capita: 2.13 / 13.12 (real GDP); 1.89* / -7.69 (LFP 15+); 4.9** / -16.87 (LFP 15–24)
  - Observations: 612 (real GDP); 534 (LFP)
  - Instruments: Income, LUR, ULCM; Underidentification test: 2.039; Endogeneity test: 9.194; Sargan-Hansen test: 0.074
  - Note: Net emigration obtained by negative net migration from WB WDI.
- Table A.10 — Remittances in 2015 constant USD
  - Real GDP growth (OLS / 2SLS): OECD emigration rate -0.06 / -9.22**; Remittances per capita 2.49 / 29.84**
  - LFP Change (15+): OECD emigration rate 1.20* / -7.16***; Remittances per capita 1.22 / 6.00
  - LFP Change (15–24): OECD emigration rate 2.95** / -14.26***; Remittances per capita 3.13 / 4.91
  - Observations: 612 (real GDP); 534 (LFP)
  - Underidentification test: 13.850; Endogeneity test: 9.127; Sargan-Hansen test: 0.265
- Table A.11 — KNOMAD remittances (2000–2019)
  - Real GDP growth (OLS / 2SLS): OECD emigration rate -0.10 / -13.19; Remittances per capita 3.41** / 13.52
  - LFP Change (15+): OECD emigration rate 1.48** / -0.42; Remittances per capita 2.28** / 6.24**
  - LFP Change (15–24): OECD emigration rate 3.84*** / -1.00; Remittances per capita 5.17** / 8.16**
  - Observations: 583 (real GDP); 509 (LFP)
  - Instruments: Income, LUR, ULCM; Underidentification test: 7.995; Endogeneity test: 5.232; Sargan-Hansen test: 1.430
- Table A.12 — Using Real GDP per capita (growth of real GDP per capita)
  - Country FE, 2SLS: OECD emigration rate -9.43**; Remittances per capita 16.20*
  - Geographic FE, 2SLS: OECD emigration rate -10.85; Remittances per capita 8.27
  - Observations: 612 (country FE); 461 (geographic FE)
  - Underidentification tests: 12.849 / 7.589; Endogeneity tests: 7.319 / 20.530; Sargan-Hansen tests: 0.026 / 0.000
- Table A.13 — Lag population specifications
  - Real GDP growth (OLS / 2SLS): OECD Emigration rate (lag pop) -0.07 / -10.15**
  - Remittances per (lag) capita: 2.16 / 20.01**
  - LFP Change (15+): OECD Emigration rate 1.12 / -7.60***; Remittances per (lag) capita 1.61 / 0.40
  - LFP Change (15–24): OECD Emigration rate 2.76* / -14.59***; Remittances per (lag) capita 4.18* / 0.35
  - Observations: 612 (real GDP); 534 (LFP)
  - Underidentification test: 12.636; Endogeneity test: 9.050; Sargan-Hansen test: 0.117

### Joint estimates and log specifications
- Table A.14 — Joint estimates of emigration and remittances on Real GDP (logs), 1999–2019
  - Country FE, 2SLS: OECD emigration rate -0.089**; Remittances per capita 0.146
  - Geographic FE, 2SLS: OECD emigration rate -0.077**; Remittances per capita 0.108
  - Observations: 612 (all columns)
  - Time-invariant controls: No (country FE columns) / Yes (geographic FE columns)
  - Instruments: Income, LUR, ULCM; Underidentification tests: 15.380 / 17.031; Endogeneity tests: 5.296 / 4.472; Sargan-Hansen tests: 0.027 / 0.00
  - Notes: Dependent variable is percentage growth of real GDP in log specification; Remittances per capita measured in US$1000.

### Methodological notes recurring across tables
- Standard errors: Robust standard errors used in two-stage least squares estimations (and noted robust in several tables).
- Significance notation: *** p<.01, ** p<.05, * p<.10.
- Remittances per capita: generally measured in US$1000 (with specific Table A.10 noting 2015 constant US$1000).
- Time-invariant controls: typically include log real GDP, log population and trade openness in 1989 or 1994/1995 depending on specification.
- Other controls: FDI as a share of GDP, ODA as a share of GNI, natural disaster damages as a share of GDP.
- Fixed effects: Time (t) and either country (o) or geographic (g, Central America, Caribbean, South America) fixed effects are used in specifications.

