Does public debt secure social peace? A diversionary theory of public debt management

In the United States, for example, Peltzman (1992) shows that electors punish politicians who let public spending increase. The explanation is related to progressive fiscal systems: on average, voters are wealthier than non-voters and penalize spender incumbents because they will pay the price of future fiscal adjustments. Such fiscal “conservatism” of electors is also found by Bertola and Drazen (1993) and Brender and Drazen (2008), who show, using a sample of 74 countries from 1960 to 2003, that pre-election deficits reduce the chances of the incumbent being re-elected. Garmann (2017) shows that if voters are highly fiscally conservative, incumbents can even decrease spending before elections.

The intuition that debt can be used to manage social conflicts is not new. Capital structure theories (see, e.g., Grossman and Hart 1982; Jensen 1986; Stulz 1990, among others) emphasize the role played by corporate debt in reducing agency conflicts among workers, managers, and shareholders. Specifically, by increasing the probability of failure, private debt exerts a threat that can discipline workers by forcing them to accept low wages and poor working conditions. In this study, we show that the same type of argument can apply to public debt, which can be used as a disciplining device to limit rent-seeking behavior during the term.

This cohesion effect has been raised by a large number of social psychology and anthropology studies. For example, Sumner (1906) first introduced the term ethnocentrism and developed the idea that “the exigencies of war with outsiders are what make peace inside. (...) These exigencies also make government and law in the in-group, in order to prevent quarrels and enforce discipline.” Such a concept is introduced in a formal model of rent-seeking by Münster and Staal (2011). Their study provides one of the first economic models with simultaneous conflicts at different levels (see also Katz et al. 1990). However, Münster and Staal (2011) use a static setup and do not study electoral competition, while our model is dynamic and mixes the rent-seeking and probabilistic voting approaches. Moreover, in their model, there is conflict either between or within groups, but not both at the same time (in the basic case in which all players decide simultaneously and independently how to allocate their resources), while both conflicts can appear simultaneously in our setup. This is an important point because by using public debt, an incumbent can move the weight of internal and external conflicts according to his/her preferences.

See, for example, Pettersson-Lidbom (2001), Hodler (2011) and Kirchgässner (2014).

In the same vein, Aghion and Bolton (1990) present an interesting model in which the incumbent has an incentive to accumulate excessive public debt because he/she can more credibly commit not to default than his/her opponent.

In Alesina and Tabellini (1990), the incumbent has a chance of being re-elected, albeit his/her exact chances are uncertain; in Persson and Svensson (1989), the current government is certain to be removed.

For example, Hodler (2011) finds that the incumbent never manipulates his/her opponent’s public spending if he/she can ensure his/her own re-election.

We restrict our analysis to internal conflicts since external conflicts are not publicly available in Canada. The data on provincial internal conflicts come from the following sources (listed in Appendix A in Supplementary Material): “A Brief History of Canada”, “Riots and civil disorder in Canada”, “List of conflicts in Canada” and “History Since Confederation”. The data on provincial debt and economic measures come from Statistics Canada and the data on elections come from the Elections Canada website.

To appropriately account for the relationship between a discrete and a continuous variable, we investigate the polychoric correlation of Drasgow (1986). The polychoric correlation measures the association between ordered or categorical variables and one or multiple continuous variables.

Arguably, the events in Quebec are often related to issues of independence. This exclusion could also serve as a robustness check.

The CF approach’s advantage is that it allows for a simple endogeneity test via a Wald test (Wooldridge 2015). Given that the annual provincial debt per federal GNP was found to be endogenous, we implement the CF bias correction approach. The CF is a correction term needed as an additional regressor in the main equation of interest, as the outcome of interest is binary. A standard instrumental variables procedure cannot be applied to discrete outcomes.

As one may think that specific economic conditions might relate the mortgage rate to social unrest, through firm closures that induce industry instability and potentially lead to individual bankruptcy and foreclosures (see Huynh et al. 2010, 2017, for Canada), we checked if this potential argument is valid for Canada. Our findings suggest that for Canada, the bankruptcies are not positively correlated with the mortgage rates (we found a correlation of − 0.105 between bankruptcy and mortgage rates). Furthermore, the correlation between mortgage rates and social unrest is insignificant, the correlation below 0.005. Additionally, to the mortgage rate we consider the lag of the mortgage rate as an exclusion restriction, and the quality of the results did not change. Therefore, for Canada’s case, the proposed exclusion restrictions (mortgage rate and its square) also satisfy the conditional independence assumption as in Abadie et al. (2002).

We also test a penalized version of the logistic model that relates to rare events and the results remain unchanged.

After the election, the debt burden can be renegotiated inside the federation. Some districts can obtain debt alleviations, bailouts, or specific assistance from the federal budget depending on their bargaining power. Since the federal budget must be balanced, other districts will have to pay for these measures, which produces inter-district conflicts about the repayment of public debt.

We could also assume that there is a private good and that citizens’ utility (6) is consumption-augmented. In this case, citizens consume their net income (\((1-\tau )Y_{t}/M\) in a symmetric configuration), and our results are qualitatively unchanged.

Strictly speaking, this public good is an impure public good (i.e., a common good).

An axiomatic foundation is provided by Skaperdas (1996).

Thus, in the second period, the mere fact of having been the incumbent or challenger in the election may roughly characterize the “type” of office-holder.

This finding echoes the traditional diversionary hypothesis (see Wright 1965; Mansfield and Snyder 2004), which suggests that leaders might trigger conflict with other groups to deflect attention from problems at home. Wright states that “foreign war as a remedy for internal tension, revolution, or insurrection has been an accepted principle of government” (Wright 1965, p. 140).

To simplify our analysis, the decisiveness parameter of external conflicts (b) does not depend on the type of office-holder, contrary to that of internal conflicts (\(a^j\)). However, we could easily introduce different parameters such that \(b^I \ne b^O\). This assumption would unnecessarily complicate the model without affecting our results.

Each citizen is endowed with one unit of time. For simplicity, citizens simultaneously and independently decide how to allocate their resources to productive and unproductive activities.

A positive (resp. negative) value of \(\theta _{n}\) implies that citizen n has a bias in favor of the incumbent (resp. the challenger), whereas citizens with \(\theta _{n}=0\) are ideologically neutral. Further, w measures citizens’ responsiveness to rent-seeking activities. As w increases, citizens care more about rent captation than ideology (see Eq. 15).

Many empirical works support this specification. For example, the well-known “rally ’round the flag syndrome” suggests that external conflicts often boost the popularity of leaders (Grieco et al. 2014, p. 199). For example, some US presidents have enjoyed extra short-run popular support during the outbreak of international crises or wars (as, e.g., Lian and Oneal 1993; Hetherington and Nelson 2003, among others).

Indeed, in the symmetric equilibrium, we have \(\partial \psi ({{\hat{{\mathbf {y}}}}}_{2})/\partial e=-b{\tilde{N}}/e\). The second-order condition is verified since \(e\mapsto \psi ({{\hat{{\mathbf {y}}}}}_2)\) is a convex mapping.

Hence, following Munster’s (2007) definition, our model exhibits only a “reversed group cohesion” effect.

The second-order condition is satisfied since \((\alpha ^I)''(\cdot )<0\) (see Appendix B in Supplementary Material).

For example, Prime Minister A. Tsipras was claimed in 2016 that creditors have worsened the Greek crisis. See, e.g., https://​www.​theguardian.​com/​world/​2016/​sep/​11/​alexis-tsipras-greece-criticises-creditors-thessaloniki.