An econometrical analysis of the interdependencies between the demographic transition and democracy

This paper investigates the effect of an exogenous demographic transition on democracy. As possible channels through which this influence makes an impact, education and per capita income and, to a lesser degree, urbanization were identified. These interdependencies were tested using pooled ordinary least squares as well as fixed effects models on the basis of panel data. In conclusion, the demographic transition affects democracy through the aforementioned channels primarily in poor countries that have experienced their demographic transition and subsequent democratization in the second half of the 20th century or later.

Échantillon de lecture
Chapter 3, Data and descriptive statistics:
3.1. Data:
In order to support these hypotheses the following section provides an empirical analysis based on different linear panel data models for more than 150 countries over the past 200 years. Panal data are advantageous compared to cross-sectional or time-series data because they usually give the resarcher a large number of data points, increasing the degrees of freedom and reducing the collinearity among explanatory variables-hence improving the efficiency of econometric estimates (Hsiao, 2003, p.3). Moreover, panel data are particularly useful to analyze the dynamics of change regarding demography and democracy across different countries. There are, however, also limitations of using panel data, such as sample selection bias because data availability might not be equal across countries and years. Demographic data, such as life expectancy and fertility rates, were taken from the World Bank s world development indicator database (WDI) and are available since 1960. This database also contains statistics on education, urbanization, female labor force participation and GDP. GDP and other economic variables were also drawn from Acemoglu & Johnson (2007) and Banks cross national time series database27. As another measure of education, the Barro-Lee dataset of educational attainment in the world is used, which provides data about average years of schooling in 5-year intervals from 1950 until 2010. As measure for democratization, Polity IV data28 as the most widely used data resource for studying regime change and the effects of regime authority 29 are employed, which cover the time span from 1800 to 2010 for more than 150 countries.
3.1.1. Demographic data:
The most important demographic data are mortality and fertility rates in order to estimate the date of the demographic transition. As one measure, a binary variable labeled demographic transition ( demotran ) was created, which turns 1 if life expectancy is higher than fifty and crude birthrates fall below 30/1000. If one of these criterions is not met the variable stays 0. Life expectancy as taken from the World Bank dataset is defined as number of years a newborn infant would live if prevailing patterns of mortality at the time of its birth were to stay the same throughout its life 30 and crude birth reflect the number of live births occurring during the year, per 1,000 population and are estimated at midyear. As the World Bank data are only available since 1960, all analyses including the binary variable demotran only take into account countries that experienced their transition after 1960.
Data about the onset of fertility decline from Reher s analysis (2004) 31 are taken as another measure for the demographic transition. The difference between these two measures amounts to 15 years on average with a standard deviation of 10.3, which is due to the fact that Reher s data underlies a different definition for the demographic transition than the one applied for the World Bank data. Also, the data of the World Bank only start in 1960 while Reher s data go back to 1900. Thus, only data for countries that experienced their demographic transition after 1960 can be compared, which are mainly African, Asian and some Latin American countries (whose data are less accurate). Taking this into consideration, the two measures from completely different datasets are quite close in terms of the time horizon they reflect.
In addition, the time difference between the present and the time of the demographic transition (yearssincedt) 32 were calculated on the basis of Reher s data. For summary statistics of the demographic data see appendix C. In the following course of the analysis WDI data underlie the binary variable demotran and Reher s data are the basis for the continuous variable years passed since (pending to) the demographic transition . It is important to keep in mind that the variables reflect different time horizons33 and