The Role of Remittances in Shaping Income Inequality in Lebanon Before and After the Crisis: An Empirical Analysis Using Macroeconomic and Financial Perspectives

Abstract

This study investigates the impact of remittances on income inequality in Lebanon using annual time-series data for the years 2000–2023. Applying Johansen’s cointegration test, with financial development (FD), GDP, and household consumption expenditure (HCE) as the control variables, the study examines the long-run and short-run relationship between remittances and inequality. The study also considers the moderating impacts of FD and HCE to account for their indirect role in the remittance–inequality relationship. Dynamic relations are also examined by using impulse response functions (IRFs) and Forecast Error Variance Decomposition (FEVD) analyses. The long-run model estimates validate that remittances and income inequality are significantly and negatively related, i.e., increased remittance receipts serve to reduce income inequality in Lebanon. Remittance effects, however, are statistically insignificant in the short run. Interestingly, the results reveal that financial development weakens the remittances’ inequality-reducing effect, dampening their impact. Contrarily, a higher household consumption expenditure slightly strengthens the inequality-reducing effect of remittances. A comparison between the pre- and post-2019 periods reveals that the explanatory strength of remittances weakened during times of economic crisis, since the function of remittances was different during times of economic distress. Based on these findings, this study recommends that Lebanon not only promote financial development but also focus on financial inclusion, improve social safety nets, and provide inclusive economic growth to maximize remittance inflow benefits and efficiently reduce inequality.

1. Introduction

In the contemporary era, remittances have emerged as a vital source of external finance for developing economies, playing an expanding role in capital inflows. The volume of remittances has skyrocketed due to the heightened labor portability brought about by globalization, prompting researchers and policymakers to seize this unique opportunity to investigate both the microeconomic and macroeconomic impacts of these financial flows [1].

Lebanon’s large diaspora and long-term mobility have rendered it one of the most remittance-dependent economies globally. Lebanon received remittances equivalent to 17.2% of its GDP in 2022, ranking it among the top 10 remittance-dependent countries globally [2]. Remittances observed a severe increase in Lebanon, driven up from 13.3% of GDP in 2003 to a peak of 24.7% of GDP in 2009. After 2020, there was also a sharp increase, reaching 35.5% of GDP in 2023 (see Figure 1) [3]. For many Lebanese homes, these remittances are a lifeline, offering vital support during unrest in the country’s economy and politics. Yet, despite this substantial inflow, Lebanon is said to be one of the countries that have the highest rates of income inequality in the world, ranking 129th out of 141 countries in the distribution of income [4]. The Gini coefficient, a measure of income inequality, rose from 0.42 in 2011/2012 to 0.64 in 2022/2023 among Lebanese households (see Figure 2) [4]. This paradox raises an important empirical question: Do remittances assist in reducing income inequality in Lebanon, or do they reinforce existing inequalities?

Although remittances are widely recognized for their poverty-reducing impact, their impact on income inequality is poorly established and context-sensitive [5]. Figure 2 shows trends in remittances as a share of GDP and the Gini coefficient (income inequality measure) for Lebanon between 2000 and 2023. The Gini coefficient is always high throughout the period, which indicates ongoing income inequality. Remittances vary greatly over time relative to this. They rose steeply during the first decade of the 2000s, declined gradually during 2005–2017, and again increased significantly after 2019. This divergence signals a rich intersection between remittance inflows and income inequality and hence is imperative to explore through how remittances influence the dynamics of inequality in Lebanon. To the best of the authors’ knowledge, while a vast strand of economic literature has analyzed the remittance effect on a range of topics, no official time-series empirical work on the impact of remittances on Lebanon’s income inequality exists.

Given the country’s economic situation, political and financial instability, and heavy reliance on remittance inflows, the aim of this study is to empirically examine the relationship between remittances and income inequality in Lebanon, particularly following the 2019 economic crisis, with a perspective to promote inclusive growth and sustainable development. Accordingly, to this end, the study pursues the following specific objectives: (1) To investigate the impact of remittances on income inequality in Lebanon using the Johansen cointegration approach, considering both short-run and long-run dynamics for the years 2000–2023. (2) To consider the key macroeconomic variables—GDP, household consumption expenditure (HCE), and financial development (FD)—to examine their direct and indirect effects on income inequality. (3) To analyze the impact of the 2019 economic crisis on the relationship between remittances and inequality using a split-sample analysis (pre- and post-crisis).

FD has the potential to affect the ability of remittances to reduce inequality because stronger financial systems will facilitate greater mobilization of remittances towards savings and investment. HCE measures household welfare, and levels of consumption could reflect the fact that remittances are increasing living standards. GDP is taken to measure overall economic performance. Remittances–FD interaction terms and remittances–HCE interaction terms are used to test if the impact of remittances on inequality depends upon financial conditions or levels of consumption in households.

This paper fills a significant gap in the regional literature by examining the pre- and post-2019 financial and economic crisis era, arguing for the functioning of remittances within the context of economic shocks. Given the limited studies on Lebanon in this area, this research offers valuable insights. The rest of the paper is structured as follows: Section 2 presents a review of the existing body of literature; Section 3 deals with the methodology; Section 4 shows the empirical results and discussions; and Section 5 offers concluding remarks.

2. Literature Review

For many years, remittances have been considered a major tool for alleviating income disparity and poverty in underdeveloped countries. Hanson [6] defines remittances as funds transferred by foreign residents to their families. Over recent decades, these remittances have increased dramatically and have been the prime source of foreign revenue for numerous developing countries. These amounts of funds are in billions of dollars and can account for a significant percentage of gross domestic product (GDP) [6]. Migrant remittances refer to goods and funds that are transferred to households by migrant workers who are employed outside the countries to which the remittances are sent [7]. Remittances are unadulterated income transfers—that is, they are the transfer of ownership of commodities from a migrant household to a recipient household with no limitations placed upon their use [8]. This suggests that remittances are not capital flows and do not include a merit good component [8].

The theoretical literature, such as that by Stark [9] and Lucas and Stark [10], was some of the first to espouse altruism and insurance as drivers of remittance flows and claim that they are private welfare transfers that will smooth out consumption and inequality.

The permanent income hypothesis by Friedman states that consumers do not change their consumption level in the short run when they are given additional income (e.g., remittances) because they view these as being temporary or indefinite. They save or smooth consumption to delay the redistributive impact [11]. Remittances are insignificant during the early years and have minimal impacts on inequality, according to the migration life-cycle theory [12]. This insignificant relation may be due to economic adjustment lags, as argued by Ajefu and Ogebe [13], in that time is needed for remittances to affect income distribution via structural change or investment. In addition, the migration selectivity hypothesis contends that middle-income households that can afford the cost of migration receive short-run remittances, and thus inequality does not fall—and even increases [14]. Subsequently, remittances in the short term tend to be consumed rather than used for productive investment and thus have narrow redistributive impacts [15].

