It Would Be a Hard Survey to Design

Nato argues in a response to my last post:

The constant references to sentiment don't seem to include links to actual surveys, but rather to people the author talks to. The counterfactual asseverations don't seem to come with any actual support. Krugman et al. can (and do) claim that the stimulus was much smaller than it should have been according to their calculations, and that seems to be a plausible and natural defense of the mismatch between claims of benefit and results. There's a significant chance that the defense is wrong, but it's not special pleading.

Now, I think this is quite wrong, and let me see if I can explain why. For conservative critics of Obama's economic policy, the counter-factual is some kind of mean reversion. The idea is that the economy has a sort of equilibrium state or equilibrium trend, and that when the economy deviates from trend, it tends to move back to it, if policy does nothing. If policy seeks to prevent deviations from trend, and deviations from trend do not occur, that suggests that maybe policy is preventing them, that is, policy is helping to smooth the business cycle, to make the economy more stable. That might have seemed true, say, in the 1990s. If, on the other hand, the economy stays a long way below trend for a long time, the counter-factual is that the economy would be on or returning to the trend, and the first hypothesis to consider for why it is not doing so is government policy. So the main basis for the counter-factual presumptions is simply the historical trend, though other evidence is nice, too, if it is available.

By contrast, it's difficult to express the kinds of assumptions that must underlie Krugman's argument for a bigger stimulus without undermine its plausibility. Thus, you might say that if there had been no stimulus, we would have been even slower to get out of the recession. To claim that, one must either deny that the economy has any inherent tendency to revert to trend, or claim that something has suddenly made that tendency inoperative. That claim is destitute of the kind of historical induction which underlies the conservative position. If one does accept that there would be a natural reversion to trend, then the counter-factual presumption has to be that the economy would have been recovering if government policy did nothing, and it's not recovery but the feebleness of the recovery that must be regarded as the result of the stimulus policy. If this is granted, one might still argue that the effects of stimulus are non-linear, that a relatively small (though still quite large!) stimulus had a negative result, but a larger stimulus would have had a positive result. But if the effects of stimulus are non-linear, it seems more likely that the effects of stimulus get worse at the margin, that the "low-hanging fruit" are picked first, and more spending means less valuable public-sector prrojects and more displacement of private sector activity. So I think Nato is quite to think that the critics of Obama's policies are more to blame for unsupported counter-factual asseverations than Krugman, it's the other way around. All Krugman has to support his claims that a bigger stimulus would have worked are deductions from a Keynesian theory that has not fared well theoretically or empirically in recent decades, and which the apparent failure of a smaller Keynesian stimulus further undermines.

However, what prompted this post is the following:

I have not yet encountered any that weren't based on assumed counterfactuals and dubious claims of sentiment. I have sometimes argued with these, and sometimes not, but it seems each time I've challenged them, the writer either retrenches somewhat or doubles down on the 'sentiment' point without offering additional evidence.

When I responded to this with a link to a survey showing pessimistic sentiment among households, Nato objected that it was "household sentiment about the economy, not businesspeople about the government." Well, but businesspeople are householders too, no? Still, it got me thinking about what the ideal instrument would be to test the hypothesis that businesses' reluctance to hire reflects heightened political risk.

First, what you'd really want to do would be to write a questionnaire and then go back in time and ask people the same questions in 2007, then ask them in 2011, and compare answers. If you just have the answers today, they would be hard to interpret. Suppose 95% of businesspeople say they are worried about higher taxes. Well, maybe businesspeople always know there's a chance taxes will rise, and always try to be ready for it. If, on the other hand, 25% of businesspeople say they're worried now, but only 10% did in 2007, that would provide strong support for the political risk hypothesis. Unfortunately, you can't do that, and it's very likely that many of the questions you would want to ask now, no one would have thought of asking in 2007, because the politico-economic environment has changed in such unanticipated ways. For example, we might now want to ask: "Do you sometimes take into account the possibility of future government bailouts when you make business decisions? How?" Who would have imagined in 2007 that that could be an important question to ask businesspeople?

