MARCO TEÓRICO
2.1. Prácticas de Laboratorio
2.1.1. Clasificación de las prácticas experimentales
In this section, the effects of a moreengaged civil society on the unique stable equilibria are analyzed. Formally, this is captured by an exogenous change in the model: the values ofp(λ)increase at all levels ofλ. In words, opportunists are more likely to get caught. LetpN(λ)denote this new function.31 To ensure
existence of a stable steady state,pN needs to satisfy the parameter restrictions
contained assumptions 1 through 5.
A change in civic engagement has an effect on the equilibrium (βA, λA). Substituting pN(λ)for p(λ)in the two differential equations governing the dy-
namics of the system (equations (3.8) and (3.9)), and solving for points at which dβ = 0 and dλ = 0 yields a new equilibrium. Graphically, the increase in p(.)
30The report, which collects specific descriptions of actions of civil society in Benin, is avail-
able at http://africastories.usaid.gov .
implies an upward shift in Figure 3.1. This shift implies a leftward shift in the locusdβ = 0at the stable steady state as illustrated in Figure 3.5.
The effect on the differential equation of bureaucrats is also intuitive: for each level of β, a lower level of λ is tolerated. Graphically, this is illustrated
by a downward shift of the loci dλ = 0 through the corner (β = 1, λ = 0).
Figure 3.5 depicts the transition of the economy from the old to the new steady state (βA′, λA′). In the new steady state, the number of corrupted bureaucrats is unambiguously lower. However, the effect onβ is ambiguous since it depends on the effects on the loci dβ and dλ: if the effect on dβ is greater (smaller), β increases (decreases).
As civil society becomes more engaged, there is an immediate decrease in the number of corrupted bureaucrats, while the level of honest citizens is un- changed; this is illustrated by a horizontal left arrow in both panels of Fig- ure 3.5. In the case where the effect ondβ is bigger (top panel), the evolution- ary forces lead to an increase in β such that the initial decline in λis followed by subsequent decreases inλ. These decreases are the result of the increasing difficulty of finding citizens willing to participate in bribery. The system finally stabilizes at the new steady state(βA′, λA′). Therefore, under this scenario, the
change in civic engagement generates a virtuous circle whereby more social capital emerges (more honest bureaucrats and citizens) and corruption falls.
In the scenario where the effect on dβ is smaller (bottom panel), the evo- lutionary forces favour opportunism and the mass of citizens willing to bribe increases. It follows that, after an initial decrease, the number of corrupted bureaucrats increases. Indeed, the increased ease with which bureaucrats can
Figure 3.5: Convergence paths: corruption falls. λt βt 1 0 λA′ λA 1 (βA, λA) (βA′, λA′)
(a) Effect ondβis greater: λdecreases andβ increases.
λ t β t 1 0 λA′ λA 1 (βA, λA) (βA′, λA′)
find citizens willing to bribe, and the increased size of bribes, attracts more bu- reaucrats into being corrupted. The system finally stabilizes at the new steady state(βA′, λA′).
Therefore, in the new equilibrium, the change in the number of honest agents is ambiguous, while the number of corrupted bureaucrats falls unam- biguously. Empowering civil society is thus effective in the fight against corrup- tion, and the dynamics of adjustment, outlined here, provide a possible expla- nation for why countries that developed engaged civil society have been able to
reduce corruption. Furthermore, the ambiguous effect onβ stresses an impor-
tant result. Countries enjoy lower levels of corruption, not necessarily because they have more honest citizens (i.e., more social capital), but rather because, for each level of honest citizens, lower levels of corruption aretolerated. This is similar to the idea of an access to a bettertechnology.
It is interesting to contrast the effect of civic engagement with the fact that wage policies are ineffective in this setup. Specifically, increasing the bureau- crats’ wageswB would have no effects on the loci dβ and dλ, and would, there-
fore, be ineffective at reducing corruption.32 This prediction is consistent with
several empirical works that found no statistical significant relationship be- tween wages and corruption.33
32It should be acknowledged that this result depends on the specification of the sanction. If
the sanction on bureaucrats is made proportional to wages, wage policy can affect corruption level. In addition, civic engagement tends to be higher in countries with higher levels of wages. Hence, it cannot be ruled out that wage policies could play a role - even indirectly through civic engagement - in the fight against corruption.
33For example, using data from public hospitals in Argentina, Di Tella and Schargrodsky
(2003) find no statistical relationship between pay and corruption - even after making adjust- ments for hospital size, experience, and other factors. Savedoff (2008), in a note on the issue, depicts similar patterns in Venezuela. For a list of papers on the relationships between wages and corruption, see Di Tella and Schargrodsky (2003).
This section has demonstrated that civic engagement can play a role in the fight against corruption by shaping citizens’ behaviour and monitoring bureau- crats. The theory predicts three important results that can be tested empir- ically. First, a negative relationship between civic engagement and corrup- tion. Second, an ambiguous relationship between corruption and social capital. Third an ineffective role for wage policies. The next section demonstrates that these results (except result three due to a lack of data) are supported by the data.
3.4
Empirical Evidence
This section provides empirical support for the predictions of the model. The data is taken from Round 3 of the Afrobarometer surveys, which were con- ducted from March 2005 to February 2006. The Afrobarometer “measures the social, political, and economic atmosphere in Africa”, and is designed for cross- country comparisons. The Afrobarometer is particularly suited for this paper since it provides direct data on corruption, and variables that perform well as proxies for the engagement of civil society. Round 3 of the Afrobarometer cov- ers the following 18 countries: Benin, Botswana, Cape Verde, Ghana, Kenya, Lesotho, Madagascar, Malawi, Mali, Mozambique, Namibia, Nigeria, Senegal, South Africa, Tanzania, Uganda, Zambia, Zimbabwe. Dropping all missing observations on age and education leaves 25,021 potential observations. The section begins with a presentation of the variables that support the analysis, followed by a discussion of the regression results.