Empirical studies have, however, reported inconclusive findings about the impact of remittances on income inequality in different contexts. Motivated by this early speculation, a meta-analysis exposed that remittances had fairly context-dependent effects on inequality [16]. Cross-country and household-level empirical studies also contribute to the evidence for remittances as drivers of inequality decline. An increase in per capita remittances by 10% was associated with a 1.6% decline in the Gini coefficient [17]. The World Bank [2] also corroborated that remittance flows are associated with more equitable income distributions in poorer economies.

Yet, more recent studies have added to the knowledge of remittance effects in a range of contexts. Moustapha [18], in Senegal, identified an “internal sharing effect” by which international remittance recipients were more likely to receive internal cash transfers, reducing inequality. In their report, Pashain et al. [19] in Pakistan ascertained that foreign remittances enhanced the well-being of households, but the heightened gains profited high-income households, thus escalating spending inequality. The effect is not everywhere an equalizing one. Evidence, as provided by Islam and Mondal [20] and Inoue [21], illustrates that where migration is expensive, the remittance gains go to middle- and upper-income groups. Vo et al. [22] examined the long-run effect of financial development on income inequality. The findings revealed an inverted U-shaped relationship that initially exacerbates inequality, then reduces it after a specific threshold. Karikari et al. [23] found a positive long-run relation and a slightly positive short-run relation between financial development and income inequality. A 2023 SSRN working paper by Ganić found short-run and long-run causality between financial inclusion and income inequality [24]. Ofori, Gbolonyo, Dossou, and Nkrumah [25] observed that both remittances and financial development have a significant impact on income inequality. Swamy [26] concluded that rising consumption inequality tends to increase income inequality. Hornok and Raeskyesa [27] studied Indonesia’s GDP impact on income inequality. The results concluded that economic growth increases inequality in the long run, and reduces it in the short run, but Kuznets’ [28] inverted-U hypothesis cautions that the initial stages of economic growth can generate rising inequality. However, a nonlinear and negative relationship was found by Wan, Lu, and Chen [29] between income inequality and economic growth. A University of Barcelona working paper found a positive short-term correlation between GDP and inequality but found that long-term effects diverge depending on the region [30]. In light of these studies, the following hypothesis is formulated:

H1. 

Remittances have a significant long-run and/or short-run effect on income inequality in Lebanon.

Acosta et al. [31] determined, when analyzing Latin America, that remittances performed most effectively at reducing inequality when financial markets were underdeveloped. Empirical proof produced by León et al. [32] corroborates these findings to suggest that remittances have a tendency to reduce income inequality in financially underdeveloped economies, but the size of the impact varies with the strength of institutions. Aggarwal, Demirgüc-Kunt, and Pería [33], in their study, say that remittances foster financial inclusion in advanced banking systems, thus enhancing their inequality-reducing effects. In the same perspective, Beine and Lodigiani [34] affirmed that financial openness heightens the effects of remittances by enabling greater room for investment. Using a dynamic panel threshold method, Mili and Tarchoun [35] confirmed that remittances replace financial development in very unequal countries but complement financial development in more equal societies. In additional evidence, Simionescu [36] employed panel quantile regression with non-additive fixed effects (QRPD) to investigate the effect of the interaction between remittances and financial development on income inequality across the EU-27 countries from 1990 to 2023. The results revealed that this interaction had heterogeneous effects on income inequality, varying significantly across different levels of the income distribution—highlighting the importance of considering distributional impacts rather than relying solely on average effects. However, Karim et al. [37] suggested that financial development strengthens the negative impact of remittances on income inequality in South Asian countries. Longitudinal studies also establish the conditional character of financial development in terms of remittances’ long-run effects on inequality. Adams and Page [17] and Gupta et al. [38] inferred that stable remittance flows can reduce long-run income inequality if financial institutions are robust and inclusive.

Giuliano and Ruiz-Arranz [39] reported that remittances may affect income inequality through enhanced consumption and investment opportunities. In addition, Banerjee and Duflo [40] note that longstanding inequality in household consumption is likely to be an approximation of underlying structural income disparities. The short-run impacts are more varied: Acosta et al. [31], Beck et al. [41], and Deaton [42] conclude that remittances increase household consumption, but the impact is not evenly distributed, where the impact accrues to higher-income groups initially, and trickle-down effects only follow subsequently, thus increasing inequality in the short run. Similarly, shocks to consumption and GDP will also disproportionately burden the poor in the short run [43]. Recent empirical evidence strengthens this view. Ojeyinka and Ibukun [44] confirm that household consumption expenditure—used as a poverty proxy—is directly impacted by remittances across 38 countries in Africa, Asia, and Latin America. Their findings show that remittances are most effective in reducing poverty where household consumption levels are already relatively high, implying that consumption mediates the inequality-reducing effect of remittances. Similarly, Keho [45] demonstrates that remittances only increase investment and reduce inequality when financial development (measured by domestic credit to the private sector) exceeds a threshold of approximately 21% of GDP. Below this level, remittances may fail to trickle down or may even worsen inequality, due to inefficient financial intermediation. Based on these historical and contemporary studies, we pondered if financial development and household consumption expenditure play a moderating role in the remittances–inequality relationship and proposed the following hypothesis:

H2. 

At least one interaction term (Remittances×FD or Remittances×HCE) has a significant effect on income inequality.

Medium-term analysis using Forecast Error Variance Decomposition (FEVD) techniques further highlights the crucial role of remittances in income inequality. Combes and Ebeke [46] found that remittance shocks contributed between 20% and 40% of the forecast error variance to the inequality measures of African and Latin American countries, particularly during economic crises, since remittances serve as prime counter-cyclical stabilizers. Ebeke and Hakura [47] present opposing evidence. For their study in Mexico, using the FEVD technique, they demonstrated that remittances contributed more to income inequality before the 2019 financial and economic crisis era than after. Their findings suggest that the stabilizing role of remittances was weakened during the 2019 financial and economic crisis periods. Their conclusion contradicts the implications that remittances always act as effective stabilizers in times of economic crisis.

Additionally, the dynamic reaction of remittance flows has also come to be increasingly emphasized. Calderón and Page [48], using the help of impulse response functions (IRFs), noted that positive remittance shocks in the short term have a modest inequality-increasing impact, as initial recipients belong to relatively better-off groups. However, at a 2- to 3-year horizon, the shocks do help in declining inequality, which is consistent with the hypothesis that remittances are an insurance mechanism in the long run. Recent work by Ojeyinka and Ibukun [44] extends this dynamic analysis to 38 developing countries across Africa, Asia, and Latin America. Using robust dynamic GMM estimation, they show that remittances significantly reduce household poverty and increase consumption expenditure over time—especially in high-remittance economies. Their findings validate the stabilizing role of remittances across multiple regions and support the use of variance decomposition and impulse modeling to assess income dynamics over different time horizons.

Additional support comes from Shair and Anwar [49], who investigated the effects of both internal and external remittances on expenditure inequality in Pakistan. Their study found that remittances increased inequality in the short term, especially among higher-income quantiles, but that these effects evolved over time, supporting a dynamic inequality transmission mechanism. Similarly, El-Sayed and Aydın [50], using data from European countries, found that remittance shocks temporarily reduced income inequality, with remittances accounting for 10–15% of the GINI index’s forecast error variance over a 5-year horizon.

