Time-varying persistence
The study of historical persistence occupies a prominent place in the HPE literature. Generally, persistence studies examine the effect of a past societal feature – an institution, a type of endowment, a cultural norm – or historical event – such as past conflict or migration – on outcomes today. Several posts of Broadstreet editors have touched upon questions of persistence (see, for instance, Volha on the legacies of the Holocaust and forced migration in Europe, Pavi on the legacy of caste inequality in India, or Sara on the effects of historical experiences on trust in medicine). The topic has also sparked a number of debates ranging from substantive questions on the relationships tested (see Jared’s summary of the Bazzi, Fiszbein and Gebresilasse study on the persistent cultural effects of the frontier in the US) to methodological disagreements (see, for instance, Morgan Kelly’s critique of spatial correlation in persistence studies).
There is no doubt that history matters for the present. The challenge of persistence studies is to figure out why and how it matters. To that end, one important task is to uncover which aspects of history affect which facets of present economic, political and social structure and behavior. As hard as this may be – a past event may affect literally everything about a society, and every past event constitutes an entire bundle of treatments – some persistence studies achieve the goal leading to important findings. For example, we now know that the slave trade had a profound negative impact on Africa’s development, not only through its direct toll on human lives, but also through an indirect – though precisely identified – pathway: by generating norms of distrust that can undermine economy and society in the long-run.
An important dimension that helps us understand how history matters is timing. The typical setup of a persistence paper examines the effect of a variable X at time t on a variable Y (or on X itself) at time t+z, where z can be anything from a few years to millennia. The choice of z matters a great deal for a study’s findings. In other words, whether the past has long-run effects, as well as what those effects are, depends on the time at which we are looking for their presence. History may play a big role in certain circumstances but not others, and the nature of its influence depends on a host of underlying factors, many of which may not be on the researcher’s radar. Since many persistence studies tend to be present-focused when it comes to outcomes, this dimension is often overlooked. Historical experience may matter in different ways today than it has in the past, and what may be irrelevant today can become relevant in the future.
This is not a novel observation. Acemoglu, Johnson and Robinson’s study on the reversal of fortunes is an early example of time-varying persistence. They showed that former European colonies with the largest population densities in 1500 had the lowest per capita incomes in the late 20th century. Building on the arguments of Engerman and Sokoloff, AJR attributed this reversal to the institutional history of the colonies. Places rich in natural endowments encouraged extraction; poorer places instead invited the development of institutions that protected private property and promoted investment. Importantly, those “good institutions” only became relevant for development in the 18th and 19th centuries, when they interacted with other fundamental changes (the industrial revolution and technological innovation) to determine country incomes. Correlating proxies of income among former European colonies between 1500 and, say, 1650, might have demonstrated remarkable positive persistence – picking 1995 as the end year leads to exactly opposite conclusions.
More recently, several studies examining shorter time horizons and higher frequencies have been able to measure the time-varying effects of the past on the present in a more targeted way. Rozenas and Zhukov examined the political legacy of repression aiming to answer the following theoretical question: does past repression generate compliance or does it incite opposition? They studied the effects of Stalin’s forced requisitions of grain from Ukrainian peasants that led to a deadly famine. Their answer is that the long-term effects of repression on political behavior depend on when that behavior is measured. In the early stages of WWII and around the time of Soviet collapse communities exposed to past famine displayed the highest opposition to Moscow. Between 1946-1958 and then again in present times, those same communities were less anti-Russian. In this case, time matters through changes in political conditions. Repression generates disloyalty when the repressive power is weak. It has the opposite effect when that power can credibly threaten to repeat past repressive acts.
In this paper with Hans-Joachim Voth we also find that timing matters. Our interest is in the long-run effects of past violence on attitudes and behavior towards the perpetrator of the violence. During WWII, the German army committed a number of atrocities in the Balkans – a precursor to even larger scale violence to follow on the Eastern front. Greece was one of the countries to suffer from civilian reprisals in response to guerilla activity. Villages were burnt to the ground and their population – including women and children – was exterminated. The memory of those events turned out to matter for high-stakes economic behavior many decades later. During the Greek debt crisis, when Germany was once again viewed as an enemy by large part of the Greek population due to its insistence on strict austerity measures, German car sales disproportionately dropped in areas that had suffered WWII massacres. Attitudes of residents towards Germany became more negative as well.
Interestingly, no indication of anti-German sentiment could be traced in these areas in earlier times. By comparing car sales across massacred and non-massacred prefectures in pre-crisis times one would have not concluded that animosity against Germans lingered on. Our research design explicitly demonstrates that time-varying political tension between Greece and Germany (as captured in the press) activated and magnified anti-German sentiment in areas with memories of past violence. Here, the explanation for time-varying persistence is more psychological than structural in nature, but the basic premise is the same: the way history’s shadow manifests depends on timing and context.
Similar conclusions are reached by other studies. Ochsner and Roesel show that the pillaging of Austrian towns by the Ottomans during the sieges of Vienna in 1529 and 1683 was exploited by far-right politicians in recent years to instigate anti-Muslim sentiment and attract votes. They show that the history of conflict did not have a persistent effect on anti-Muslim attitudes until political entrepreneurs capitalized on it to promote their platforms. Belmonte and Rochlitz study the interaction between memories of the Soviet collapse and regime transition in Russia and state-led “recollection campaigns”. Regions that suffered more during the transition cast more votes for the government after, but not before, negative memories were made salient in state-controlled media. In both these cases, the time-varying persistence of past events is engineered by political elites.
In a somewhat different setup, Cantoni, Hagemeister and Westcott uncover another instance of time-varying persistence of political behavior. German areas that supported the Nazi party in 1933 voted more for the AfD – but only after 2017, when the AfD moved from an economically conservative to an anti-immigrant and nationalist platform. Even if far-right ideology was more prevalent in those areas prior to the AfD’s ideological repositioning, it is not possible to identify this with available observable metrics; these areas did not vote more for far-right parties that existed in the years before 2017, likely because of those parties’ links to neo-Nazi activity and the associated social stigma. Once an appropriate, “respectable” party platform was supplied, latent demand arose to meet it. One may have drawn very different conclusions from correlating 1933 NSDAP votes to 2016 voting patterns, than from examining the same correlations a few years later.
Collectively then, these papers suggest that in the study of persistence the choice of timing matters a great deal. In some cases, understanding whether something has an effect in the present may be an end in itself. For more generalizable conclusions, however, one needs to broaden the time scope and examine persistence at different points in time. Time-varying effects provide crucial insight on the latent variables that interact with past events to influence the present, helping us shed light on the mechanisms of persistence.