Most positivist political scientists look for causal relationships inspired by Hume’s diction that we can only ever infer that A causes B to the extent that we observe A and B occurring together at the same time. So, as much as common parlance usually cautions against inferring causation from correlation, it’s actually a pretty important criteria in assessing causal relationships. Of course the difference between scientists and common parlance is that, when faced with a correlation, most scientists spend a lot of time clarifying that correlation before declaring it causal, i.e. they ensure that A in fact occurs before B temporally speaking and they try to determine whether there are any unobserved, third variables that make the correlation spurious. So, when scientists say a correlation is causal, it’s usually worth paying attention, much more so than when your astrologer notes that your birth was correlated with the ascendancy of Jupiter, or whatever.
I raise this because political scientist Salomon Orellana has published a new book on the relationship between parties, party systems and governance. One of his findings, outlined in a blog post at the Monkey Cage, is that there is a relationship between the number of parties in a party system and the incarceration rate in a country. In short, he finds that two-party systems tend to adopt more punitive, rather than rehabilitative, corrections policies. Continue reading
One of his graphs is here:
Clearly, there is a negative correlation there between the fractionalization (a measure of the number of parties in a party system) and the number of prisoners a country incarcerates. I have no reason to doubt Salomon’s evidence, modelling or reasoning about the possibility of unobserved control variables, so I’m fine to see that as a causal relationship.
But look at the United States! Orellana correctly identifies it as an extreme outlier, writing that: “although the American two-party system certainly does not explain everything about the U.S. incarceration rate, the country would nevertheless benefit from the presence of a “consistently heard” dissenter that can help break the vicious cycle of pandering.”
Later, Orellana concludes with this comment, calling for a reform to its electoral system.
Regardless of how reform might be implemented, the important point remains: the high rate of incarceration in the United States has roots in its electoral system. More political parties could ultimately mean fewer people behind bars.
I won’t dispute the second part of that statement – giving more access for third parties could reduce the incarceration rate – but I do dispute the first part of that sentence, that the high rate of incarceration has its roots in its electoral system. Here, Orellana is practicing good positivist social science, and it’s a great example of how good positivist social science can walk right into big problems. Namely, he is privileging the causal correlation that he has identified as the primary causeof his outcome of interest. One of the most important features of social life that complicates the positivist tendency to think about causation in correlational terms is that most outcomes are not determined by one single thing, but they can be determined by many different things.
A different way of assessing causal relations in empirical (note that I did not say positivist) social science is to look not for correlations but for the structures and mechanisms that underlie and create observed regularities. This is premised on a realist view of social science which has deep roots, but you can get a flavour of it by reading Daniel Little’s blog Understanding Society or by reading books like George and Bennett.
The thing that these scholars continually drive home is that correlations are usually only the starting point for empirical social science research. It’s more important to look under the hood of regularities and directly examine the structures and mechanisms (realists have those two words on auto-complete in their word processors) that create them. Doing this type of research is more loyal to the subject matter because human society is an open system where conditions at time 1 do not determine conditions at time 2. By contrast, Hume’s vision of causality, and the one that inspires most positivist social science, was developed very much with closed systems, such as the solar system, in mind.
Realist-inspired empirical research on this specific question would look at Orenella’s identified correlation, accept it as causal, and then looked at that glaring outlier and wondered about what structures and mechanisms caused such a deviation from the observed relationship.
I had the good fortune of taking a really amazing course on the contemporary political economy of the United States with Prof. Phil Wood at Queen’s University and his interest in the politics of prisons led me to some fascinating literature seeking to do what Orenella is *not* trying to do, namely, to explain the reason for the United States’ truly exceptional incarceration rate. Reading that literature gets you very quickly to the interaction between structures of racial inequality and the importance of the southern United States in shaping policy. A short version of this narrative speaks about the migration of African-American populations from the south to the north in the World War II era, leading to the creation of urban ghettos and difficult ethnic politics between non-white and black populations, and then the emergence of the New Right in the south and west of the United States amidst a backdrop of stagnating wages since the 1970s.
Orenella is probably right in his diagnosis: if there were a third party, or more attention to a third party paid, the incarceration rate would probably drop a little. But I think he is dead wrong when he states that the “primary root” of the American outlier is in its electoral system. He is only assigning it such an important role because, I suspect, as a solid positivist social scientist, he is thinking in terms of correlations, rather than structures and mechanisms.
I will close by simply saying that this is not just an arcane matter over which social scientists quibble. This matters for making good recommendations for policy-makers and for making our own choices as citizens. Orenella recommends modifying the party and electoral systems to give more attention to the margins. I recommend looking harder at the underlying structures of this problem, particularly the racial dynamics such as racist attitudes and institutional relationships between the US’ ethnic minorities? I wonder what is easier? There isn’t a clear answer to this, but I will note that the two-party duopoly has been more or less permanently for around 50 years. Since that time we have had massive transformations in people’s attitudes toward minorities, policies of managing ethnic diversity and the legal structure that governs ethnic relations. That might provide a hint as to what is actually easier to change.