1.4 Estimation and Identification
1.4.2 Discussion of the empirical model
My modeling approach impose several restrictions on players’ behaviors. Moreover, there are certain aspects of the setting that I choose not to model, which also grants discussion. In this section, I go through these and discuss their potential impact on the parameters estimated.
Finite horizon
As discussed while presenting the model, the finite horizon assumption has important impli- cations with respect to the possible equilibria of the game. For example, an equilibrium with logrolling where one type of projects are implemented first and the other second is ruled out. In finite horizon, players cannot credibly commit to stick with such a strategy. The OCADs continued through to 2015 with the same players. However, six months after the span of my data ends, ta local election took place. This is particularly important because the term-limits for mayors and governors in Colombia is a single cycle. Therefore mayors and governors in the OCADs necessarily changed soon after my data ends. However, it is not uncommon that a strongly connected candidate runs in place of the incumbent and represent its legacy.
It is also useful to think about the implications of the alternative model with infinite horizon. In this class of models, instead of a utility from savings, players would value the funds in the bank because of the future projects. At the end of each cycle, OCADs would draw
another budget and continue with the same stage game. Now, in this setting, consider a municipality with low administrative efficiency. For this municipality, the expected utility from the next budget cycle would be lower than that of the saving motive finite horizons model. This is because it is very hard to draw projects for this municipality in the next cycle as well and hence a dollar has a different value to it than it does for an efficient municipality. This would imply that the inefficient municipalities would be more willing to take on a project at any time as their opportunity cost of forgoing a project is lower. The finite horizon assumption does not completely undermine this effect, however. In any period other than the last one, inefficient municipalities have lower opportunity costs simply because it is less likely for them to draw projects in the future. However, in the last period, finite horizon model assumes an equal opportunity cost for all types of mayors.
Moreover, the infinite horizon model would be able to generate logrolling. However, notice that in this class of models, the set of equilibria is very large and it is empirically impossible to differentiate the effect of Supreme Court decision from that of a potential logrolling.
Presidential Election of 2014
The first round of Colombian Presidential Election of 2014 was held on May 25, 2014. This date coincides with the span of my sample. In particular, it is after the central government lost its veto power. Both this election and the second round on June 15th were close elections which resulted in the reelection of President Santos. It is not clear whether Santos used the OCADs as a method of clientelism. It is possible that the central government acted more lenient in accepting transfer projects around the election to collect the support of local mayors. If this channel truly exists in my sample, this would upward bias my estimate of the impact of the veto.
No anticipation of the Supreme Court Decision
An important assumption I make is that players are not able to anticipate the change in the institution. There are several reasons to believe that this was actually true. A lawyer, who was not affiliated with any local or state government, filed a suit to the Supreme Court suggesting that the central government’s veto power was unconstitutional. This case did not make the headlines in the media until the actual Supreme Court decision. In fact, most of the DNP bureaucrats were not even aware of the case until after the decision was made. It is also very unlikely any mayors anticipated the decision, as the lawyer was based in Bogota and was not representing any entity.
It is still useful to discuss the implications of this assumption. Suppose that a mayor anticipates the decision. If her preferences are aligned with that of the center, anticipation will not have an impact on the approved projects. On the other hand, if the preferences of the mayor and the center is different, she will be more likely to not to propose projects while the central veto lasts. Proposing these projects in the non-veti period will have a higher probability of success with the vote of the governor. As will be discussed in the next section, most mayors prefer transfer (public goods) projects more (less) compared to the central government. Hence, if all the mayors had anticipated the decision, we’d expect mostly the municipalities whose mayors prefer public goods to be active in the veto period. Hence, anticipation by the mayors will upward bias the estimate of veto’s effect.
Alternatively, if the central government anticipates the Supreme Court decision, its con- tinuation payoff will be lower. This is straightforward to see, as there will be a positive probability of future projects that the center would otherwise veto. Under some assump- tions, this will imply that the representative will be more lenient toward projects –that is, it will be more likely to accept projects that it did not before. Hence, the bias on the estimate of veto’s effect would be downward.