Based on these dynamic findings, the following hypotheses are put forward:

H3. 

Remittances explain a significant share of the forecast error variance of the GINI index.

H4. 

A shock to remittances has a significant effect on income inequality over time.

The Lebanese context offers a specific context within which to view these processes. Remittances have been a feature of the Lebanese economy for many years, and they have represented a significant proportion of GDP as well as a major source of family incomes [51]. The 2019 financial and economic crisis essentially altered this equation, throwing doubt on whether remittances would be capable of continuing their redistributive role. Additionally, Bouhga-Hagbe’s [52] study had argued that in economies like Lebanon’s, remittances are diverted away from investment into present consumption, thereby lessening their long-term implications for inequality. Subsequently, more precisely, El Khoury [53] has suggested that during the post-crisis period, Lebanon has seen a declining developmental role of remittances, and their role in aiding equality on an income basis needs to be downwardly modified. Hence, the paper adopts a split-sample approach to investigate empirically if the impacts of remittances on inequality are heterogeneous between the pre- and post-crisis periods. Figure 3 shows the conceptual model underlying the study, reflecting the variables used, the hypothesized relationships among them, and how the model is derived from both theoretical foundations and empirical studies.

3. Methodology

3.1. Research Design

The study adopts annual time-series analysis based on secondary quantitative data to evaluate the impact of remittances on income inequality in Lebanon, covering the period from 2000 to 2023. The years chosen are limited to 24 years due to the unavailability of data for the remaining years. The study investigates both the long-run equilibrium relationships and the short-run dynamics between remittances and inequality. The data are obtained from reputable international sources, and the period is divided into a full sample and two sub-periods (pre- and post-2019) to capture any potential effects of the 2019 economic crisis in Lebanon. All empirical results were obtained using Eviews 10 Software. Following the methodologies adopted by authors like Strak et al. [9], León [32], Chukwuone [54], Borja [55], Dada and Akinlo [56], and Mili and Tarchoun [35], the paper integrates their approaches into the analysis within the Lebanese scope. Their contribution was linked to several macroeconomic controls, in addition to using interaction terms, to evaluate how financial development and consumption behavior would act as moderators of the remittances–inequality relationship.

3.2. Variables Description and Data Sources

The dependent variable in this study is income inequality, with the Gini coefficient (Gini) used as its proxy. The data were collected from the Standardized World Income Inequality Database (SWIID) [57] and the datapoints range from 0 to 1, with 0 indicating perfect equality, where everyone has the same income, and 1 indicating maximum inequality, where only one person receives the whole income and everyone else receives nothing. Remittances (Remit) are the main independent variable, measured as personal remittances received as a percentage of GDP, and their data were collected from the World Bank’s Migration and Remittances Database [2]. Several control variables are used in this study, which are the log of gross domestic product (log(GDP)), household consumption expenditure (HCE), and financial development (FD). The first two were sourced from the World Bank and the third from the IMF Financial Development Index Database [58,59]. Log GDP is taken as the natural logarithm of real GDP measured in constant 2015 US dollars. The natural logarithm is used to mitigate skewness and serves as a proxy to reflect the economic size. Household consumption expenditure (HCE) is used as a proxy that measures the final consumption expenditure of households, measured in constant 2015 US dollars. Financial development (FD) comprises two indices, which are the Financial Markets Index and the Financial Institutions Index, capturing the inclusiveness and performance quality of the financial markets and institutions.

To capture the interaction effects, the model includes the following researcher-constructed interaction terms: Remittances × Financial Development (RemitFD), and Remittances × Household Consumption Expenditure (RemitHCE). These are included to examine how the impact of remittances on income inequality varies with financial development and household consumption when they are acting as moderators.

Two dummy variables are utilized to account for significant structural breaks in the Lebanese economy. D2019, a crisis dummy equal to 1 for the 2019 Lebanese economic crisis and beyond (0 otherwise), is constructed to capture the onset of Lebanon’s financial and political crisis. D2006, a war dummy equal to 1 for the year 2006 (0 otherwise), captures the impact of the 2006 war on inequality. Both are constructed by the researchers based on historical events.

3.3. Empirical Models

The following equations describe the relationships between the studied variables.

Model 1: $\mathit{G}\mathit{i}\mathit{n}\mathit{i}=\mathit{f}(\mathit{R}\mathit{e}\mathit{m},\mathit{F}\mathit{D},\mathit{l}\mathit{o}\mathit{g}(\mathit{G}\mathit{D}\mathit{P}),\mathit{H}\mathit{C}\mathit{E},\mathit{D}\mathbf{2019},\mathit{D}\mathbf{2006})$

Model 2: $\mathit{G}\mathit{i}\mathit{n}\mathit{i}=\mathit{f}(\mathit{R}\mathit{e}\mathit{m},\mathit{R}\mathit{e}\mathit{m}\mathit{i}\mathit{t}\mathit{F}\mathit{D},\mathit{R}\mathit{e}\mathit{m}\mathit{i}\mathit{t}\mathit{H}\mathit{C}\mathit{E},\mathit{D}\mathbf{2019},\mathit{D}\mathbf{2006})$

The following equations represent the long-run and short-run model specifications adopted in this study (Model 1 and Model 2, respectively) based on methodologies from the prior literature:

Model 1:

• Long-Run Equation:

$$\begin{gathered} Gin{i}_t={\beta }_0+\sum _{i=1}^v{\beta }_{1i}Gin{i}_{t-i}+\sum _{j=0}^m{\beta }_{2j}Re{m}_{t-j}+\sum _{k=0}^n{\beta }_{3k}F{D}_{t-k}+\sum _{l=0}^r{\beta }_{4l}{log\left(GDP\right)}_{t-l}+\sum _{m=0}^s{\beta }_{5m}HC{E}_{t-m} \\ +{\beta }_{6}D2019+{\beta }_{7}D2006+{\epsilon }_t \end{gathered}$$

M1-A

• Short-Run Equation:

$$\begin{gathered} \Delta Gin{i}_t={\alpha }_0+\sum _{I=1}^p{\alpha }_{1j}{\Delta Gini}_{t-i}+\sum _{j=0}^m{\alpha }_{2j} \Delta Re{m}_{t-j}+\sum _{k=0}^n{\alpha }_{3k} {\Delta FD}_{t-k}+\sum _{l=0}^r{\alpha }_{4l} {\Delta log\left(GDP\right)}_{t-1}+\sum _{m=0}^s{\alpha }_{5m} {\Delta HCE}_{t-m}+{\alpha }_{6}D2019 \\ +{\alpha }_{7}D2006+{\lambda ECT}_{t-1}+{\epsilon }_t \end{gathered}$$

M1-B

Model 2:

