Page 89 - Azerbaijan State University of Economics

Fadai Mardanli Mehman, Vildan Zahidkizi Rizayeva: Do Remittances Compensate for the
Labor Market Gaps Created by Emigration?

Regression analysis
To control for time trends and unobserved country-level factors‚ we estimate linear
regression models of the unemployment rate‚ with remittances (as a percentage of
GDP) as the independent variable of interest. We estimated individual country models
and a pooled model with all the data. All analyses were conducted using SPSS version
26. The data are panel data‚ and while a pooled OLS regression may violate some
independence assumptions‚ we use it for our exploratory approach and to control for
the unobserved differences across countries by including country fixed effects. In
concrete terms‚ the following models are run:

Separate country regressions:
UnemploymentRate = α + β × Remittances + ε
Where UnemploymentRate is the unemployment rate‚ Remittances is the share of
personal remittances in GDP‚ α is the constant of the regression‚ β is the slope
coefficient showing how remittances affect the unemployment rate‚ and ε is an error
term. The functional form is estimated separately for four countries (Kyrgyzstan‚
Moldova‚ Nepal‚ Tajikistan) because remittances affect unemployment differently in
each country.

Pooled regression with fixed effects:
UnemploymentRate = α + β × Remittances + γ₁ × D_MDA + γ₂ × D_NPL + γ₃ ×
D_TJK + u.

UnemploymentRate is the dependent variable‚ while Remittances is the ratio of
remittances to GDP. D_MDA‚ D_NPL‚ and D_TJK are dummy variables for
Moldova‚ Nepal and Tajikistan‚ respectively. Kyrgyzstan serves as the base category.
This means no dummy variable is needed for the Kyrgyzstan. The resulting equation
is α + βRemittances + γ₁ + γ₂ + γ₃ + u‚ where α is the intercept for Kyrgyzstan‚ β is
the slope on Remittances for all countries‚ γ₁‚ γ₂‚ and γ₃ are the fixed effects‚ and u
is the error term. This only specifies the model appropriately by controlling for
country intercepts so that differences in baseline unemployment do not bias our
estimate of β‚ which measures the correlation between remittances and unemployment
conditional on country fixed effects (i.e. the average unemployment of a given
country). Alternatively‚ it tells us the correlation between the level of remittances and
the level of unemployment in a country in a given year (the specification does not
include a time trend or year fixed effects). This is because it is a cross-country
regression‚ and the sample of four countries does not allow many fixed effects.
However‚ we treat β as only a partial correlation indicator‚ and do not imply that the
estimate is causal.

89