T H E L O NG S HA D O W O F I N F O R MA L I T Y
C H A P T ER 4
169
ANNEX 4B Regression analysis Correlation between informality, poverty, and income inequality. The following crosscountry ordinary least squares regression model is estimated to show the association between informality and levels of extreme poverty and income inequality:
y i = α 0 + θ1 xi + θ 2 LnGDPpc i + ∈i
(4B.1)
The results are reported in table 4D.6. The dependent variable ( yi ) includes a range of measures for levels of poverty and income inequality averaged over 1990-2018 in country i. The level of poverty is proxied by the poverty headcount ratio at $1.90 a day (2011 purchasing power parity [PPP]) in percent of the total population. Measures for income inequality include the Gini coefficient (range from 0 to 100, with 0 being perfect equality and 100 being extreme inequality), survey mean consumption or income per capita of the lowest-income 40 percent of population, and the difference in consumption or income per capita levels between the bottom 40 percent of population and the total population (World Bank 2018). Last, the progress in shared prosperity, measured as the difference in the average annual growth in income or consumption of the poorest 40 percent of population and that of total population, is used as the dependent variable in column (6). The variable of interest, xi , is the average level of informality in country i over the period 1990-2018, including the share of estimates based on DGE and multiple indicators multiple causes (MIMIC) models of informal output in official GDP and the share of self-employed in employed. All regressions control for income per capita, measured as the logged real GDP per capita in 2010 U.S. dollars averaged between 1990 and 2018 ( LnGDPpci ). The proxies for poverty, income inequality, and shared prosperity are taken from World Development Indicators (WDI). Declines in informality, poverty reduction, and income equalization. The association between within-country changes in informality and poverty and inequality reduction is explored using a similar sample setup and methodology as in Dollar and Kraay (2002) and Dollar, Kleineberg, and Kraay (2013). In particular, the sample of country-year observations is assembled by starting with the first available observation for each country and selecting all available consecutive observations with at least a five-year distance between them (sampling window). This approach yields 428 country-year pairs for 32 advanced economies 119 EMDEs with at least two observations per country and a median of four observations per country. A median distance between observations is 5.5 years for EMDEs and 5.0 years for advanced economies. The sample excludes fragile and conflict-affected states. In table 4D.7, the dependent variables are changes in poverty rates at $1.90 and $3.20 per day (in PPP terms) poverty lines at the end of the sampling window. In table 4D.8, the dependent variables are changes in the Gini coefficient and shared prosperity at the end of the sampling window, where shared prosperity refers to the income share of the bottom 40 percent of the population.