• Long-Run Equation:

$$Gini\_t={\beta }_0+\sum _{j=0}^m {\beta }_{1j}Re{m}_{t-j}+\sum _{k=0}^n{\beta }_{2k}{(Rem\times HCE)}_{t-k}+\sum _{l=0}^r{\beta }_{3l}{(Rem\times FD)}_{t-l}+{\beta }_{4}D2019+{\beta }_{5}D2006+{\epsilon }_t$$

M2-A

• Short-Run Equation:

$$\Delta Gin{i}_t={\alpha }_0+\sum _{j=1}^m{\alpha }_{1j}\Delta Re{m}_{t-j}+\sum _{k=0}^n{\alpha }_{2k} {\Delta (Rem\times HCE)}_{t-k}+\sum _{L=0}^r{\alpha }_{3l} {\Delta (Rem\times FD)}_{t-l}+ {\alpha }_{4}D2019+{\alpha }_{5}D2006+\lambda EC{T}_{t-1}+{\epsilon }_t$$

M2-B

• where:
• LOGGDP_t−1: Lagged natural logarithm of GDP.
• εₜ: error term.
• Δ: First difference operator.
• ${\mathrm{E}\mathrm{C}\mathrm{T}}_{\mathrm{t}-1}$: Error correction term from the long-run cointegration equation.

This specification allows us to examine the long-run relationship between remittances and income inequality while controlling for financial development, economic size, consumption expenditure, and structural shocks in Lebanon.

3.4. Estimation Tests

Johansen’s cointegration test is used in this study to check for the number and the existence of long-run cointegrations among the variables. This test is more robust than the Engle–Granger procedure, especially with more than one cointegrating relationship [60]. This test is utilized when all the variables are stationary at first difference I(1) and not stationary at level I(0); it consists of five steps:

• Unit root tests are conducted using the augmented Dickey-Fuller (ADF) test to assess the stationarity of each variable.
• The optimal lag length for each equation is calculated using VAR selection criteria
• Johansen’s cointegration test is applied to estimate any long-run cointegration among the variables.
• Once long-run cointegration is proved, the VECM is used to generate the long-run and short-run coefficients.
• Granger causality tests are performed to evaluate the direction of causality between the variables.

As a last step, diagnostic tests, including heteroskedasticity, serial correlation, autocorrelation, and model stability tests, are generated to ensure the robustness of the model used and the results. In this study, Model 1 is also evaluated using Forecast Error Variance Decomposition (FEVD) and impulse response functions (IRFs).

4. Results and Discussion

4.1. Testing for Data Stationarity

The augmented Dickey–Fuller (ADF) test is a widely used statistical method for assessing the presence of a unit root in a time series [61], which helps determine the series’ stationarity. In this study, the ADF test is applied to ascertain the order of integration of the variables utilized in the two models. Taking household consumption expenditure (HCE) and the Gini coefficient as examples, the ADF model tests for a unit root as follows:

$$\Delta HCE=\alpha +\beta t+\gamma {HCE}_{t-1}+\sum _{i=1}^p\delta i{\Delta HCE}_{t-1}+\epsilon t$$

$$\Delta Gini=\alpha +\gamma {Gini}_{t-1}+\sum _{i=1}^p\delta i{\Delta Gini}_{t-1}+\epsilon t$$

The ADF test examines the null hypothesis that a unit root exists in the series, indicating non-stationarity, against the alternative hypothesis that the series is stationary. According to

Table 1. ADF stationary result.

Variable	Level I(0)	1st Difference I(1)	Decision
Gini	0.1586 (Intercept)	6.472733 *** (Intercept)	Not-Stationary at I(0), Stationary at I(1)
FD	−2.165934 (Intercept)	−6.474767 *** (None)	Not-Stationary at I(0), Stationary at I(1)
Rem	0.576431(None)	0.0085 * (None)	Not-Stationary at I(0), Stationary at I(1)
log(GDP)	−2.396210 (Intercept)	−2.457267 ** (None)	Not-Stationary at I(0), Stationary at I(1)
HCE (Trend & Intercept)	−2.762224 (Trend & Intercept)	−2.760455 * (None)	Not-Stationary at I(0), Stationary at I(1)
Remit × FD	0.396866 (None)	−1.692846 * (None)	Not-Stationary at I(0), Stationary at I(1)
Remit × HCE	0.551861 (None)	−1.954341 ** (None)	Not-Stationary at I(0), Stationary at I(1)

Note: * shows p < 0.1; ** shows p < 0.05; *** shows p < 0.001. Source: Authors’ Computations.

, it indicates that the variables are stationary only after first differencing. When at level, a unit root exists where we failed to reject the null hypothesis of the presence of the unit root. As a result, all the tested variables are non-stationary at level. Consequently, Johansen’s cointegration test is suitable for this condition, which requires all the variables to be I(1) and not I(0). Some tests were conducted with “None”, others with “Intercept” or “Trend & Intercept”. The choice of specification was based on the graphical representation of each variable to detect the presence of a trend, as well as the statistical significance of the intercept and trend components in the ADF test.

Based on the shape of Figure 4, and the p-value of the trend coefficient and intercept (p = 0.989, p = 0.0068, respectively), the trend is not statistically significant, but the intercept is. This indicates that a linear deterministic trend is not present in the LOGGDP series. Thus, it does not exhibit a distinct increasing or decreasing trend, despite a sudden spike between 2006 and 2008, presumably caused by a structural break.

In Figure 5, the HCE series shows a sustained upward movement in 2019, with a peak in 2022. This sharp rise reflects a strong deterministic trend, consistent with the statistical significance of the trend component in the ADF test. The intercept (constant term) and trend are both statistically significant (p = 0.0163, p = 0.0119, respectively) under the trend-and-intercept specification. This confirms the existence of a deterministic trend in the series, not a purely stochastic trend or drift. Consequently, the long-term upward movement in HCE is systematic and time-driven rather than purely the result of random shocks.

4.2. VAR Lag Length Selection Test

Selecting the optimal lag length is a significant step before performing the Johansen test because it is highly sensitive to the number of lags included in the model. Various criteria were used for this analysis, such as the SC, LR, AIC, or HQ.

Table 2. VAR lag order selection criteria of Model 1.

Order	LL	LR	FPE	AIC	SC	HQ
0	−73.90031	NA	0.020735	7.469592	8.062024	7.618587
1	−10.66561	87.97871 *	0.000367 *	3.362227 *	4.744568 *	3.709882 *

* indicates lag order selected by the criterion, LL: Log Likelihood, LR: sequential modified LR test statistic (each test at 5% level), FPE: Final prediction error, AIC: Akaike information criterion, SC: Schwarz information criterion, HQ: Hannan-Quinn information criterion. Source: Authors’ Computations.

and

Table 3. VAR lag order selection criteria of Model 2.

