Program evaluation: estimation vs. simulation

Simulations are a tool that is more and more used to evaluate policies. This is in particular important when these policies have never been implemented before and thus there is no historical data to draw on. How good are such simulations? One way to check this is to do an ex-ante simulation of a policy change that has actually been implemented thereafter, and then estimate ex-post its effect.

Fabian Bornhorst does this with the PROGRESA program in Mexico. There, some families receive transfers if their children go to school. Bornhorst first draws a model of occupational choice and then estimates it using data that was available before PROGRESA was implemented. He then looks at outcomes if PROGRESA is applied in this model economy. This is then compared to actual outcomes. How well does the simulation fare? Not bad at all, but not perfectly either. It tends to indicate somewhat stronger outcomes than those actually observed. Also, it does not capture some differences across groups, but this can mainly be attributed to the level of heterogeneity within the simulated model, so this is fixable. Overall, very encouraging.

How to forecast inflation in Sudan

We tend to get interested in large developed economies because this is where we live and where loads of data are available to make interesting observations and run empirical exercises. But there are also lots of issues elsewhere that need to be studied and that are relevant, at least locally. Those are harder, because data is scarce and also because theory that could guide us may not be available.

Take as an example a study by Kenji Moriyama and Abdul Naseer that tries to forecast inflation in Sudan. This is very important in an economy as disrupted as this one, because the lack of efficient financial markets and banking leaves only currency as a tool for savings. If inflation is high or uncertain, using it for savings is not likely either. The problem is that there is very little data available for Sudan. The authors use ARMA techniques, which at least rely on a limited number of series but require rather long samples. But they can rely only on eight years of monthly data.

Another way to work this out would be to have a structural model, but this requires quite a few additional data series, which are not likely to be available. Maybe then, letting theory guide us could be a solution. This is particularly important in a country where shocks are very important and regime changes are likely. The Lucas Critique has a lot of bite here, and you want to go as deep as possible in the structure in order to study the reactions of agents and markets to situations that may never have happened before. But do we really have a good theory to describe Sudan, with its civil war, population displacement, humantarian aid, etc.? This is the kind of theory that needs development, as this is where the marginal return of theory to real-world well-being is the highest. And this is not just about inflation in Sudan.

Many people do not plan ahead

It is a well know story: people do not save enough for retirement, which is inconsistent with standard theory. Hyperbolic discounting comes here to the rescue, justifying this with the twisted preferences people have: they just value the present a lot. And what if people are just not planning ahead?

John Bone, John Hey and John Suckling conduct an experiment that allows to separate preferences from planning ahead. In particular, players should be anticipating what other players are doing. It turns out about a third do not. And asking them to pre-commit does not change this.

What should we take from this? The conclusion of the three Johns may only apply to the experimental environment, but not when it really matters, like planning for retirement. But there is plenty of evidence that many do not plan. While it may be rational for some (who do not expect to stop working or live long enough), many are still caught empty-handed at retirement. Should government then plan for them?

Are children a source of happiness or not?

I think that in every model with fertility choice, children are modeled as beneficial but costly. The number of children enters positively in the utility function, and there is some cost, either monetary or in time, of raising them. Yet, it appears that quite consistently empirical studies on the happiness of people reveal that the number of children has a negative impact on happiness indicators. Then why would people have children? Social pressure, bad luck with contraception or stupidity? What if the empirics are flawed?

Leonardo Becchetti, Elena Giachin Ricca and Alessandra Pelloni find that the typical way the empirical studies are conducted is flawed. Among the controls, there is usually income. But this make it difficult to disentangle the monetary for the non-monetary impact of children. Using equivalised household income (income adjusted for the number of people in the household) allows to focus only on the impact on happiness, and children now have a positive impact. Decomposing their sample (the German Socio-Economic Panel), they find that opportunity costs do matter. Theory is saved.

Is crime a habit?

Why are there repeat offender and career criminals? The obvious answers would be that this is where there comparative advantage is, or that they have nothing to lose once they served the first sentence. But what about crime being addicting?

This is what Vladimir Kuhl Teles and Joaquim Andrade explore what would happen if utility exhibit habit formation. Note surprisingly, it affects positively criminal activity. Note sure that this is very convincing. But the authors also embed an interaction with capital: Higher capital increases the rewards of crime, but also increases the opportunity cost of crime, the latter less apparently. Thus one should see more crime in capital-intensive locations, say cities, where habit formation is also easier. We should thus see more career criminals in cities. And I thought it was all about the anonymity that cities provide.

Growth leads to savings, not vice-versa

Fast-growing countries, like currently China, have very high savings rates. Data indicates that causality runs from growth rates to savings, and not the reverse. In theory, this is puzzling. Such high growth rates originate in rapid productivity improvements. This leads to high returns for capital and thus one should see high investment (and savings). However, returns for savings in such countries are very low. Why are people savings so much then?

Yi Wen finds one way to justify this: precautionary saving. We know that whenever there is a motive for precautionary savings, this can be rewarded with interest rates below the discount rate. And this is triggered by borrowing constraints. And it is well known that the Chinese financial sector is still severely underdeveloped.

The actual mechanism at play is obscure to me. The paper reasons that when permanent income increases, it is savings that increase instead of consumption, because of the borrowing constraint. The only way I can see this happening is when the uncertainty increases faster than incomes, or if utility is twisted in some way. But I do not seem to see either. While the author claims to have made a model that is tractable and analytically solvable, it does not appears to help in any way to understand what is going on. And Yi Wen does not seem to offer any explanation either.

Houses are a poor way to share risk

Owning a house is a terribly risky investment, not only because of the volatility of its value, but also because of the high correlation between house value and local economic conditions which also determine labor income. In other worth, owning a house you live in is among the worst things one can do in terms of diversification. Two recent papers revisit some aspects of this.

Dmytro Hryshko, María José Luengo-Prado and Bent Sørensen show that houses still provide substantial consumption smoothing ability to households suffering from a job loss. Obviously, when there is job loss and house equity loss, things look bad, but home owners still do better than renters. What this means is that the fact of holding some equity is more important that the lack of diversification. But it remains that better diversified households would still do better.

Todd Sinai and Nicholas Souleles argue that the risk is not that bad once you consider that when you sell a house you need to buy another one. So if you have to sell low, you are likely to buy low as well. This works obviously if you move locally, but appears also to work quite well for moves to another metropolitan area, as house prices across such areas in the US have an expected correlation between 0.35 and 0.60. So things are not as bad as one would have thought.