Constructing our own cage (3-6-2020)

Written by: Jan Luiten van Zanden

In medical research there is the following problem: if some kind of experiment with a new drug is carried out, positive results are easy to publish. The sponsor, often the company which has developed the new drug, will certainly encourage it, and journals are of course keen at such ‘good news’. But when no effects of the new drug are found, there is no incentive to publish about the test; moreover, a journal will not be interested in a paper which presents no clear results. There is therefore a strong selection bias in published outcomes of medical tests – good news gets a lot of attention, bad news (and no news) is swept (or kept) under the carpet.

Do similar mechanisms select research outcomes in economic history? Of course, like academics in medical sciences, economic historians want to score, to draw attention to their work with new, surprising results. To get a paper published,  you have to find ‘something’. And a negative outcome – “we did test all the hypotheses and did not find any statistically significant coefficient” – is not appealing. Such papers will not be published, and probably not even be written as there is ‘nothing’ to report. But from a broader perspective, as in medical experiments, ‘nothing’ is also a result, and in theory equally interesting.

There is perhaps one topic that is relatively sensitive to this kind of selection bias. In the past 20 years or so a small cottage industry has emerged to study the phenomenon of ‘path dependency’. Economists and economic historians have argued that contemporary phenomena – say the divide between the north and the south in Europe (or the east and the west), or between south and north of Italy – have deep historical roots. Institutions are the product of long historical processes, and to make the point that ‘history matters’ many papers have been published which established some kind of link between something in a very distant past – sometimes going back to the Neolithic Revolution, or before that, the Exit from Africa – and contemporary variables of interest (GDP per capita, or female labour force participation). The problem is that we probably never see the results of failed experiments to establish path dependency, in which no significant link between the event in deep history and the contemporary world is demonstrated.  So we have witnessed a cumulation of papers showing that we are the prisoners of our past, and no papers have been published arguing that we are in fact free, that history did not dominate this or that institution or outcome. I know at least of one paper which tried to establish a link between VOC activities in the past and current outcomes, which has not been finished and published, perhaps because the authors could not find such a link. But instead of publishing this liberating outcome – ‘free at last! from our past’ would be a nice title – the paper simply disappeared.

Perhaps it is a more general problem: theories always imply that we are bound by the causes identified, that our behaviour is not entirely free but constrained by the ‘independent variables’ specified by the model. The backlash of the scientific project started in the 17^th and 18^th century to make us masters of the universe, is that we tried to include homo sapiens and her society in the big model as well. As economic historians we share in this tension between scientific law and human freedom – the more we are able to understand how the past ‘determines’ the present, via complex forms of path dependency, the less real freedom we seem to have. But the selection bias we started with means that the arguments that we indeed live in the cage of our past are over-represented in academic debate, and that we tend to ignore the evidence that we are basically free and able to recreate the present and the future.

Continue reading: My first desk rejection (4-6-2020)