Second, what is your sample, and how do you weight it? If your sample is of "firms," represented by, say, CEOs, a big problem is that firms vary hugely in size and are not of anything like equal importance. So, do you interview the bosses of Fortune 500 companies, or mom-n-pop stores, or something in between? If you interview firms of different sizes, do you give firms of greater size greater weight? Once you do that, though, your statistics start to become strange. What are you actually reporting? You can't say "55% of bosses said that..." Your number might accurately be stated as "55% of firm value is represented by bosses who said that..." Worse, there will be so many different weird schemes you could adopt to deal with the problem that people could suspect you of manipulating the results, and someone else using the same data but adopting a different, and equally defensible, weighting procedure, might find a different result. And yet, if 95% of bosses don't think at all about government bailouts, but 5% have switched from productivity-maximization to rent-seeking in the past five years, and especially if those 5% are highly placed, that might explain much of the deterioration of the economy.

Or is it businesspeople you want to interview at all? Maybe the businesspeople are as confident as ever, but they can't get credit to expand because the people with the money are worried about political risk. How would you find that out? You'd need a sample of bankers or investors. Also, remember that the sample of businesspeople is endogenous. Some people might be businesspeople now because they got thrown out of jobs by the recession and so tried self-employment instead-- or because they are politically well-connected and found that a more useful asset in a more interventionist time. Others may have been businesspeople two years ago and aren't now because their businesses went bust in the recession. More subtly, some people would have been businesspeople but for the recession, that is, they had business ideas but didn't pursue them because the environment seemed unfavorable. Or conversely, they might have planned to abandoned self-employment, then decided to stick with it because the times were unfavorable to looking for a salaried job. In short, if businesspeople are your sample, the sample is endogenous. How do you deal with that? (Some of these people might not even be in the country: immigrants who went home, or didn't come, because of the bad economy.)

There might be clever ways to deal with these statistical problems. But I think the demand for "evidence" in the form of polls or surveys might be a little hasty. There's merit in just grabbing stylized facts and thinking about them. The economy's performing way below trend. Why? Obamacare and the removal of the Bush tax cuts are going to cause substantial tax hikes for the rich, and Obama still seems to think there's room for more. How does that affect the incentives of entrepreneurs and investors today? No one seems to understand Obamacare, what obligations it places on employers and individuals, how it's going to affect prices. How is that going to affect hiring decisions? Corporations have been making big profits in the past couple of years, and P/E ratios are low. If capital's earning a lot, why don't people want to make more of it by investing? Political risk makes a lot of puzzle pieces fit together.

Comments

[Nato]

"Political risk makes a lot of puzzle pieces fit together."

This is a good hypothesis, and I think it's certain that it forms some part of the answer. However, I don't see any good reason to assign it a large percentage. Some reasons, in approximately descending order: 1)The elevation in unemployment is largely structural. Hiring manager surveys have shown that companies want to hire but can't find qualified candidates. It doesn't surprise me, considering the skills of the hardest-hit industries don't apply well to growth industries. Looser immigration restrictions would help even unemployed Americans, as more jobs would be created, and though the 'primary' positions might be filled by imported labor, secondary support jobs would open up for the not-quite-qualified. Of course, in *very* strong hiring environments, companies make do with the not-quite-qualified, but the point here is that actual hiring seems likely to understate the degree to which companies would prefer to expand, when the mismatch is extra problematic 2)Financial crisis blah blah blah no need to belabor that point 3)Even leaving the overall financial crisis cycle argument, credit spreads remain such that credit is tight at the same time as it's cheap, showing that traditional lenders continue to be leverage-averse. VCs, meanwhile, have become much more active, implying that there is appetite for investment despite uncertainties. 4)The slack in the housing and commercial real estate markets still hasn't been worked off, meaning there remains a deep sink trapping capital and labor(mobility).

Regarding surveys, they do exist:
http://www.dailymarkets.com/stock/2011/05/31/businesses-growth-outlook-remains-positive-but-economic-sentiment-dampens-as-hiring-expectations-drop/

Their design may not be everything we might wish for in terms of answering questions about the political economy, but they do at least stay consistent and show variation over time on given topics. I suspect relatively good polls exist on or near management's perception of regulatory threat and wish they were easier to google*. However, I feel like these kinds of polls would have come to my attention (and probably have become political footballs) if they had shown really marked shifts.

*I definitely know polls exist on rankings of threats to business, because I've seen them many times over the years. The problem is finding directly-comparable polls, whatever their methodology.

[allenleephillips@gmail.com]

I've enjoyed reading your blogs, I put it on my list to follow.
http://allenthemetaphysician.com