Order	LL	LR	FPE	AIC	SC	HQ
0	−177.5677	NA	170.4805	16.48415	17.07658	16.63315
1	−127.8146	69.22169 *	9.749764 *	13.54910 *	14.93144 *	13.89675 *

* indicates lag order selected by the criterion, LL: Log Likelihood, LR: sequential modified LR test statistic (each test at 5% level), FPE: Final prediction error, AIC: Akaike information criterion, SC: Schwarz information criterion, HQ: Hannan-Quinn information criterion. Source: Authors’ Computations.

show that for both models, Lag 1 is optimal according to all standard selection criteria. This implies that the dynamics of income inequality (GINI index) in response to remittances and macroeconomic controls (Model 1), as well as to remittances and their interaction terms (Model 2), are best captured with one period of lag in the VAR specification.

4.3. Johansen’s Cointegration Test

The Johansen test with an optimal lag length of 1 was conducted utilizing trace and maximum Eigenvalue statistics. Dummy variables as exogenous regressors were included for the years 2006 and 2019.

The outcomes for Model 1 and Model 2 using both tests are presented in

Table 4. Johansen Cointegration Test Results for Model 1 and Model 2 (Trace and Max-Eigen Tests).

Cointegration Rank	Trace p-Value (M1)	Max-Eigen p-Value (M1)	Trace p-Value (M2)	Max-Eigen p-Value (M2)	Significant at 5%
None	0.0400 *	0.2712	0.0005 *	0.0028 *	Yes (Trace M1 & Both M2)
At most 1	0.0838	0.3131	0.0572	0.1456	No
At most 2	0.1103	0.3602	0.1727	0.2925	No
At most 3	0.0263 *	0.0263 *	0.0967	0.0967	Yes (M1 only)

* denotes rejection of the hypothesis at the 0.05 level. Source: Authors’ Computations.

. The table shows the existence of at least one cointegrating relationship in both models. In Model 1, the Trace test indicates one cointegrating equation at the 5% significance level, but the Max-Eigenvalue test does not match this result. However, the two tests confirm a statistically significant equation at the “At most 3” level. Both tests indicate one cointegrating relationship at the 5% level in Model 2. This reflects a stable long-run equilibrium relationship among the variables in both models, with stronger statistical evidence in Model 2. The outcome justifies proceeding with a Vector Error Correction Model (VECM) to analyze both long-run and short-run coefficients.

4.4. Vector Error Correction Model (VECM) Results and Analysis

Table 5. Long-Run Estimates.

Model	Long-Run Equation	Interpretation
Model 1 (M1-A)	$$\begin{gathered} \mathrm{G}\mathrm{I}\mathrm{N}\mathrm{I}=0.875-{0.0036}^{*}\mathrm{R}\mathrm{E}\mathrm{M}\mathrm{I}\mathrm{T}- \\ {0.7297}^{*}\mathrm{F}\mathrm{D}+{0.0809}^{*}\mathrm{L}\mathrm{O}\mathrm{G}\mathrm{G}\mathrm{D}\mathrm{P}+ \\ {0.0007}^{*}\mathrm{H}\mathrm{C}\mathrm{E} \end{gathered}$$	REMIT and FD coefficients are negative and statistically significant; LOGGDP and HCE have a positive and significant effect on income inequality
Model 2 (M2-A)	$$\begin{gathered} \mathrm{G}\mathrm{I}\mathrm{N}\mathrm{I}=-0.779-{0.0261}^{*}\mathrm{R}\mathrm{E}\mathrm{M}\mathrm{I}\mathrm{T}+ \\ {0.0892}^{*}\mathrm{R}\mathrm{E}\mathrm{M}\mathrm{I}\mathrm{T}\mathrm{F}\mathrm{D}- \\ {0.000086}^{*}\mathrm{R}\mathrm{E}\mathrm{M}\mathrm{I}\mathrm{T}\mathrm{H}\mathrm{C}\mathrm{E} \end{gathered}$$	REMIT, REMITFD, and REMITHCE are significant

* indicates significance at 5% level based on t-statistics. Source: Authors’ computations.

shows the long-run cointegration equations for Model 1 and Model 2 that determine significant relationships between remittances and income inequality (GINI index), with macroeconomic variables and interaction terms as controls.

In Model 1, all the coefficients are statistically significant, indicating that remittances (Remit), financial development (FD), and household consumption (HCE) are related to declining income inequality in Lebanon. Yet, economic growth (LOGGDP) is linked with rising income inequality. Interestingly, the negative sign on FD (−0.7297) and REMIT (−0.0036) indicates that both enhanced financial access and remittances make the distribution more equal in the long run. The positive sign of LOGGDP (0.0007) indicates that the gains of economic growth in the long run were not evenly distributed among all groups of people. It, therefore, indicates that growth in Lebanon, in the years of study, was not equal since more prosperous regions absorbed more than their due share of growth, leading to greater income disparity. The same conclusion was reached by Adams and Page [17] and Acosta et al. [31] regarding the significant and negative relationship between remittances and income inequality.

In Model 2, multicollinearity was not a problem; VIF values were comfortably within the range of 1.36 to 2.07. The model shows that remittances by themselves reduce inequality, but the effect is reduced when financial development is high, as evidenced by the strong positive interaction between remittances and financial development. That is, financial development is a moderator in the remittances–inequality relationship.

The Lebanese experience can explain this result. The country’s financial and economic system has been in crisis since 2019, but even before that, it only served the political elites and the privileged classes. Poor and rural households lack easy access to banks, credit, or formal remittance services, while wealthier households have financial instruments like foreign currency accounts or investment products. Therefore, when remittances are flowing into such an uneven system, the richer can gain more, which can cause inequality to increase rather than reduce.

Although financial development is often considered a force for equity, Lebanon shows that if access is uneven, it can have the opposite effect. The estimates of our model reflect this fact: in elite-dominated and fragmented systems, remittances lose their potential to bridge income gaps. Our result corroborates the findings of Aggarwal et al. [33] and Beine and Lodigiani [34]. The former established that remittances reduce inequality more intensely when financial development is low, since in highly developed systems, the wealthy capture more, and the latter argues that financial development raises the extent to which remittances are used for investment, but this will work in favor of the non-poor. This is also corroborated by Acosta et al. [31], who point out that remittances are more capable of reducing inequality in countries having poorly developed financial institutions, where they impact poorer households more directly.

By contrast, the spillover effect with consumption in the household (REMITHCE) has a very limited negative impact, which suggests an inequality-reducing impact through increased consumption, but in a very weak context. The preceding relationship suggests that remittances, if channeled into consumption, namely, among poor families, can potentially decrease income inequality through enhanced welfare and poverty reduction. This confirms that consumption is a good but weak moderating channel through which remittances can reduce inequality in Lebanon and hence enable remittances to promote the welfare of poor households in the long run. However, the weak impact could be an indication of what is known as “investment poverty”, where households are too poor to invest in productive activities, and remittances are not used for transformative investment. Poor households are incapable of channeling funds from their daily necessities to other uses such as investments in education, health, and sustainable development projects. The study findings align with those of Taylor et al. [62] and Ratha [15], which illustrate how household spending can be a good medium through which remittances concentrate income inequality, particularly by improving the welfare of poor households.

Table 6. Short-Run Estimates (ECM).

