Trust and Financial Crisis Experiences

Also trust in central banks has declined. See Ehrmann et al. (2013) for an analysis of trust in the European Central Bank.

Source: http://​www.​financialtrustin​dex.​org/​resultswave22.​htm. URL last accessed 6 March 2015.

Source: http://​www.​gallup.​com/​poll/​162602/​european-countries-lead-world-distrust-banks.​aspx. URL last accessed 4 July 2013.

In contrast, there is more research on trust in public institutions. For instance, Marozzi (2014) uses the European Social Survey to construct an index of trust in public institutions in Europe. Likewise, several studies examine trust in government (e.g. Poznyak et al. 2014). Kong (2013) provides a further discussion of this strand of literature.

For further details we refer to Banfield (1958), La Porta et al. (1997), Bjørnskov (2007) and Uslaner (2002, 2013) and references therein. Bjørnskov and Sønderskov (2013) offer a critical discussion of the concept.

Some studies reach more qualified conclusions on the relationship between trust and performance. Roth (2009) argues that increasing trust may be detrimental for economic performance when trust starts from a high level. Dearmon and Grier (2011) find that promoting investment through institutional reform is less effective when trust is high.

Some other papers focus on cross-country differences in generalized trust. For instance, Bjørnskov (2007) considers several potential determinants of cross-country differences in trust. In line with Uslaner (2002), he concludes that income inequality is the most important determinant of generalized trust. Likewise, using individual panel data for Swedish counties, Gustavsson and Jordahl (2008) find that income inequality is an important driver of generalized trust. They also report that the proportion of foreign-born inhabitants in a geographical region is negatively related to trust. Based on data for U.S. localities, Alesina and La Ferrara (2002) reach a similar conclusion. These authors also find that low generalized trust is related to recent traumatic experiences, such as illness or divorce. From the perspective of the present study, it is interesting that their evidence also suggests that financial misfortune is closely associated with low trust.

Information on the CentERpanel is available at http://​www.​centerdata.​nl/​en/​about-centerdata/​what-we-do/​data-collection/​centerpanel. URL last accessed on 18 February 2015.

The CentERpanel and DHS have been used in several recent articles such as Van Rooij et al. (2011, 2012) and Van der Cruijsen et al. (2012, 2013a). Van der Cruijsen et al. (2012) examine how crises experiences affected respondents’ saving decisions. They report that respondents who were customers of troubled banking institutions were subsequently more likely to spread their savings across accounts at several banks. Likewise, these respondents were also more likely to move funds across banks. Finally, Van der Cruijsen et al. find that the size of the shock is important as the strongest effects are found for respondents who experienced both a bank bailout and a bankruptcy. Van der Cruijsen et al. (2013a) use the DHS to examine what the general public knows about banking supervision. They conclude that the public’s knowledge about banking supervision is far from perfect and that respondents often expect more from supervisors than they can realistically achieve. In addition, their findings suggest that better-informed people have more realistic views on banking supervision.

Van Beuningen and Schmeets (2013) construct an index for social capital for the Netherlands, which includes generalized trust using a similar question.

Respondents who answer “I don’t know” are not included in the analyses.

One could expand the set of control variables, for instance, by including information on the amount of financial assets and the number of accounts. When we include these variables in the regressions, our general conclusions hardly change. However, the number of observations is reduced substantially. We thank the referee for this suggestion.

If there is no direct effect on trust from these two major events, it is not likely that crisis experiences will have an impact in later years.

As explained in Sect. 2, we prefer estimating separate regressions for the 2006–2012 period and for 2013. However, we have also estimated the models for the 2006–2013 period. Appendix 4 reports the results. The findings are comparable to those presented in Tables 1 and 2.

Q1, Q2, Q3, Q4 and Q4a are standard trust questions, which are asked every year. Q27 and Q28 are questions which were only asked in 2010 to measure crisis experiences.

These questions were included to measure recent crisis experiences (i.e., the nationalization of SNS Reaal).