In the last decade we have witnessed two major systemic financial crises, namely, the 1998 Russian crisis and the current crisis, the latter initially associated with the subprime mortgage market (henceforth, subprime crisis). A critical event in the subprime crisis was the Lehman Brothers’ episode in September 2008. Lehman’s collapse, coming on the heels of the sell-off of Bear Stearns, took the market by surprise. The ensuing about-face regarding AIG was perhaps less surprising but still added a heavy dose of policy uncertainty.
Many observers have been critical of that erratic policymaking and see it as directly responsible for the worldwide collapse of stock markets and near panic that took place soon afterwards. Repeated bouts of time inconsistency – as I will characterise this type of episode – have also been argued to have triggered the spreading of the Russian crisis across most emerging market economies in August 1998. As the argument goes, the market was aware that Russia was facing an unsustainable fiscal deficit before August but it was expecting that if a run against Russian public debt obligations materialised, the IMF and other multilateral institutions would rise to the occasion and bail them out. One often-heard reason for this was that Russia was “too nuke to fail.” However, the bailout did not happen, Russia was forced to default and, as if coordinated by a magic wand, the JP Morgan EMBI for all emerging markets went through the roof, with the average interest spread exceeding 1500 basis points. This episode raised fears that Brazil – Latin America’s kingpin – would follow suit and collapse, and apparently led the IMF to soften its stance and extend a very generous standby loan to Brazil on the basis, according to rumour, of skimpy fiscal account information – likely lowering IMF’s credibility as enforcer of market discipline.1
These episodes are cases of what one might call “triple time inconsistency”. First, a public institution is expected to depart from earlier statements and offer a bailout to prevent a major crisis (this is the first round of time inconsistency); then, in an attempt to regain credibility, the bailout is pulled back (the second round) and, finally, having witnessed the wreckage caused by the policy surprise, it resume bailouts of the still-standing dominoes (third round). This seesaw policymaking cannot be right. The initial refusal to continue offering bailouts can only be justified as a warning signal to market players against getting involved in situations in which they will need a bailout. But this “investment in credibility” goes to waste as the policymaker chickens out and bailouts resume.
The example below aims at making these intuitions more precise. I think the effort is worth it, given that the similarity between the episodes highlighted above suggests that the phenomenon is likely to repeat itself unless we better understand it and develop ways to prevent it. I would like to point out, however, that the example focuses on the costs of different rounds of time inconsistency but stops short of addressing the policy uncertainty generated by policies that are not in line with private-sector expectations. The latter is a major task that requires more substantive research.
A simple model of “triple time-inconsistent” actions
Consider an economy with a two-period time span: today and tomorrow. Individuals have an output endowment today that they could consume or allocate to capital accumulation. This is the only decision that they have to make today. Tomorrow’s output can be produced by two independent technologies: (1) output proportional to the capital stock (which is predetermined as of tomorrow) and (2) output proportional to labour supply (which individuals will choose tomorrow as a function of the wage rate net of taxes). Net (consumable) income is output minus taxes. Taxes are of two types: (1) a tax proportional to capital holdings and (2) a tax proportional to wages. The government collects taxes in order to pay back outstanding debt. In a fully conventional way, I will assume that social welfare depends on today’s consumption, tomorrow’s consumption, and tomorrow’s leisure.
From today’s perspective, the two taxes are distorting. The capital tax distorts allocation between consumption today and tomorrow, and the wage tax distorts the allocation between consumption and leisure tomorrow. Therefore, it is intuitive that there should be a robust set of cases in which optimal taxation from the perspective of today calls for setting positive taxes on capital and wages. However, if a benevolent government is free to reset taxes tomorrow, it will set the wage tax equal to zero and finance government expenditure entirely out of capital taxes (assuming that government expenditure does not exceed the value of the capital stock). This is because tomorrow the capital accumulation decision has already been taken and there is no output or welfare cost (i.e., no distortion) involved in changing the tax on capital. Thus, there are solid grounds for individuals to expect that a benevolent government will be time inconsistent and that government expenditure will be fully financed from capital taxes. This outcome, of course, is suboptimal from today’s perspective but, unless institutional constraints bar a benevolent government from engaging in time inconsistency, once tomorrow arrives, the best available policy will take that form (This suboptimality of time-inconsistent policy is well known; see Kydland and Prescott 1977 and Calvo 1978).
Let us consider the case in which the private sector has reached the conclusion that the government will follow the time-inconsistent optimal policy (in the above-mentioned episodes this would correspond to the cases in which, counterfactually, the IMF bailed out Russia in 1998, and the Fed/Treasury bailed out Lehman Brothers in 2008) and suppose that, contrary to those expectations, the government decides to stick to today’s announcement and avoid the first round of time inconsistency. Notice that tomorrow’s economy is already exhibiting the effects of expected time inconsistency. Therefore, in that context, from the private sector’s point of view, the government not behaving in a (first round) time-inconsistent way boils down to a second round of time inconsistency. Could this be optimal? The second round of time inconsistency takes place tomorrow when the cost of lower capital accumulation (as a result of expected first round of time inconsistency) has already been incurred. Thus, the best policy from tomorrow’s perspective is to eliminate wage taxation; in other words, to behave as expected by the private sector. In this simple model, the second round of time inconsistency pointlessly reduces welfare. In a richer and more realistic model, though, the second round of time inconsistency may have some redeeming value because it could send a strong signal that the government is prepared to incur severe costs to ensure the credibility of time-consistent policy. But the third round of time inconsistency (corresponding to the about-face after failing to bailout Russia and Lehman Brothers) destroys the credibility investment in one fell swoop.
Improving public policy responses
It is far from me to chastise or ridicule those involved in triple time inconsistency. There are always good reasons why bright and well-intentioned public officials make serious mistakes during major crises. The two cases singled out in this note took place in arguably “unprecedented” circumstances. In situations like these, “shooting from the hip” becomes the rule, and errors are to be expected. However, I believe that there are at least two lessons that we could draw from these episodes, which could help to lower the incidence of triple time inconsistency and other inefficiencies:
- A financial crisis is not the best time for reform or building credibility, especially if those actions go against the private sector’s expectations. Policymakers should focus their attention on putting out the fires and minimise the short-run social costs.
- Policymakers should spend more time discussing worst-case scenarios before crises occur. These discussions should be carried out with some regularity (much like fire drills) and involve a wide spectrum of public officials that might eventually have to be involved in rescue operations during crisis. This will ensure a better understanding of the involved risks and tradeoffs, and improve the effectiveness of policies that need to be implemented in the spur of a moment.
Calvo, G. (1978), “On the Time Consistency of Optimal Policy in a Monetary Economy,” Econometrica 46, pp. 1411-1428.
Calvo, G. (2005), Emerging Capital Markets in Turmoil, Cambridge, MA: MIT Press, Chapters 5 and 12.
Kydland, Finn E. and Edward C. Prescott (1977), “Rules Rather than Discretion: The Inconsistency of Optimal Plans,” Journal of Political Economy, 85, 3, June, pp. 473-492.
1. More precisely, the erratic policy may have led key market participants to make horribly wrong bets, which triggered margin calls and induced them to sell, or stop from buying, bonds from other emerging markets. Less informed investors, in turn, could have (rationally) inferred that something was amiss in emerging markets, increasing the selloff. This argument is more fully developed in Calvo (2005).