Model	Short-Run Equation	ECT Coefficient (D(GINI))	ECT and Variables Significance
Model 1 (M1-B)	$$\begin{gathered} \Delta \mathrm{G}\mathrm{I}\mathrm{N}\mathrm{I}=-0.001860+0.0028\Delta \mathrm{R}\mathrm{E}\mathrm{M}\mathrm{I}\mathrm{T}(-1)- \\ 0.2122\Delta \mathrm{F}\mathrm{D}(-1)+0.0644\Delta \mathrm{L}\mathrm{O}\mathrm{G}\mathrm{G}\mathrm{D}\mathrm{P}(-1)+ \\ 0.0010\Delta \mathrm{H}\mathrm{C}\mathrm{E}(-1)-{0.7465}^{*}\mathrm{E}\mathrm{C}\mathrm{T}(-1)+ \\ 0.015423\mathrm{D}2006-0.021089\mathrm{D}2019 \end{gathered}$$	−0.7465	ECT Significant at 5% (t = −2.34)
Model 2 (M2-B)	$$\begin{gathered} \Delta \mathrm{G}\mathrm{I}\mathrm{N}\mathrm{I}=-0.002786-0.0012\Delta \mathrm{R}\mathrm{E}\mathrm{M}\mathrm{I}\mathrm{T}(-1)- \\ 0.0145\Delta \mathrm{R}\mathrm{E}\mathrm{M}\mathrm{I}\mathrm{T}\mathrm{F}\mathrm{D}(-1)+ \\ {0.000076}^{*}\Delta \mathrm{R}\mathrm{E}\mathrm{M}\mathrm{I}\mathrm{T}\mathrm{H}\mathrm{C}\mathrm{E}(-1)-{0.2970}^{*}\mathrm{E}\mathrm{C}\mathrm{T}(-1)+ \\ 0.0101\mathrm{D}2006-0.0202\mathrm{D}2019 \end{gathered}$$	−0.2970	REMITHCE is significant and ECT Significant at 5% (t = −2.87).

* indicates significance at 5% level based on t-statistics., Source: Authors’ computations.

presents the short-run dynamics and error correction. The short-run equations reflect the pattern of adjustment of inequality towards the long-run equilibrium as well as the direct effects of the explanatory variables.

In Model 1, the statistically significant error correction term (ECT = −0.7465) (t = −2.34) indicates a strong and instant adjustment towards the long-run equilibrium. This captures the reality that any short-run disequilibrium in income inequality is corrected by about 74.7% each year. All the short-run coefficients are statistically insignificant. This implies that even though these variables have a presence in influencing inequality in the long run, their effects in the short run are weak or lagged. D2006 and D2019 (crisis dummies) are also insignificant in the short-run GINI equation, meaning that responses to these 2019 financial and economic crisis episodes by inequality were likely captured more in the long-run dynamics or other controls. The result of the short-run model is supported by Gupta, Pattillo, and Wagh [38], who point out that the contribution of remittances to poverty and inequality may be modest in the short run but becomes increasingly robust over time as remittances contribute to overall welfare gains.

The adjustment towards the long-run equilibrium is corroborated by the statistically significant ECT of Model 2 (−0.2970; t = −2.87). The model has achieved an approximately 30% correction of last year’s disequilibrium this year, a relatively modest rate of adjustment. The slower rate of adjustment compared to Model 1 may be a reflection of the more complicated nature of the interaction terms in the model. Among the short-run variables, only the change in remittances through household consumption (ΔREMITHCE) is positive and statistically significant at the 5% level (t = 2.44). This suggests that when remittances are used for consumption, they can immediately help reduce inequality. Interestingly, the interaction term REMITHCE shows a dual effect: in the short run, the positive coefficient suggests that increased consumption might be concentrated among wealthier households, which could temporarily widen inequality. But in the long run, the coefficient turns negative and significant, indicating that remittance-driven consumption eventually helps narrow inequality, likely by funding essential needs like health, education, and daily living for poorer households. Meanwhile, the lagged values of remittances and their interaction with financial development (ΔREMIT(−1), ΔREMITFD(−1)) are not significant in the short term, implying that their inequality-reducing impact emerges gradually rather than instantly. Overall, the findings highlight that how remittances are spent, especially on household needs, matters considerably, and their positive effects on reducing inequality take time to unfold. These findings are consistent with theoretical accounts and empirical evidence discussed in the literature, which indicate the lagged and conditional impact of short-term remittances on income inequality, as outlined by the permanent income hypothesis [1], the migration life-cycle theory [2], the adjustment lag theory [3], and the migration selectivity hypothesis [4].

4.5. Pairwise Granger Causality Test

Pairwise Granger causality tests were employed to Model 1 only, to focus on primary relationships without complications from the interaction terms of Model 2.

Table 7. Pairwise Granger causality test for Model 1.

Null Hypothesis	F-Statistic	p-Value	Conclusion
REMIT does not Granger Cause GINI_INDEX	0.0196	0.8899	Not significant
REMIT does not Granger Cause FD	4.4970	0.0466	Significant
REMIT does not Granger Cause LOGGDP	2.3403	0.1417	Not significant
REMIT does not Granger Cause HCE	0.0170	0.8975	Not significant
GINI_INDEX does not Granger Cause REMIT	0.0149	0.9040	Not significant
FD does not Granger Cause REMIT	2.1552	0.1576	Not significant
LOGGDP does not Granger Cause REMIT	7.8042	0.0112	Significant
HCE does not Granger Cause REMIT	4.2809	0.0517	Marginally significant
FD does not Granger Cause GINI_INDEX	2.5741	0.1243	Not significant
FD does not Granger Cause LOGGDP	3.2946	0.0845	Marginally significant
FD does not Granger Cause HCE	0.0917	0.7652	Not significant
GINI_INDEX does not Granger Cause FD	0.0633	0.8039	Not significant
LOGGDP does not Granger Cause GINI_INDEX	1.2330	0.2800	Not significant
LOGGDP does not Granger Cause REMIT	7.8042	0.0112	Significant
LOGGDP does not Granger Cause HCE	0.2229	0.6420	Not significant
GINI_INDEX does not Granger Cause LOGGDP	0.0042	0.9488	Not significant
HCE does not Granger Cause GINI_INDEX	0.0056	0.9413	Not significant
HCE does not Granger Cause REMIT	4.2809	0.0517	Marginally significant
HCE does not Granger Cause FD	0.0006	0.9815	Not significant
HCE does not Granger Cause LOGGDP	0.3377	0.5677	Not significant
LOGGDP does not Granger Cause HCE	0.2229	0.6420	Not significant

Source: Authors’ computations.

investigates the direction of causality between the variables, allowing for the identification of unidirectional or bidirectional causality relationships.

