Is trust the missing root of institutions, education, and development?

It bears mentioning that the direct link between education and development is not undisputed. Pritchett (2001), for example, notes several problems relating to measuring average impacts and achieving the ‘right’ mix between primary, secondary and tertiary education. Berggren et al. (2008) and Beugelsdijk et al. (2004), in extreme bounds analysis, also find only weak support for education as an important growth factor.

In principle, one could think of arguments similar to discussions of the duration and protections offered by patent rights that would indicate some optimum level of trust. For example, extreme levels of trust may imply that industrial secrets are transmitted so slowly that the diffusion of technology is hampered.

The early literature on social capital and trust outlined a number of direct mechanisms connecting social trust to output, summarized in Fukuyama (1995), Putnam (2001) and Bornschier (2005). They pertain to reduced transaction costs in society as a whole, and a general acceptance of anonymous cooperation in high-trust societies, as the perceived risk of someone taking advantage of another is lower. Ikeda (2008) also suggests that trust allows entrepreneurs to have more impersonal contacts, thereby accessing a wider range of knowledge resources. In general, though, the empirical literature has failed to support such direct transmission mechanisms.

We thank Martin Paldam for popularizing this type of easily interpretable figure.

Descriptive variables are provided in Appendix A.2.

More precisely, we use the variable RGDPCH of the Penn World Tables for per capita GDP, GDPWOK for GDP per worker, and the ls25_2005 variable in Barro and Lee’s dataset for education.

We get an additional observation for Jamaica from a large, representative survey. The survey is described in Bourne et al. (2010). The inclusion of Jamaica is the only difference from the dataset in Bjørnskov (2012). The list of trust scores is displayed in Table A.1.

We acknowledge that the assumption may be questionable in the case of the United States, where trust levels fell in the 1970s and 1980s.

Reverse causality would, for instance, be a concern if trust increases as countries grow richer from total factor productivity growth, as Paldam (2009) argues. We note that the contention has, however, been rejected by Delhey and Newton (2005) and Bjørnskov (2007). In general, recent studies suggest that trust in most countries is remarkably stable over time (e.g., Uslaner 2008; Nunn and Wantchekon 2011). As corruption certainly has declined and property rights institutions are stronger in most of the world, compared to immediately after World War II, while trust seems to have remained approximately stable in most of the Western world in the same period, it seems logical to assume that causality mainly runs from trust to institutional quality (Bjørnskov 2007).

With Hippocrates, one can find arguments along these lines in his “On Airs, Waters, and Places”, in particular, parts 12 and 23. The Aristotelian view, expressed in “The Politics”, book 7, claims that populations in colder climates have more freedom, yet have no (formal) political organization.

An example often mentioned in this strand of the literature is that of Denmark. The last successful attempt at killing a leading politician dates back to 1286, while the last attempt occurred in 1885. Norwegian history, although substantially shorter as an independent nation, is equally peaceful.

A potential worry is that the findings are overidentified, even though the Sargan tests in Table 3 remain insignificant. As noted in Sect. 3, we have addressed this by correlating the residuals from the 3SLS GDP estimates with social trust. While there is a significant negative association, it is limited in size. This suggests that the bulk of the impact of trust on GDP is mediated by education and institutions. Technically, this result also implies that our coefficients are approximately unbiased.