*Source: Annex II. Additional Tables and Figures (IMF Working Paper excerpt).*

### References

### References

### Remittances and Economic Growth, Financial Development, and Development Finance
- Acosta, P., Calderón, C., Fajnzylber, P. and López, H. (2006), ‘Remittances and development in Latin America’, World Economy 29(7), 957–987.  
- Acosta, P. A., Lartey, E. K. and Mandelman, F. S. (2009), ‘Remittances and the Dutch disease’, Journal of International Economics 79, 102–116.  
- Aggarwal, R., Demirgüç-Kunt, A. and Pería, M. S. M. (2011), ‘Do remittances promote financial development?’, Journal of Development Economics 96(2), 255–264.  
- Barajas, A., Chami, R., Fullenkamp, C., Gapen, M. and Montiel, P. J. (2009), ‘Do Workers’ Remittances Promote Economic Growth?’, IMF Working Paper 2009/153, International Monetary Fund.  
- Beaton, M. K., Cerovic, M. S., Galdamez, M., Hadzi-Vaskov, M., Loyola, F., Koczan, Z., Lissovolik, M. B., Martijn, M. J. K., Ustyugova, M. Y. and Wong, J. (2017), ‘Migration and Remittances in Latin America and the Caribbean: Engines of Growth and Macroeconomic Stabilizers?, IMF Working Paper 2017/144, International Monetary Fund.  
- Chami, M. R., Barajas, M. A., Cosimano, M. T. F., Fullenkamp, C., Gapen, M. T. and Montiel, M. P. (2008), ‘Macroeconomic Consequences of Remittances’, IMF Occasional Paper, No. 259, International Monetary Fund.  
- Chami, R., Fullenkamp, C. and Jahjah, S. (2005), ‘Are Immigrant Remittance Flows a Source of Capital for Development?’, IMF Staff Papers 52(1), 55–81.  
- Chami, R., Ernst, E., Fullenkamp, C. and Oeking, A. (2018), ‘Are Remittances Good for Labor Markets in LICs, MICs and Fragile States?’, IMF Working Paper, No. 2018/108, International Monetary Fund.  
- Clemens, M. A. and McKenzie, D. (2018), ‘Why Don’t Remittances Appear to Affect Growth?’, Economic Journal 128, F179–F209.  
- Cazachevici, A., Havranek, T. and Horvath, R. (2020), ‘Remittances and Economic Growth: A Meta-Analysis’, World Development 134.  
- Imai, K. S., Gaiha, R., Ali, A. and Kaicker, N. (2014), ‘Remittances, Growth, and Poverty: New Evidence from Asian countries’, Journal of Policy Modeling 36, 524–538.  
- Lim, S. and Simmons, W. O. (2015), ‘Do Remittances Promote Economic Growth in the Caribbean Community and Common Market?’, Journal of Economics and Business 77, 42–59.  
- Feeny, S., Iamsiraroj, S. and McGillivray, M. (2014), ‘Remittances and Economic Growth: Larger Impacts in Smaller Countries?’, Journal of Development Studies 50, 1055–1066.  
- Saadi, M. (2020), ‘Remittance Inflows and Export Complexity: New Evidence from Developing and Emerging Countries’, Journal of Development Studies 56, 2266–2292.  
- Fromentin, V. (2018), ‘Remittances and Financial Development in Latin America and the Caribbean Countries: A Dynamic Approach’, Review of Development Economics 22, 808–826.  
- Askarov, Z. and Doucouliagos, H. (2020), ‘A Meta-Analysis of the Effects of Remittances on Household Education Expenditure’, World Development 129.  
- Buettner, T. (2022), ‘Stocktaking of Migration Data’, KNOMAD Paper NA(42).  
- Cazachevici, A., Havranek, T. and Horvath, R. (2020), ‘Remittances and Economic Growth: A Meta-Analysis’, World Development 134.  

