EBP2: Is Democracy Good for the Poor? - Michael Ross

Previous studies report that democracies outperform non-democracies in regards to the improvement of the welfare of the poor - these claims are consistent with political economy models. This is because democracies fund public services at a higher level, however, this does not mean that this funding reaches the poor, or if better outcomes are actually produced. Michael Ross argues that prior studies are flawed by a selection bias, as they tend to not include non-democratic states that have good economic and social records - which creates false impressions that democracies outperform non-democracies when it comes to the improvement of the welfare of the poor. Theories of Regime Type and Poverty: There are three common theories that suggest that a democracy will raise the living standards amongst the lower income quintiles. The first theory suggests that democracies allow for the poor to penalize the governments, whom allow famines to occur, via the electoral process. To try and avoid famines from occuring, political leaders are thus required to act strategically. Sen, who is associated with the first two theories argues that: “Famines...don’t kill the ruler...if there are no elections, no opposition parties, no scope for uncensored public criticism, then those in authority don’t have to suffer the political consequences of their failure to prevent famines. Democracy, on the other hand, would spread the penalty of famines to the ruling groups and political leaders as well.” (Sen 1999, 180) The second theory by Sen suggests that democracies are better than non-democracies at relaying information from poor areas to the government because of the freedom of press. “...elementary source of basic information about a threatening famine are enterprising news media...for bringing out facts that may be embarrassing to the government (facts that an authoritarian government would tend to censor out).” (Sen 1999, 181) Even if democracies and non-democracies are just as determined as each other, democracies are more likely to know when to take action. The third theory suggests that through producing more public goods, and an increase in income distribution, democracies are more successful than non-democracies. Some scholars further go on to suggest that as a result of the electoral process, democracies spend more on public goods as opposed to autocratic governments, who have no constraints. It is said that this is because democratic governments have a wider range of supporters to please, thus leads to an increase production of public goods instead of private goods. Meltzer and Richard (1981) developed a model on the distributional effects of democracy, from this they have concluded that democratization happens when political rights are extended from the wealthy downwards, to the rest of the citizens. This expansion results in the position of the median voter to move downwards in income distribution. In democracies, people with lower incomes tend to have higher turnouts to the polls, which forces the income redistribution to occur. However, even if democracies increase their social funding, it does not guarantee for an improvement in the infant and child mortality rates. This is is suggested to be an accurate indicator of how the social welfare is amongst the poor. Empirical Studies: Academics, such as Moon and Dixon, and Dasgupta, find that democracies generally produce better welfare outcomes, and an improvement in living standards. Other academics conclude that democracies should decrease the infant and child mortality rates. Furthermore, Siege, Weinstein, and Halperin reported that “poor democracies also suffer 20% fewer infant deaths than poor autocracies”, their statement indicates that democracies do outperform non-democracies in a wide variety of social indicators. However, there are three important problems with past quantitative studies of democracy and the poor. Firstly, country specific effects are not accounted for, even if other studies of infant and child mortality rates have found country specific effects. As nation states are heterogenous, this is not surprising. In states, the welfare levels are influenced by factors such as; geography, culture, colonial legacy etc. When a state is influenced by any of these factors, and it affects the state’s regime type and poverty level, the absence from the model will lead to a bias. Secondly, neglect of exogenous global health trends. Due to low cost health interventions. The global infant and child mortality rates dropped drastically between the period of the 1970s and 2000s. This however, could have been wrongly accounted for in research due to other variables that were trending during that period, such as the global increase of democracies. Thirdly, sample bias. ‘Global studies’ are based on small states that produce easily obtainable information. For example, Dasgupta only gathered data from 51 states, and Przeworski et al. on 135 states - the missing data will lead to a reduction in the actual sample size, as Przeworski et al. only used about one third of their gathered information in the final data set. It has been found that data sets from non-democratic states are more likely to be missing. Results would not turn bias if the authoritarian states that have lower observations had similar results to authoritarian states with higher observations. However, the poor in lower observation states were usually better off - where there was a higher number of improvements as opposed to the poor in higher observation states. With the exclusion of low observation states from data sets, academics are able to produce results that show that authoritarianism is worse for the people as opposed to democracy. Data, Mode, and Variables: There are two widely used models to estimate the effect regime type has on infant and child mortality rates, these are the fixed effects model, and a model created by Beck and Katz - the OLS with ‘panel-corrected standard errors’, that has a lagged dependent variable. When running the tests there are two dependent and two independent variables. These are the log of the infant mortality rate and the log of the child mortality rate, and the two independent variables is two alternative measures of regime type. Recent studies have shown, when tested with these variables, that new democracies compared to established democracies perform worse. The model further uses four control variables, including dummy variables for each period. These variables are; the log of income per capita, the log of population density, economic growth, and the impact of HIV/AIDS on the poor. Results show that the measures of democracy fail to reach significance when based on the original dataset (table 3) with the inclusion of fixed effects and period dummies, even in the filled-in datasets when these are not included. Tests run with table 4 produces similar results. It has been found that if an authoritarian state were to change to a democratic state, it would not have a large effect on child mortality rates, these rates would only drop by about 0.2 death per thousand per year. The magnitude the change from complete authoritarianism to democracy would have to infant and child mortality rates is much less than scholars have previously claimed. These tests have shown that the gap between these two governmental systems is close to zero.

Theories of Democracy and Redistribution Revisited: It is expected that in democratic countries the infant and child mortality rates are low, especially in the states that spend a large amount of their funding on social services, in particular health care. This, however, can only happen if the governments of these states focus on the low-income households. Instead, it has been found that these mortality rates are not affected by the government subsidies if they are allocated to include the middle- and upper- income households. These households still have the ability to purchase the required goods with or without the subsidies. The mortality rates are only able to be reduced if these goods are received by low-income households. Basically, the government’s impact on these mortality rates largely relies on the assistance they give to these low-income households. The Meltzer-Richards model determines that a government’s policies targets the median voter - these voters generally lie in the middle-quintile of voters. Furthermore, the additional public goods that are produced by poor democracies do not benefit the poor, instead are more beneficial for the middle- and upper-quintile. Filmer and Pritchett further confirm that public spending has little to no impact on infant and child mortality rates unless the benefits are delivered to low-income households. Further research suggests that the poor in democratic countries do not receive more benefits and subsidies than the poor in non-democratic countries. Conclusion: Social scientists have little knowledge on the types of governments that actually benefit the poor, and improves their welfare, furthermore, the understanding of a government’s role in countries where poverty is severe is even less. During the period between the 1970s and 2000s the mortality rates improved drastically, these rates dropped in 163 countries while it rose in just six countries. Observers during this time period agreed that factors such as; rising incomes, and the increase in dispersion of low-cost health interventions (e.g immunizations and antibiotics) played a role in the improvement of mortality rates. During this period, however, there was a rise in global democracy, but there is evidence yet to be found as to whether this had anything to do with the fall in mortality rates. The article concludes that democracies have over all produced very little improvement in the area of material well being of the people categorized in the lower quintiles. This could be because of the incomplete information provided by voters, social polarization or the inability of politicians to keep to election promises. The article highlights the type of selection bias used when constructing the data sets - countries have a tendency to produce little data the less democratic and unconstrained they are. By saying this, it suggests that authoritarian states that have good records will most likely be under counted, which leads to an underestimation of the performance of authoritarian regimes, and thus overestimating the performance of democratic regimes. If this were to be corrected, the beliefs that are held about democracies could be altered significantly.

(It will be interesting to do some research into what sorts of response this article has garnered since it has been published, and whether there are other academics that agree with what Ross has argued.)