Believe those who are seeking the truth. Doubt those who find it. Andre Gide


Saturday, May 23, 2009

Religiosity in Macroeconomics and the Sad Case of Father Wilder

A booming economy suddenly grinds to a halt. Stock prices are plummeting, wealth evaporating, investors fleeing to safe assets, consumer and investment spending contracting, deflationary forces appearing evident, banks and other businesses failing, employment falling, and unemployment rising. The forecast is for a long recession; possibly even a depression. What is to be done?

Gosh...I'm not really sure. Naturally, and like most people, I have some opinions on what sort of policy responses are more likely to work than others. My opinions on this question have evolved over the last 20 years. I interpret this process as learning (or unlearning, in some cases). I've learned a lot of new theories; theories that have opened my eyes to how any given set of observations might interpreted in wildly different, yet plausible, ways. I've also learned something about the history of similar (yet each in their own way, distinct) economic events. Nevertheless, I remain very ignorant; there is still much to learn. I expect that my opinions--or my belief system, if you prefer--will continue to evolve in the light of new theory and evidence.

Recently, I've become interested in the theory of religion. I'm not even sure yet how one might usefully define the term. I have something like the following in mind: "An unshakable belief in some preconceived notion." Formally, one might begin by saying that a person has some prior belief b that A is true; where b is between 0 and 1. This belief may evolve over time according to some learning algorithm, b' = B(b,e), where b' is a posterior belief, updated from a prior belief b, and new information e. I like to think of B(.) as Bayes' rule; but other learning algorithms are possible as well.

I define a person to be religious along dimension A if b=0 or b=1. Now, one interesting thing about Bayes' rule is that b' = b if b = 0 or b = 1. In other words, a Bayesian who begins with an absolute certain prior (one way or the other) will hold on to an unshakable belief (will not learn from any new information). We are all likely religious, in this sense, along at least some dimensions. If we are, then we might do well in recognizing the phenomenon when it occurs and at least be honest enough to own up to it.

On to some macroeconomics. In reply to the question posed above, imagine that someone proposes policy A; where A = a large fiscal stimulus policy is socially desirable in the event of an economic "crisis" as described above.

Now, if you would have asked my opinion on policy A several years ago, I think I might have assigned belief b = 3/4 that A is true. Given what I know today (which is still not very much), I would assign something closer to b = 1/4. In short, I would say that policy A is likely to fail. My good friend, Nick Rowe of Carleton University, expresses an opinion that is closer to b = 3/4. He believes that such a policy is likely to succeed. While we both have very different opinions, we each acknowledge that we really do not know for sure. That is, we are not religious on this matter.

Such differences of opinion are quite natural. The differences can likely be accounted for by the fact that both Nick and I have been exposed to different theories, have read different historical accounts of past crises, and have generally had different personal experiences that, together, have led our beliefs to follow different paths. But because we are not religious, and because we recognize this is so, we find it fruitful to persuade each other (and in the process, ourselves) on the validity of our beliefs. This is, I think, the best we can probably hope for in any science.

This all leads up to the topic of religion in macroeconomics. This field, more than any other it seems, appears to populated by true believers who like to view themselves as priests of some sort. One sure way to identify a priest is in how "exasperated" they become when failing to gain converts on the basis of what they know to be an "obvious" truth. Take Bruce (Father) Wilder for example. He knows that b = 1 when it comes to policy A.

Father Wilder, I have a confession to make.
What is it, son?
I'm not sure that hell exists.
What?! This is blasphemy!
So hell exists?
Most certainly!
And how do you know?
Poor, wretched and miserable child!
Um, yeah...OK. But really, how do you know?
It is obvious, you cursed flea-brain; the Bible says so!
Forgive me father, but there appear to be some inconsistencies in the Bible...
You are damned, child...you must endeavor to dismount the fortress of your conceit!
Maybe you can help me, Father...just a little bit of evidence...please?
Volcanoes.
Volcanoes?
The flames of Satan himself! Are you blind?
I'm not sure, Father...I have heard that volcanoes might be plausibly explained in some other manner.
How dare you question my authority!?
Sorry! I'm sorry I asked! Yes, I see it now...volcanoes. Got it.

I now direct you to Father Wilder and his volcanoes here.

Can fiscal stimulus spending increase aggregate output and employment? Yes.
Um, OK. But this was not the question. Even in the simple static labor-leisure model, output can be made to increase with an increase in G. We know this, you dope. We also know from history that conscripted labor can achieve the same result. But the associated increase in (measured) GDP is not necessarily welfare-improving (although it could be for certain individuals or sectors). What evidence can you bring to bear that a fiscal stimulus actually worked in a time of crisis?

Is there good evidence that this is the case? Yes. What is this good evidence? Well, pretty much the whole history of macroeconomic fluctuations in output and employment is evidence in favor of fiscal stimulus "working" in this positive "mechanical" sense.

