Saturday, August 18, 2007

Business Contractions by Month-of-Decade

I went through the last 15 decades (actually starting in 1855, when the NBER dating starts, and ending in 2004 to give an integer number of total decades) and calculated the frequencies of business cycle contractions (using the NBER dates) for each month of the decade. For example, January of the “zero” year is the first month, February of the “zero” year is the second month,…,January of the “one” year is the 13th month,…,January of the “two” year is the 25th month, and so on. (You can click on the picture to see a larger version.) These results are purely empirical and may not mean anything, and I’m not sure how to do statistical tests on them, because I find it hard to make a rigorous distinction about what is a sensible hypothesis to test.



But there is at least one thing that seems hard to attribute to chance: the highest frequency of contractions is just after the beginning of the decade, and the lowest frequency of contractions is just before the end of the decade. Since 1855, there has been only one contraction that included the second quarter of a "nine" year (1949). Whereas in 10 out of 15 of the "zero" years, at least November has been a recession month. And at least 3 of the remaining 5 are not very impressive: November 1940 wasn’t a contraction, but the nation was still recovering from the Great Depression, so it wasn’t exactly business as usual either; November 1980 technically wasn’t a contraction, but it was part of a very short expansion sandwiched between two contractions; November 2000 wasn’t a contraction just yet, but as I recall, the winds of recession were in the air. (Maybe somebody who knows the history better than I do will have stories about 1880 and 1950.)

If we take this chart seriously as a prediction for the future, it has an interesting implication for the present time. There is a peak late in the “seven” year, after which the chance of contraction drops rapidly. That suggests that right now is a critical time to avoid recession, and if we can get through the next few quarters without one, we can expect smooth sailing for the remainder of the decade.

The chart may also be depressing for Democrats and heartening for Republicans. During the first few months of President Obama’s term, it will look like he inherited a strong economy. Then once the unimpeded Democratic initiatives are implemented, it will appear that they are destroying the economy. The trick, though, for a Democratic president, will be to reduce initial expectations and have people expect a longer-run payoff. That seemed to work for Reagan, anyhow. Alternatively, one could hope that the recession will be as shallow as possible and the recovery as strong as possible, which worked for Nixon. My advice to the next president is to figure out now a way to blame 2010 on Bush and then save your real economic ammunition for 2011 and 2012. With any luck, it could be the second Clinton boom. (As per tradition, Clinton will be facing a Republican congress.)

Labels: , , ,

5 Comments:

Blogger Dan said...

As for using the chart to predict the future and therefore make policy guidelines, surely you have not forgotten the Lucas Critique. These data were drawn over a time period in which monetary policy looked very different, and you have to figure that avoiding recessions in the economy is not like avoiding blizzards in Boston: if you make it through the treacherous months by going to Florida, warm weather will not necessarily await you in Boston.

Mon Aug 20, 03:40:00 PM EDT  
Blogger knzn said...

The applicability of the Lucas Critique would depend on what theory explains the pattern. It's certainly not obvious (to me anyhow) that there is something about past monetary policy (or other public policy) that could explain it. As far as the broad pattern of declining probabilities over the course of the decade, my first guess would be that it is explained by the psychology of "animal spirits". If that's the correct explanation, then the Lucas Critique wouldn't apply, since the psychology doesn't depend on policy, and it would then be reasonable to recommend that monetary policy should tighten during the "8" year and ease at the end of the "9" year to offset the expected changes in animal spirits.

Mon Aug 20, 05:23:00 PM EDT  
Blogger reason said...

Curious - a strong hint of political cycle (4 year), that restarts every decade - but why?

Wed Aug 22, 03:26:00 AM EDT  
Blogger mranonymous said...

Please do a significance test before drawing any conclusions. One possibility is a Chi-Squared test on the underlying counts.

Thu Aug 23, 06:22:00 AM EDT  
Blogger knzn said...

I think a chi-square test on the monthly counts would be invalid because consecutive months are so highly correlated. One can collapse them into annual counts by asking "Were there any recession months in year X?" (or in any of various other ways), and the annual counts are still not independent, but the dependence is fairly weak. I get an insignificant chi-square if I do it that way.

But suppose one wants to test a more specific hypothesis? I will tell you that, based on earlier research using a different data set (which began during the 1950s), my hypothesis going into this analysis was that there would be more recessions at the beginning of the decade. If I test, using a 2x2 chi-square, the hypothesis that the "zero" year has more recessions than others, it is borderline significant at the .05 level. If I test "first 2 years of the decade" vs. "all other years", it is unambiguously significant at the 5% level.

All of which leaves me still unsure whether the broad tendency is due to chance, but it seems to be at least worthy of further investigation.

Wed Aug 29, 01:03:00 PM EDT  

Post a Comment

<< Home