Last week, Alex Nowrasteh and Andrew Forrester published a Cato working paper arguing that the empirical trust literature is so rife with weaknesses that we can’t learn much from it. Their central finding is that sub-regional trust levels in the United States don’t correlate with economic output, despite trust theorists arguing that trust promotes economic growth (primarily, probably, because trust lowers transactions costs, increasing the amount of exchange, and because of better policy implementation through a more trustworthy civil service).
My sense from the trust literature is that many of these concerns have been amply addressed. I’ve spoken with some trust researchers I know and thought I’d blog some of the responses. The basic worry is that Nowrasteh and Forrester are sort of selective in the papers they feature. The trust literature is actually pretty self-reflective, and have addressed a number of these worries.
Here are what Nowrasteh and Forrester claim are sins in the literature. SIN 1:
The first deadly sin of the trust-growth literature is that it contains no macroeconomic growth model that incorporates trust, either in its micro-foundations or otherwise (Beugelsdijk and Maseland, 2011, p. 213). Furthermore, the trust-growth literature does not contain a formal theory of social capital formation broadly or one of trust specifically (Guiso, Sapienza and Zingales, 2011, p. 469). Most trust researchers aggregate assumed effciencies at the microeconomic level up to the macroeconomic level and assume that trust creates economy-wide growth: an illegitimate leap in the logic of micro- to- macro functioning (Beugelsdijk and Maseland, 2011, p. 208). The relationship between an individual’s trust and income may not be true for society and cannot be aggregated up to form a truthful representation of the whole (Beugelsdijk and Maseland, 2011, p. 208).
Response: the authors don’t engage with the part of the trust literature that explores transmission mechanisms, writing just a bit about theoretical models in pp. 1-2. A number of papers, including this one, find that human capital and the quality of judicial-bureaucratic institutions are the two main transmission channels. In the US context, the first is hard to pinpoint because people often don’t work where they were educated. The second is also a problem since much institutional quality is defined at the federal level.
Another difficulty Nowrasteh and Forrester cite is supposedly that the literature “contains no macroeconomic growth model that incorporates trust.” Apparently this is false. If the likely transmission mechanisms are through institutions and human capital, there are plenty of growth models illustrating such effects. What the formal models miss is how to explain how trust affects institutions and human capital. However, see this paper, this paper, this paper, and Zak and Knack’s highly cited investment driven model. Oh, and this paper and this paper too. On to SIN 2:
The second deadly sin of the trust literature is that the trust question itself does not produce internally valid responses. Recall, the trust question is: \Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?” There is no universal measure of trust because, in part, its meaning is culturally and contextually specific (Beugelsdijk and Maseland, 2011, p. xvii). The responses to the trust question are \Most people can be trusted,” \Can’t be too careful,” and \Depends.” The meaning of those responses is also unclear.
It’s not true that the trust question doesn’t produce internally valid responses. The primary argument they give is a twenty-year-old quote from Robert Putnam. But a number of studies do show that the generalized trust question does rather well: here, Sapienza 2013 which the authors cite but not on this point, and a Naef and Schupp 2009 working paper using the GSOEP dataset all show clearly that the question produces valid responses. Even Nannestad’s well-known 2008 review notes that the trust question has excellent test-retest stats. Now SIN 3:
The third deadly sin of the trust literature is that responses to the trust question do not generally predict trusting behavior in real-world micro-level experiments or in trust games.
The papers above also address this deadly sin, such as the Sapienza paper, which the authors cite, but, again, don’t draw the right conclusions from. Here’s another paper connecting trust responses and actual behavior. SIN 4:
The fourth deadly sin is that many of the major papers in the trust literature are contaminated by various types of sample biases.
The main findings in the literature can be replicated using much larger samples. And trust scores are so stable over time that slight demographic changes like those they describe are not relevant. SIN 5:
The fifth deadly sin is that even if the trust question were free from measurement error or sample selection bias, trust may be a proxy measurement for other deeper causes of economic development.
The trouble here is that the literature hasn’t found any examples of this. It’s also a bit of a cheap shot because that can be used as an argument against pretty much all findings in the empirical growth literature. SIN 6:
The sixth deadly sin of the empirical trust literature is that sub-national level data in the United States that is collected under better conditions do not indicate a robust positive relationship between trust and growth.
The sixth sin is only a problem if you think that regional trust should cause regional growth. But it might be that the different regions are sufficiently economically integrated with each other that isolating trust and growth in each region doesn’t tell us much.
In sum, Nowrasteh and Forrester seem to have missed some important papers that I think weaken their argument. It will be interesting to see how their paper develops once they take these papers into account. I think they’ll probably need to stop framing the paper in terms of identifying deadly sins, though.