Demographics and inflation: international graphs
Mark Sadowski points out that that the price level in the Penn World Tables represent an attempt at a global price level, rather than a country-specific domestic index, based upon the price of US GDP. Thus changes in the real exchange rate between a country and the US show up as changes in the price level, but don’t always represent “inflation” or “deflation” within the local economy. Combining these flawed price levels with what we already knew to be a flawed proxy for labor force, I think the graphs below are too noisy to be meaningful. This renders much of my discussion, which was based on the flawed graphs, essentially bullshit.
At Sadowski’s suggestion, I’ve regenerated graphs where I could based entirely on data from the OECD’s Main Economic Indicators database, as hosted by FRED. (Sadowski suggests AMECO data for CPI, but I’ve not had a chance to look at that yet. I’ve used the OECD CPI series.) The OECD database (at least as hosted by FRED) is not complete. There were surprising omissions. In particular, the lack of UK and Austrian Civilian Labor Force series prevented me from reproducing those graphs from the original post.
I’ve generated graphs for a larger set of countries that I did originally. (I basically generated graphs for all countries for which both OECD data series were reasonably complete.) In an Appendix to this post, I’ll first place the six graphs from the original piece I could reproduce, and then all the rest. I’ll defer any discussion to a later post. I’m going to strike through the original text, and add cautionary watermarks to the flawed graphs, in order to emphasize the essentially bullshitty nature of the original discussion.
The two previous posts on demographics and the Great Inflation remain very much not bullshit. They are…→ Not a monetary phenomenon
→ Agreeing in different languages
See also a whole lot of excellent discussion of this topic by others, links are here.
Mark Sadowski (in comments here and here) and Scott Sumner challenge me to support my thesis that inflation is related to demographics with international data. Doing that right would be hard — inflation correlates across borders, as did World War II and its effect on postwar fertility. Even if international data appeared to support my hypothesis perfectly, one could argue that the whole thing is just a correlated coincidence. I’m not going to make a project of teasing out causality econometrically. But the least I can do is show you some graphs.
Graphs in economics are always Rorschach Tests. They are invitations to confirmation bias. But I’m going to go out on a limb and say there’s some evidence for my thesis in these pictures. You, of course, will decide for yourself.
You will be looking at data from Penn World Tables version 8.0, which I am supposed to cite as
Feenstra, Robert C., Robert Inklaar and Marcel P. Timmer (2013), “The Next Generation of the Penn World Table” available for download at www.ggdc.net/pwt
This was a data source of convenience. It’s designed for international comparisons, and includes population and price-level data beginning in 1950. I have graphed 5-year trailing growth rates of the price level (i.e. 5-year compounded average inflation rates) against 20-year lagged population growth rates, also trailing 5-year averages. Lagged population growth is an imperfect proxy for what I really want, which is contemporaneous labor force growth. It misses entirely the effect women’s entry in the workforce, which I think is an important part of the story, and ignores variations in the rate of labor-force exit. Nevertheless, it is the best I could (easily) do.
There’s nothing special about the 5-year lookback window or the 20-year lag. I haven’t data-mined them. In my initial post I chose a 10-year window, and it seemed to work. This time I tried a 5-year window for the hell of it. It’s still fine. Note that growth rates are reported in gross terms, that is “1.1” refers to a 10% growth rate, not a 110%.
Without further ado, here is my base case, the United States:
Unsurprisingly, as this is the economy from which I drew the thesis, it matches up pretty well.
Let’s try Australia next:
Again, to me, the smoothed inflation rate and population growth level follow one another during the disinflation rather strikingly. The inflation changes seem to lead the population growth rate changes a bit, which is somewhat surprising. But the population growth is 20-year lagged. Inflation in 1970 isn’t likely to be causing changes in the 1950s birthrate. Here, as in many of the graphs, I’m going to write off this lead of price changes off as due to some combination of an ill-chosen lag, missing effects of workforce composition changes, and forward-looking policymakers anticipating near-future entrants to the labor force. Reasonable or confirmation bias? You be the judge!
Next let’s look at Sweden.
Sweden is one of four countries I looked at that Sadowski cited as a counterexample. But once you look at time series rather than individual data points, I think it’s okay. Again, the price level leads a bit, but otherwise the two series track one another fairly well.