The test outcomes indicate that remittance–income inequality and corresponding macroeconomic variables relationships are primarily unidirectional, and bidirectional causality null hypotheses fail to be rejected at the 5% level. Notably, two causality relationships are strong in the one-way direction—from remittances to financial development and from economic growth (LOGGDP) to remittances—suggesting that remittances have a causative impact towards inducing financial activity, but receipts of remittances are macroeconomically determined. While two other relationships, i.e., HCE to REMIT and FD to LOGGDP, are marginally significant, they fail to attain the 5% criterion; thus, the causality is not strong, and the relationship may be context-dependent or unstable over time. Interestingly, no short-run causality in either direction between remittances and income inequality (GINI) was detected in Model 1, corroborating the absence of short-run forecast ability between the aforementioned variables. This corroborates the usefulness of the long-run analysis cointegration model and the VECM, because the impact of remittances on inequality appears to take effect in the long run via indirect transmission channels like financial development and the consumption patterns of households.

4.6. Diagnostics Tests

In order to validate the reliability and validity of the VECM estimates, several post-estimation diagnostic tests were applied to Model 1 and Model 2, i.e., stability analysis (CUSUM and CUSUM of Squares), heteroskedasticity, serial correlation, and Portmanteau autocorrelation tests, and the result was determined by taking up to one lag into account.

As presented in Table 8, all the diagnostic tests subsequent to estimation show that Model 1 and Model 2 are statistically well behaved. Both the CUSUM and CUSUM of Squares tests, along with the inverse roots of the AR characteristic polynomial, indicate structural stability for the two models. There was no heteroskedasticity, as both the joint and individual tests gave p-values well beyond 0.10. Additionally, the serial correlation LM tests and Portmanteau tests confirm the non-existence of autocorrelation among residuals even at lag 5 for Model 1 and lag 2 for Model 2. In all, the diagnostic results confirm the consistency and reliability of the estimated VECM specifications.

Table 8. Summary of Post-Estimation Diagnostics.

Diagnostic Test	Model 1	Model 2
Stability Test (CUSUM)	Stable (see Figure 6)	Stable (see Figure 7)
Stability Test (CUSUM of Squares)	Stable (see Figure 8a)	Stable (see Figure 9a)
Inverse roots of AR	Stable (see Figure 8b)	Stable (see Figure 9b)
Heteroskedasticity (Joint Chi-square)	χ2(210) = 199.88, p = 0.6805	χ2(120) = 108.93, p = 0.7562
Heteroskedasticity (Individual)	All p > 0.10 → No issue	All p > 0.10 → No issue
Serial Correlation LM Test	χ2(25) = 18.50, p = 0.8206	χ2(16) = 17.39, p = 0.3611
Rao F-stat (LM)	F(25,16.4) = 0.617, p = 0.8657	F(16,22.0) = 1.129, p = 0.3884
Portmanteau (Lag 2)	Q = 30.07, Adj Q = 32.47, p = 0.9187	Q = 19.61, Adj Q = 21.04, p = 0.8239
Portmanteau (Lag 4/5)	No autocorrelation up to lag 5	No autocorrelation up to lag 2

Source: Authors’ Computations.

Table 8. Summary of Post-Estimation Diagnostics.

Diagnostic Test	Model 1	Model 2
Stability Test (CUSUM)	Stable (see Figure 6)	Stable (see Figure 7)
Stability Test (CUSUM of Squares)	Stable (see Figure 8a)	Stable (see Figure 9a)
Inverse roots of AR	Stable (see Figure 8b)	Stable (see Figure 9b)
Heteroskedasticity (Joint Chi-square)	χ2(210) = 199.88, p = 0.6805	χ2(120) = 108.93, p = 0.7562
Heteroskedasticity (Individual)	All p > 0.10 → No issue	All p > 0.10 → No issue
Serial Correlation LM Test	χ2(25) = 18.50, p = 0.8206	χ2(16) = 17.39, p = 0.3611
Rao F-stat (LM)	F(25,16.4) = 0.617, p = 0.8657	F(16,22.0) = 1.129, p = 0.3884
Portmanteau (Lag 2)	Q = 30.07, Adj Q = 32.47, p = 0.9187	Q = 19.61, Adj Q = 21.04, p = 0.8239
Portmanteau (Lag 4/5)	No autocorrelation up to lag 5	No autocorrelation up to lag 2

Source: Authors’ Computations.

Figure 6. Stability test using CUSUM for (M−1).

Figure 6. Stability test using CUSUM for (M−1).

Figure 7. Stability test using CUSUM for (M−2).

Figure 7. Stability test using CUSUM for (M−2).

Figure 8. (a): Stability test using CUSUM (squares) for M−1. (b): Inverse roots of AR for M−1.

Figure 8. (a): Stability test using CUSUM (squares) for M−1. (b): Inverse roots of AR for M−1.

Figure 9. (a): Stability test using CUSUM (squares) for M−2. (b): Inverse roots of AR for M−2.

Figure 9. (a): Stability test using CUSUM (squares) for M−2. (b): Inverse roots of AR for M−2.

4.7. Forecast Error Variance Decomposition (FEVD) and Impulse Response Functions (IRFs)

To have a better understanding of the dynamic impact of remittances and other corresponding controls on income inequality, the present study employs Forecast Error Variance Decomposition (FEVD) and impulse response functions (IRFs) based on the estimated VECM for Model 1. FEVD does decompose, but error forecast variance in the GINI index can be explained by the share attributed to shocks in each explanatory variable. IRFs are employed in simulating the time-path of income inequality (GINI) concerning a one-period shock in remittances and other endogenous controls, such that insight is obtained into the direction, magnitude, and persistence of effects over time.

For any potential structural breaks due to the 2019 financial and economic crisis in Lebanon, the model is initially estimated on the full sample period (2000–2023) and subsequently split into two sub-periods: pre-2019 and post-2019. This enables comparative assessment of the dynamics of remittances and inequality under different regimes of economic evolution and stability. Our findings align with those of Calderón and Page [48], who reached nearly the same conclusion.

4.7.1. Variance Decomposition Analysis

Table 9. Comparative Summary Table (REMIT → GINI).

Period	Pre-2019	Full Sample	Post-2019
1	0.00%	0.00%	0.00%
5	30.95%	8.30%	4.58%
10	38.8%	19.68%	-

Source: Authors’ calculations based on FEVD results from VAR model estimations.

and Figure 10 show a comparative summary table for the Forecast Error Variance Decomposition (FEVD) results for the three samples based on Model 1. The FEVD outcome reveals that the explanatory power of remittances on income inequality (GINI index) is rather heterogeneous across horizons. Over the whole sample period (2000–2023), remittances account for a significant but increasing share of GINI variation (19.68% by period 10), while context and macroeconomic variables (LOGGDP and FD) account for more. In the pre-2019 sub-sample, remittances are, however, a top contributor, explaining up to 38.8% of inequality variation and pointing to their central redistributive role in a fairly stable economic environment. After 2019, their ability to explain falls precipitously to a modest 4.6%, because inequality is progressively self-reinforcing, implying that economic crises, devaluations, or remittance consumption changes may have cut their long-term impact. The results confirm the split-sample approach and point to the importance of context to specify to what extent remittances drive income inequality in the long term. The comparison amply brings out the diminished role of remittances in inequality for Lebanon’s recent crisis. Our results are consistent with those of Ebeke and Hakura [47], who reported similar evidence. For more detalils on each sub-sample of FEVD individually, please refer to the Appendix A.

