Today I was looking for time series data that was not macroeconomic in nature. I wanted something fresh to keep my time series class engaged. There are only so many macroeconomic vector autoregressions one can do before your students begin to lose interest. To my delight, I stumbled upon a time series set of alcohol consumption per capita in the United States going back to 1934. The data is reproduced in the figure below:
I thought the figure might be interesting to others and posted it to twitter. Later, I got this reply from Geoff Turner:
I first chuckled at his comment, but then started looking closer at the figure too. Whereas Geoff Turner saw his favorite TV show in the data, I began to see macroeconomic history. I quickly began downloading data from FRED and messing around with these series and to my surprise discovered the true underlying cause of the Great Inflation was not the usual suspects. It was not the breakdown of Bretton Woods, misguided U.S. monetary policy, negative supply shocks, or demographics. Instead, it was per capita consumption of beer and spirits as seen in the figure below:
It is amazing how the World War II surge in inflation and the Great Inflation both track per capita consumption of beer and spirits. (Adding wine to the mix weakens the relationship.) And here is the scatterplot of this data:
Clearly we have uncovered the holy grail of central banking. All the Fed needs to do is raise per capita beer and spirit consumption and our low inflation problem will be solved. No need to raise the inflation target or adopt NGDPLT.
Of course, this finding raises an interesting question: if more beer and spirits consumption leads to more inflation, why is inflation so low in Germany? They are one the largest per capita consumers of beer. Maybe the answer is that their inflation would be even lower if were not for all their beer consumption.
P.S. This post is a joke of course and I am NOT promoting more alcohol consumption. Just having fun with spurious relationships.
David,
ReplyDeleteThis reminds me of some of those posts that argue that demographics was the reason for inflation. Time series econometrics is a very peculiar art.
Yichuan, see Mark's comments below on demographics. I concur it was a hard argument to swallow. Regarding time series, like anything it can be misused. Have you started doing VARs?
DeleteSince we are having fun with spurious relationships, it is worth observing that this finding implies Prohibition caused the Great Depression!! ( Andy Harless first noted this on Twitter.) Alcohol is not only a social lubricant it is an economic lubricant.
ReplyDelete"...It was not the breakdown of Bretton Woods, misguided U.S. monetary policy, negative supply shocks, or demographics. Instead, it was per capita consumption of beer and spirits as seen in the figure below:..."
ReplyDeleteI ran some Granger causality tests on CPI and the NIAAA data from 1934-2011 using the Toda and Yamamato method and I found no significant relationship between beer, wine, spirits or total alcohol consumption per capita and CPI. However I did find that the sum of beer and spirits consumption Granger causes CPI at the 10% significance level, so maybe you're on to something here. (Just kidding!)
The Interfluidity post you linked to led to a back and forth between Steve Randy Waldman and I on labor force and CPI. In particular these two posts:
http://www.interfluidity.com/v2/4624.html
http://www.interfluidity.com/v2/4706.html
Despite the fact that only three countries out of the 26 I looked at supported the kind of demographic story Steve was trying to tell his conclusion was still:
"Alas, I am not at all dissuaded from my view."
P.S. In Japan, Korea and Finland labor force growth and CPI inflation are positively correlated, and labor force Granger causes CPI. In the US, Greece and Iceland labor force growth and CPI inflation are positively correlated, but Granger causality runs the other way, namely from CPI to labor force. In Poland the correlation between labor force growth and inflation is negative, and labor force Granger causes CPI.
In five other countries there is a positive correlation between labor force growth and CPI inflation but no Granger causality, and in three other countries there is a negative correlation between labor force growth and CPI inflation but no Granger causality. In 11 countries there is no significant correlation between labor force growth and CPI inflation at all.
In short the results are decidedly mixed, although evidently not mixed enough to dissuade those with strong priors.
Lol!! ... as soon as I read this post my first thought was to ask David if beer and spirits consumption "Granger cause" inflation, and **of course** Sadowski already thought of that and ran the numbers! Ha...
ReplyDeleteMark and Tom, got a chuckle out of this. And thanks Mark for the points on the demographics. I never bought into it.
ReplyDeleteDavid and Mark, how about this one:
ReplyDeletehttps://twitter.com/gavyndavies/status/423114837263020032/photo/1
Have you seen that before? Another job for Granger?
The t-value is unbelievable! I wonder where Davies got the data?
DeleteThe "T-stat?" I skimmed Wiki on T-statistic, but I'm still not sure:
Delete1. Why are there two of these numbers? (28.4) and (5.5)
2. Good, bad, "unbelievable"... what's it mean?