If you say so, Father Wilder. Can you show me some evidence...please? Show me, so that I too may be saved from the flames of damnation!

Every year, employment increases with the seasonal rise in Xmas retail sales, and that fluctuation in output and employment is evidence that the economy works in a way that makes fiscal stimulus spending effective in increasing aggregate output and employment (when output is low relative to capacity and unemployment is high).

Volcanoes?! Father, I too am aware that GDP drops by about 90% every night; and there are frequently two consecutive periods (called weekends) in which GDP drops by a similar amount. I am also aware that seasonal fluctuations are huge. Having worked in the construction sector myself for many years, I know what cold weather can do to productivity. I am aware of the boom in expenditure that accompanies seasonal sales. I also know that these fluctuations can all be easily explained by a simple labor-leisure model with time-varying shocks to productivity and preferences. How is this relevant for understanding how fiscal stimulus is likely to mitigate what happens during an economic crisis? I hate repeating myself, but is there any evidence that you might bring to bear on this question?
Every business cycle provides plentiful evidence that rising aggregate spending drives increased output and employment. That kind of aggregate fluctuation kind of defines the business cycle. In the U.S. historical experience, two prime applications of fiscal stimulus as policy, inadvertently in WWII and deliberately during the Kennedy-Johnson Administration, certainly demonstrate increases in aggregate output and reductions in unemployment and chronic poverty. If you want to, you can find plenty of evidence in regional fluctuations as well -- the effect, for example, of the Reagan defense spending increases on Boston's Route 128 were pretty dramatic.

(The man must be hard of reading). Regional fluctuations? I already admitted the case in favor of helping out depressed sectors or regions. I am talking aggregate here. I am not talking about WW2 or the 1960s. I am talking about severe (non-war) economic crises, like the one we're currently experiencing, and like those experienced in history. (And the idea that aggregate spending "drives" the economy is an hypothesis, not a fact. You are confusing correlation with causation).

Let me try to help you out, Father Wilder. One legitimate question that could have been raised is why I think b = 1/4. One piece of data that led me to revise this belief downward was the case of Japan; see, in particular,



This is annual data from the Penn World Tables; and I've talked about it before here. The blue line measures real per capita GDP in Japan relative to that of the U.S. The red line measures the (G/Y) ratio in Japan, relative to that in the U.S. The blue and red lines are virtual mirror images of each other. In particular, note how late in the sample, the relatively expansionary fiscal stimulus in Japan is associated with a relative decline in their GDP.

Now, this is a rather interesting diagram, I think. Doesn't prove a damn thing of course. But it should lead one, I would think, to step back a moment and ponder the validity of b = 1.

Different interpretations are possible. One is that the data is lousy and hence conveys no useful information (so that prior beliefs may remain unchanged). I have some sympathy for this view. Another view is that the large fiscal stimulus in Japan may indeed have worked; in the sense that absent any stimulus, the blue line would have plummeted far more than it did. Yes, this is a possibility; although I wonder how many people might find it persuasive? The other interpretation, of course, is that the fiscal stimulus did not actually work at all; and in fact, likely made things much worse. This may or may not be true, but it certainly does not sound crazy to me. Certainly not crazy enough to deserve the pompous and pretentious sermons of an exasperated preacher trying to convince someone otherwise.

So, I am done asking the same question over and over again to Father Wilder. But I would like to direct this question to others. I am almost certain that one of you can come up with some data over some relevant time period for some crisis episode that might lead me to revise my belief in the other direction. I am very curious to see what the data looks like for the 1930s (a cross-country sample would be desirable).

Monday, May 18, 2009

Bruce Wilder on Andolfatto

Some time ago, I posed a question to those who were advocating "fiscal stimulus" in the form of increasing government spending as a weapon to combat a recession that typically follows a financial market panic. The question was: What evidence can you bring to bear that would persuade an agnostic that "fiscal stimulus" has "worked" in the past? ("Worked" in the sense of leading to a desirable increase in aggregate output and employment).

I have yet to receive what I consider to be a satisfactory answer to this question. I have, however, received a considerable number of responses that appear to be based on three basic principles hardwired into peoples' heads: [1] Econ 101 teaches us that increase in G leads to increase in Y; [2] Private markets are screwed up (and government agencies are not); and [3] Evidence that an increase in G leads to an increase in Y based on some wartime episodes.

In a nutshell, these people believe that: [1] leaves them with a firm theoretical foundation; that [2] legitimizes their belief in the principle of government intervention; and that [3] supports their view on the practical and desirable consequences of such intervention. In the language of climate alarmists, it is "obvious" that the "science is settled."

And what sort of response is in store for those who dare question [1]-[3]? Here is an example of Bruce Wilder comments on my earlier post. (Note: I have no idea who Bruce Wilder is, but I am pretty sure he cannot be an economist).
Libertarianism -- in all its mindless glory -- is often less a political philosophy than it is a personal conceit, a pose of intellectual superiority coupled with a demand that others assault one's own tower of skepticism, and, should one choose not to surrender, those assaulting will be found wanting. Much of what follows, in Andolfatto's posting, suggests that he can be as dense as a box of rocks. I just find him exasperating.

