In the aftermath of the financial crisis, we have seen two major approaches to bringing the economy back on track. In the U.S. a huge stimulus package was passed, and double-digit deficits were run; in Europe on the other hand government spending was restrained, and austerity ruled, even in countries with no sovereign debt problems. Which one of these approaches is better, and which one is generally recommended by economists?
This post presents the results of an agent-based computational model of the economy, which considers different policies and examines their implications for things like growth, unemployment, or likelihood of crises.
Thinking of the macroeconomy in terms of a macro 101 AS-AD diagram, one can understand the dilemma. As GDP shrinks, aggregate demand drops. With more unemployment and lower tax revenues, automatic stabilizers kick in, i.e. government spending increases while revenues drop. This usually results in a deficit. This is generally desirable, as the drop in aggregate demand has to be offset somehow. The problem is that as the government is running big deficits, it becomes more likely that it will default on its bonds. Thus the interest rates it has to pay on the bonds will go up. This is indeed a real problem in certain economies (think Greece or Portugal), but for countries that kept their finances relatively in order prior to the crisis (like the United States), this should not be an issue.
Consider the agent-based economic model of Dosi et al. (2014). This is a relatively big model with intermediate good (i.e. machines) producers, final good producers, banks, consumers, a central bank and a government. Banks can provide credit to firms, the central bank sets the interest rate, and the government collects taxes (on banks and firms) to pay unemployment benefits. This is the very basic description of the model.
First, the authors see whether the model reproduces empirical facts. Indeed it does. There is trend growth, and business cycle fluctuations in consumption, investment and GDP. The relative volatilities of these variables check out too. Their comovements with the business cycle are as in the real world as well. The model also produces a GDP growth rate distribution with fat tails, which is something mainstream business cycle models (DSGE) cannot do (according to the authors at least). The available credit and bank/firm profits also move along nicely with the business cycle as in real life. The frequency, length, and magnitude of crises in the model represent reality well too (see figure).
Some micro data is also reproduced by the model. For instance, the right-skewed distribution of firm size, or a fat-tailed distribution of firm growth rate. There is also heterogeneous productivity, which is persistent over time. These are all features that agree with empirical evidence and that mainstream macro models (DSGE ones) cannot reproduce.
It’s important to note that all these empirical regularities are created by the interaction of a large number of agents in a complex system. There is no exogenous shock or anything that makes these crises happen. This is a model that is calibrated, and then the authors let it run. This is quite amazing if you think about it, especially if you compare it to traditional business cycle models.
Now that we’ve established that we have a good model, what can we do with it? The authors run a couple of policy experiments. Specifically, they check what effect certain mixes of fiscal/monetary policies have on some economic variables such as growth, unemployment, or frequency of crises.
For fiscal policies they consider: no restraint (norule), deficits cannot exceed 3% (SGP), public debt needs to be below 60% (FC), SGP that’s not binding in a recession (SGP_ec), and FC that’s not binding in a recession (FC_ec).
For monetary policies they consider: an inflation targeting Taylor rule (TR_pi), a dual-mandate Taylor rule (inflation and unemployment targeting, TR_[pi,U]), TR_pi and TR_[pi,U] with the central bank also acting as a lender of last resort (LLR_pi and LLR_[pi,U]), and finally a situation where a higher public debt leads to higher interest rates on bonds (spread).
The authors take norule-TR_pi to be the baseline policy, and compare every other policy mix relative to the baseline. For GDP growth, this relative comparison is shown in the table below.
What we see is that GDP growth is definitely maximized with a dual-mandate Taylor rule monetary policy (what the Fed is doing at the moment). As for fiscal, the best seems to be FC_ec (which is probably the least restrictive fiscal policy after norule), but norule is a close second.
GDP volatility is again minimized by a dual-mandate monetary policy in most cases (in one case LLR_[pi,U] is marginally better). Furthermore, one should opt for no restraint on fiscal policy (norule) to minimize GDP volatility. Now SGP_ec and FC_ec are fighting for the second place, but FC_ec is somewhat better with the dual-mandate Taylor rules.
Unemployment is minimized by the dual-mandate Taylor rules (TR_[pi,U] and LLR_[pi,U]) with TR_[pi,U] being considerably better in one case. In other cases, the differences between these two are negligible. As for fiscal policy, the picture gets somewhat complicated. No restraint (norule) is generally the best, but FC_ec does marginally better if TR_[pi,u] is employed.
The likelihood of economic crises is once again minimized by the dual-mandate Taylor rule TR_[pi,u] with LLR_[pi,U] being once again a close second. For fiscal policy, norule is again the best.
As for inflation, perhaps unsurprisingly inflation-targeting monetary policies (TR_pi, LLR_pi) do the best job, but spread is also quite good. The differences compared to dual-mandate policies are, however, not so huge. The best fiscal policy is quite hard to determine here as the differences are not too large. SGP that is limiting deficits to under 3% does the best with inflation-targeting monetary policies, for the dual-mandate policies the differences are very negligible between fiscal policy options.
Finally, the amount of public debt is minimized by dual-mandate policies again. The best fiscal policy seems to be norule, but with the dual-mandate policies in particular FC_ec and SGP_ec work well too.
Thus in conclusion, it appears that a dual-mandate Taylor rule coupled with an unrestrained fiscal policy is the best for the economy. Perhaps, a limit to public debt (at 60% of the GDP) can be considered as an alternative to this unrestrained fiscal policy, but only with an escape clause during recessionary times.
This seems to be a nice conclusion, somewhat even obvious. I think the main problem may be that some economies had bad histories, and are currently stuck in a recession with high debt (Greece). It would have been interesting to see how the simulation performs, if there is a very high initial stock of debt. Thus in general, it would be nice to see what this model implies if all its initial conditions are calibrated to the current Greek situation for instance.
A second thing that comes to my mind is that unrestrained fiscal policy in the framework of this model is different from what it is in the real world. In this model, the government is in a sense a rational agent. It collects taxes only to pay unemployment benefits, nothing funny is going on. In the real world, politicians who introduce fiscal policy are not economists, and they have their own self interests. Thus, it is not at all guaranteed that if fiscal policy is unrestrained, then politicians will behave the same way the government in this model does. For instance, they may be more prone to spend more even in times when they shouldn’t just to win over voters, especially prior to elections.
It could be interesting to see a model where even political economy considerations are included. However, the biggest problem is that it is impossible to predict what “silly politicians” will do at any given point in time. Thus I’m not sure that there is a general model of politician behavior vis-a-vis their generosity when in power, that holds up well to empirical evidence. In addition, this may very well vary from country to country.
To sum up, this paper is a nice attempt to show that in general an unrestrained fiscal policy along with a dual-mandate monetary policy is probably the best way to go. The two weaknesses are that the authors do not consider the model in a situation where the initial conditions of the economy are like those of Greece for instance. So I would not treat this paper as a prescription to what Greece or other countries in similar situations should do. It is more about what the ideal thing to do is in the long run. Second, the model does not deal with the fact that politicians will abuse the lack of rules on fiscal policy. The results would probably still hold in case a technocrat government is in power though.