4.7.2. Impulse Response Function (IRF) Analysis

To complement the FEVD results and better visualize the dynamic response of income inequality to remittance shocks, impulse response functions (IRFs) were generated for Model 1 across three samples: the full period (2000–2023), pre-2019, and post-2019. The IRF plots (Figure 11 and Figure 12a–c) show comparisons for the three samples and depict how a one-standard-deviation shock to remittances affects the GINI index over a 10-period horizon. In the full sample, the IRF reveals a modest and delayed increase in income inequality, indicating a slight inequality-worsening effect of remittances. The direction of the response is positive, suggesting that remittances might increase inequality, but the impact is weak and statistically insignificant in the full sample. This finding aligns with the FEVD results, which show that remittances explain only 19.7% of the forecast error variance in income inequality by period 10. This is a moderate rate because the whole sample period (2000–2023) captures the effect of the 2019 economic crisis, which suppressed the dynamic behavior of remittances compared to the pre-2019 period.

This effect is more pronounced in the pre-2019 period, where the response is stronger and more persistent, supporting the view that remittances had greater redistributive power before Lebanon’s economic crisis. This complements the FEVD result, where remittances explained up to 38.8% of the variance in inequality pre-2019. In contrast, the post-2019 period shows a flat and statistically insignificant IRF, suggesting that the impact of remittances on inequality almost disappeared after the crisis. This also supports the FEVD finding that GINI is almost entirely self-driven in this period (remittances explained only ~4.6% of its variance). This divergence across sub-samples highlights the importance of macroeconomic context and supports the split-sample approach to better capture structural breaks and shifts in remittance effectiveness. The IRF plot and the comparison table suggest that remittance shocks have no significant short-term effects on income inequality across all horizons, with all the responses being statistically insignificant. This implies that the dynamic effect of remittances on inequality is negligible or delayed rather than contemporaneous.

4.8. Hypothesis Testing

Based on the empirical findings, H1 is partially supported. The results indicate that remittances significantly impact income inequality in the long run, which supports the long-run role of remittances in explaining income distribution in Lebanon. However, the short-run impact is not statistically significant, which means that remittances have no short-run impact on levels of inequality. For Model 2, in H2, we reject the null hypothesis for both the long run and the short run. REM × FD and REM × HCE are significant in the long run, but only REMITHCE is significant in the short run. Finally, in H3 and H4, the null hypotheses are rejected because impulse response functions show that remittance shocks reduce inequality over time, and variance decomposition confirms that remittances explain a significant proportion of inequality variation, especially before 2019.

5. Conclusions and Policy Recommendations

In this study, the authors tested the impact of remittances on income inequality in Lebanon from 2000 to 2023. The outcome overall strongly demonstrate that remittances help reduce inequality in the long run, but are not significant short-run forces; remittances have an insignificant or negligible impact, particularly amid Lebanon’s prolonged economic downturn. This suggests that remittances perform as a gradual equalizer towards long-run equilibrium free from very strong short-term shocks. In other words, remittances are not an immediate fix. This conclusion corroborates the findings of the FEVD and IRF analyses, and both tests validate the same result.

The FEVD reflects that remittances explained most of the GINI variance before 2019 (up to ~39% during period 10), yet merely ~19.7% for the overall sample and ~4.6% post-2019. The IRF analysis also validates that in all samples, remittance shocks have no statistically significant dynamic effect on inequality, as the responses are close to zero and inside confidence bands. The findings show that economic growth in Lebanon has been non-inclusive, and the income gap has been widening. The financial sector and household consumption play a crucial role in shaping the remittance-inequality relationship. A sizable financial sector lacking financial inclusion can reverse the inequality-reducing effect of remittances, whereas directing remittances towards productive purposes enhances their long-run impact. Remittances in times of economic crises are spent mainly on basic needs (as shown by the significant REMITHCE effect), limiting their role in reducing inequality. Short-run consumption by itself does not affect inequality unless supported by stable remittance inflows.

In light of these observations, a set of targeted policy recommendations is proposed to enhance the effectiveness of remittances in reducing income inequality, particularly within Lebanon’s fragile financial system and weak economic environment. These recommendations are as follows:

• The Lebanese government and the central bank need to concentrate on financial inclusion rather than just expanding the banking sector. This includes encouraging microfinance, mobile banking, digital wallets, and no-minimum-balance accounts to target low-income households with financial services. Given the challenges in the Lebanese commercial banking sector, policymakers should, for example, encourage fintech programs like Purpl, which are working to make remittances accessible for everyone in Lebanon by offering digital wallets and also connecting users with services through which they can receive and withdraw money.
• Strengthen social safety nets and broad-based growth by improving low-income households’ access to health, education, and small business financing for productive remittance use and long-term inequality reduction.
• Foster financial inclusion through the provision of financial literacy to remittance-receiving households, especially from high-migration rural areas, to enable saving, investment, and long-term financial planning.
• Expand donor-supported programs like Anera’s vocational training and LIM’s microfinance tools to enhance skills, provide financing to small businesses, and channel remittances to productive income generation.

Remittances are important for household survival and, with proper management, can promote more equity in Lebanon’s fragile economy. Policymakers are able to enhance the contribution of remittances to reducing income inequality through specific policies, and they can also promote inclusive, sustainable economic growth. The research had some limitations; most notably, data availability was a challenge, as some of the variables were not available for all years, which limited the authors’ ability to extend the time frame and consequently resulted in a small sample size. Future research can explore other potential moderating factors, such as governance indicators or labor market institutions, and utilize other econometric techniques, like nonlinear or structural models, to better reflect the complex dynamics between remittances and inequality.

Table 10. Summary of Methods and Findings: Remittances and Income Inequality.

Method	Sample	What It Measures	Result
VECM Long-Run Equation	Full sample (2000–2023)	Long-run cointegrating relationship	REMIT reduces GINI significantly
VECM Short-Run Equation	Full sample	Immediate impact	REMIT effect is insignificant
IRF (Full Sample)	2000–2023	GINI’s dynamic response to REMIT shock	Slight positive effect in the medium term
IRF (Pre-2019)	2000–2018	GINI’s dynamic response to REMIT shock	Initial decline followed by an increase in GINI
IRF (Post-2019)	2019–2023	GINI’s dynamic response to REMIT shock	Flat and statistically insignificant
FEVD (Full Sample)	2000–2023	% of GINI variance explained by REMIT	Moderate (up to ~19.7%)
FEVD (Pre-2019)	2000–2018	% of GINI variance explained by REMIT	High (up to ~38.8%)
FEVD (Post-2019)	2019–2023	% of GINI variance explained by REMIT	Low (up to ~4.6%)

Source: Authors’ own calculations based on VECM, IRF, and FEVD estimations in EViews.

provides an overview of the empirical approach employed and the most significant results regarding the remittances–income inequality relationship. It is a concise comparison between the different empirical approaches employed, the time span involved, the specific measure of the particular aspect in question, and the corresponding results.