The UK is the first country that contains a really interesting anomaly:
Note the downspike in inflation in the mid-1980s, entirely unmatched by any fall in population growth. I think this anomaly is very easy to explain, and very informative. Concomitant with the Volcker disinflation in the US, the Bank of England allowed interest rates to climb very high. Sure enough, tight money, killed inflation! But inflation seems to be very zombie-like when the labor force is rapidly growing! You can kill it, but unless you painfully hold it down, it comes right back. That was the US experience in the late 1960s and 1970’s: each attempt to tighten money (reflected in high interest rates) provoked a sharp bout of unemployment, until the Fed cried uncle. Was the problem insufficient grit and determination on the part of the Arthur Burns Fed? Or did Volcker get lucky, in that his tightening cycle happened to coincide with the peak growth rate of the US labor force? Maybe a bit of both? (I know we love to fight, but most hypotheses are not mutually exclusive!)
Anyway, the UK experience was very clear. There was a lot of grit and determination in the early 1980s, it was temporarily effective, but it didn’t take. The inflation zombie rose again. It was slain for good when, perhaps coincidentally, the lagged population growth rate collapsed.
How about Japan:
I don’t know what to say about Japan. In a general sense, the two series track. Population growth collapsed and the economy went into deflation. But the two series certainly don’t wiggle together in any obvious way. Totally a Rorschach Test.
Next let’s look at Austria.
Again, you can call this one either way. On the one hand, Austria’s disinflation corresponded very well to a collapse of putative workforce growth. But the initial inflation did not. I could wave my hands about commodity prices and the Great Inflation being a global phenomenon, so that explains the early start. But you know, confirmation bias. You decide, I’ll call it a draw.
France graphs like the surprisingly sexy love child of the UK and Austria:
Like Austria, France suffered from embarrassing premature inflation. It did the job, though! Babies were conceived and born in the 20-year-lagged time machine, and eventually there was a nice boom of workforce entrants, timed to match the Volcker disinflation in the 1980s. As in the UK, there was a satisfying downspike then a zombie-like return of price level growth. Inflation then fell roughly — very roughly! — coincident with the decline in population growth. Like Austria, the nadir of the two series match very well but any comeandering of the rest is less clear. There is something interesting in the tail of the France graph. Post-2000, inflation rose sharply, in a manner that doesn’t seem at all proportionate to population growth. We’ll see there was a similar spike in Spain. (Also in Austria, but that did track population growth.) One explanation for these spikes would be post-Euro capital flows: Price level growth is often attributed to foreign lending, and both France and Spain were current account deficit nations. (Do foreign capital flows fall under the definition of a “monetary phenomenon”? You tell me!)
So, Spain. Let’s do Spain:
Of the countries I examined, Spain was the one that least fit my story. Spain disinflated before population growth collapsed, and reinflated just when it did. So Spain is a legitimate counterexample. I have my stories: There may have been unusual political will towards disinflation in the high population growth 1990s Spain, so the country could qualify for Euro entry. Once the Spain entered the Euro, it was a prime destination for speculative lending. The timing of these events was completely decoupled from Spain’s population dynamics.
But, confirmation bias is a real danger. I have to concede that, on face, Spain’s inflation dynamics looks nothing like I would expect from my demographic hypothesis.
Overall, I think my claim that labor force expansion imparts an inflationary bias is supported by these graphs. The Great Inflation and its unwind, I score four clear “wins”, three iffy cases, and one clear loss. These were the only countries I examined. (I’m not cherry picking from a larger group.)
Throughout the graphs, there are some recurring themes. Inflation downspikes in the 1980s are nearly universal, but they “take” as persistent disinflation only when ratified by falling labor force growth. The relationship between disinflation and slowing labor force growth seems more precise than the relationship between labor force growth and the initial inflation. Current account deficit countries (including the UK and Australia and Eurozone countries, but not the US) tended to experience inflation unrelated to population growth in the period following Y2K.
I think demographics has had a very great deal to do with inflation dynamics. My cards are on the table. Obviously this is imperfect evidence. So what do you think?
Appendix: Please see the update at the top of this post. The spreadsheet from which these new graphs were generated is here. These are graphs of 5-yr trailing growth rates, annualized. They are presented as gross rates, so that 1.2 means a 20% growth rate in more conventional terms.
Graphs from the original post, redone (corrected!):
[UK omitted, missing data]
[Austria omitted, missing data]
- 8-Sept-2013, 8:30 a.m. PDT: Corrected name (twice) “
- 9-Sept-2013, 1:40 a.m. PDT: Retraction via bold update at beginning of piece + strike-through of original, recomputed graphs with more appropriate data in the Appendix.
- 9-Sept-2013, 4:15 a.m. PDT: Added “BAD” watermarks to the original graphs.
- 11-Sept-2013, 10:00 p.m. PDT: “