### Migration, Labor Markets, Emigration, Brain Drain, and Household Effects
- Acosta, P. (2020), ‘Intra-household labour allocation, migration, and remittances in rural El Salvador’, The Journal of Development Studies 56(5), 1030–1047.  
- Amuedo-Dorantes, C. and Pozo, S. (2006), ‘Migration, remittances, and male and female employment patterns’, American Economic Review 96(2), 222–226.  
- Anelli, M., Basso, G., Ippedico, G. and Peri, G. (2023), ‘Emigration and Entrepreneurial Drain’, American Economic Journal: Applied Economics 15, 218–252.  
- Antman, F. M. (2013), ‘The Impact of Migration on Family left behind, in International Handbook on the Economics of Migration, Edward Elgar Publishing, pp. 293–308.  
- Berger, S. (2022), ‘Brain Drain, Brain Gain and Its Net Effect’, KNOMAD paper 46.  
- Bhagwati, J. and Hakamada, K. (1974), ‘The Brain Drain, International Integration of Markets for Professionals and Unemployment: A Theoretical Analysis’, Journal of Development Economics, 1, 19-42.  
- Clemens, M. A. (2011), ‘Economics and Emigration: Trillion-dollar Bills on the Sidewalk?’, Journal of Economic perspectives 25(3), 83–106.  
- Clemens, M. A. (2020), ‘The Emigration Life Cycle: How Development Shapes Emigration from Poor Countries, IZA Discussion Paper.  
- Docquier, F. and Rapoport, H. (2012), ‘Globalization, Brain Drain, and Development’, Journal of Economic Literature 50(3), 681–730.  
- Kerr, S. P., Kerr, W., Özden, Ç. and Parsons, C. (2016), ‘Global Talent Flows’, Journal of Economic Perspectives 30(4), 83–106.  
- Munshi, K. (2003), ‘Networks in the Modern Economy: Mexican Migrants in the US Labor Market’, The Quarterly Journal of Economics 118(2), 549–599.  
- McKenzie, D. and Rapoport, H. (2010), ‘Self-selection Patterns in Mexico-US Migration: the Role of Migration Networks’, the Review of Economics and Statistics 92(4), 811–821.  
- Imbert, C., Seror, M., Zhang, Y. and Zylberberg, Y. (2022), ‘Migrants and Firms: Evidence from China’, American Economic Review 112, 1885–1914.  
- Lim, S., Morshed, A. K. and Turnovsky, S. J. (2023), ‘Endogenous Labor Migration and Remittances: Macroeconomic and Welfare Consequences’, Journal of Development Economics 163.  
- Serrano, J., Gasparini, L., Marchionni, M., & Glüzmann, P. (2019). Economic Cycle and Deceleration of Female Labor Force Participation in Latin America. Journal for Labour Market Research 53, 1-21.  
- Jaeger, D. A., Ruist, J. and Stuhler, J. (2018), ‘Shift-share Instruments and the Impact of Immigration’, Technical report, National Bureau of Economic Research.  
- Tabellini, M. (2020), ‘Gifts of the Immigrants, Woes of the Natives: Lessons from the Age of Mass Migration’, The Review of Economic Studies 87(1), 454–486.  
- Mayda, A. M., Peri, G. and Steingress, W. (2022), ‘The Political Impact of Immigration: Evidence from the United States’, American Economic Journal: Applied Economics 14, 358–389.  
- Lessem, R. (2018), ‘Mexico–US Immigration: Effects of Wages and Border Enforcement’, The Review of Economic Studies 85(4), 2353–2389.  
- Kleemans, M. and Magruder, J. (2018), ‘Labour Market Responses to Immigration: Evidence from Internal Migration Driven by Weather Shocks’, The Economic Journal 128(613), 2032–2065.  
- Fouka, V., Mazumder, S. and Tabellini, M. (2022), ‘From Immigrants to Americans: Race and Assimilation During the Great Migration’, The Review of Economic Studies 89(2), 811–842.  
- Autor, D. H. and Duggan, M. G. (2003), ‘The Rise in the Disability Rolls and the Decline in Unemployment’, The Quarterly Journal of Economics 118(1), 157–206.  

### Education, Child Labor, and Household Expenditure
- Alcaraz, C., Chiquiar, D. and Salcedo, A. (2012), ‘Remittances, schooling, and child labor in Mexico’, Journal of Development Economics 97(1), 156–165.  
- Askarov, Z. and Doucouliagos, H. (2020), ‘A Meta-Analysis of the Effects of Remittances on Household Education Expenditure’, World Development 129.  

### Methodology, Instruments, and Technical Contributions
- Pacini, D. and Windmeijer, F. (2016), ‘Robust Inference for the Two-Sample 2SLS estimator’, Economics Letters 146, 50–54.  
- Wang, Y. and Bellemare, M. F. (2019), ‘Lagged variables as instruments’, Technical report, Working Paper, Department of Applied Economics, University of Minnesota.  
- Jaeger, D. A., Ruist, J. and Stuhler, J. (2018), ‘Shift-share Instruments and the Impact of Immigration’, Technical report, National Bureau of Economic Research.  

### Regional, Historical, and Data Sources
- Ratha, D. (2003), ‘Workers‚ Remittances: an Important and Stable Source of External Development Finance’, Global Development Finance.  
- UNPD (2020), ‘International Migrant Stock 2020’. URL: https://www.un.org/development/desa/pd/content/international-migrant-stock  
- Babii, A, A. Carare, D. Vasiliev, and Y. Yakhshilikhov (2022), ‘Evolution of Remittances to CAPDR and Mexico During the COVID-19 Pandemic’, IMF Working Paper 2022/092.  
- Beltran, P. and M. Hadzi-Vaskov (2023), ‘Climate and Cross-Border Migration”, IMF Working Paper 2023/250, International Monetary Fund.  

*Source: wpiea2024175-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024175-print-pdf.pdf_