Now, I confess that I can be a rather slow learner. And maybe I am as dense as a box of rocks. But I am also being sincere in my question (and I have been known to change my mind in the past). Mr. Wilder's response to this question is basically this: How Dare You Ask Such a Question, You Mindless and Conceited Fool? I am reminded of the response by Brian's mother (Monty Python's Life of Brian) when she was asked whether she was a virgin or not.

Luckily, Andy Harless, a commenter plays the part of the voice of reason: "It seems to me that the burden of proof is on those who say that fiscal spending doesn't work. As you note, the visible evidence suggests that it works: the roads and bridges get built; there are many people looking for work, some of whom can presumably be employed to build those roads and bridges. If you want to argue about the general equilibrium effects, present a model in which those do erase the advantage of the stimulus, and someone else will present a model in which they don't. Then we can test the two empirically. Since, as your responders noted, there is econometric evidence which suggests that the government spending multiplier is greater than one, I'm guessing that the empirical analysis will tend to reject your model."

Ah, sweet reason. Somehow, I doubt that Harless will be judged to have penetrated the libertarian fortress of Andolfatto's conceit, at least by Andolfatto.

Actually, I thought that Andy Harless provided a thoughtful reply. I did, however, find his response wanting. According to Mr. Wilder, this should be construed as being attributable to the libertarian fortress of my conceit. This guy sounds like your common schoolyard bully (I suspect that he is more like the little Wizard of Oz).

My response to Andy Harless' reply is as follows:

[1] Yes, I did admit that there is ample evidence which suggests that fiscal stimulus "works" at the microeconomic or sectoral level. I even suggested that such a policy may desirable from the perspective of providing a form of social insurance to those who work in particularly hard hit sectors. But this is a far cry from saying that such a policy will work at the aggregate level. When the government makes an appropriation of funds from the private sector, this money is no longer available for private use. There has to be some crowding out. How much crowding out is an empirical question. So what is the evidence that a large fiscal stimulus will "work" in an economic crisis?

[2] Yes, there have been some econometric studies which purport to show that the multiplier on government spending is greater than one. But there are also econometric studies that suggest something closer to zero. In any case, no one is likely to be persuaded one way or the other by the results reported from running regressions on simple linear and reduced-form representations of historical data. People are much more likely to be persuaded by actual and dramatic historical episodes where such a policy can plausibly be interpreted as "working." Can you give me a few historical examples where this has been the case?

[3] What we do know from history is this. During the panics that afflicted the U.S. during the National Banking Era (1863-1913), recessions were typically short and severe; and the economy recovered without fiscal stimulus. This was also true during the Swedish financial crisis in the early 1990s. The only two dramatic examples I can think of where a recession prolonged into a depression involved massive fiscal stimulus; like the U.S. during the Great Depression, and Japan during its "Lost Decade." Is this the type of evidence that people have in mind in favor of fiscal stimulus?

Mr. Wilder: Please forgive me for remaining skeptical. I wish that what seems "obvious" to you might be made to seem obvious to me too. I ask that you persuade me, not by challenging my intellect or calling me names, but providing me with some evidence in support of your religion.

Tuesday, May 5, 2009

B.C. Election Campaign 2009

There is something very important happening in BC at the moment. The Vancouver Canucks are in a death struggle with the Chicago Blackhawks (series tied 1-1). Things are not looking good as the series moves to Chicago. Game 3 is on tonight. Oh, the drama!

There is also something less important happening in BC at the moment. We are in the midst of a provincial election campaign. I heard that the provincial party leaders are engaged in a series of debates. Naturally, I checked out the comedy network for the full schedule. No luck. But I did find a summary here of some recent "highlights."

The cast of characters are in the finest tradition of BC politics (they are all nuts). But I think that NDP leader Carole James really takes the cake. Here is a quote:
"We’re doing that (lowering taxes, increasing spending) within a fiscally responsible plan, a balanced budget in four years, and putting you and your family first. That’s what people are looking for in these tough times.”

That's right: she will put you and your family first. The presumption, I suppose, is that the other leaders are promising to put you and your family second.

Ms. James' platform reminds me of the fabulously funny radio campaign put on by my favourite London hangout Joe Cools in the 1980s: "Come to Joe Cools...where we'll screw the other guy and pass the savings on to you!"

Relative to Ms. James, Mr. Cools is to be commended along three dimensions. First, Mr. Cools understands the concept of feasibility (it is impossible to put everyone first; unless, of course, everyone is simultaneously put last). Second, Mr. Cools is honest. And third, Mr. Cools was actually trying to be funny. I vote for Joe Cools!

Now...back to the NHL playoffs...