Why do democracies survive? – 02/26/2023 – Marcus Melo

The title of this column will not be on the cover of some bestseller. Yes, the discussion around how democracies die has given rise to a growing body of qualitative and quantitative studies on why predictions about the collapse of democracies have failed.

These studies point to a critical weakness of the analyzes already pointed out here in the column at least four years ago in relation to the best-seller by Levitsky and Ziblatt: there is selection bias in the dependent variable (in this case, the death of democracy). The sample would have to include cases of the survival and death of democracy. Generalizations from case studies of Orban, Chávez, Erdogan, etc —often combined with references to Hitler— have feet of clay. Furthermore, marginal changes in India or the US cannot be reported population-weighted, as if reflecting global trends.

However, some analyzes do not make this primary error and are based on indexes prepared by institutions such as Freedom House, V-DEM, The Economist (EIU), or Polity2. Some are based on objective indicators (eg percentage of elections in which the incumbent or his party won) or number of journalists murdered. Others are subjective (eg, “how do you assess whether the elections were clean and fair and whether the popular will prevailed” and are subject to bias (the number of experts consulted is very small).

Little and Meng disaggregated dozens of indicators from available sources, concluding that over the last four decades there is no evidence of the decline of democracy according to dozens of objective indicators; in many cases the opposite is observed. They also examined subjective indicators and investigated possible reasons for the biases found. They concluded (using a formal model) that the increasing salience and level of information about democratic setbacks explains part of the bias. The impact of the Trump phenomenon is an example.

Some fonts have specific problems. The V-DEM —the gold standard— anonymously publishes the responses of experts (whose number ranges from 5 to 11) who suggest a confidence interval for their answers. Little and Meng showed that it is greater than the standard deviation of responses for the levels of the scales used, compromising some indicators. Polity2, on the other hand, lost a lot of credibility due to its lack of plausibility. After Trump’s election in 2016, the score for the US declined to 5 (the same as Haiti and Somalia), the lowest in its history since 1801, lower even than the Jim Crow Era, when blacks and women (until 1920) were barred from voting.

PRESENT LINK: Did you like this text? Subscriber can release five free hits of any link per day. Just click the blue F below.