Apropos of the recent discussion of unemployment insurance in the context of demand shortages, it’s worth mentioning Daniel K. Benjamin’s and Levis A. Kochin’s “Searching for an Explanation of Unemployment in Interwar Britain.” During the 1920s, despite strong real growth in national income and wages, Britain was plagued by persistently high unemployment. Between 1921–1938, British unemployment never fell below 9.7 percent, and only in one year (1927) did it fall below 10 percent. At the time, there were several economists who looked at the incentive effects of unemployment insurance, but this explanation was soon drowned out by Keynes’ theory of an underemployment equilibrium. Benjamin’s and Kochin’s research, although not without controversy (as is usually the case in economics), shows that unemployment insurance (UI) did play a large role in increasing the British unemployment rate.
My intention is not to mislead, and it should be said that the situation between 1922–1938 is not necessarily comparable to that of 2007–present, although the longer high unemployment persists against the backdrop of positive income growth, the more similar the two situations become. Also, Benjamin and Kochin include Great Depression years in their research, and they find that unemployment insurance does explain a good chunk of the unemployment rate, although it cannot explain all of it (or even most of it) — this is where monetary disequilibrium explanations come in. Further, I coincidentally became aware of this paper recently (while reading David Glasner’s book on free banking), and I thought that it’s interesting enough to share.
Between 1920–21, like in the United States, Britain underwent a sharp period of deflation, as the monetary authorities attempted to arrest the wartime inflation. Unemployment rose from 3.9 to 17 percent, but this was to be expected — actually, the surprise was how much the recession was represented by the overall decline in prices, rather than a decline in output. However, despite a recovery and strong growth after 1923–24, unemployment remained inexplicably high. Between 1930–31, growing unemployment is explained, in large part, by the Great Depression and its demand-side causes. However, again, as the British economy began to recover after 1932, unemployment rates failed to fall below 15 percent until 1935, and they remained above 10 percent by 1938. By the way, between roughly 1924–28 and 1932–38, this high unemployment is in the context of demand management by central banks.
This situation was an empirical influence on J.M. Keynes. Originally, Keynes advocates the standard interpretation of unemployment, which was that unless nominal wages were adjusted, unemployment would persist until real wages fell to their market clearing level. But, this theory couldn’t make much sense of the persistently high unemployment of the 1920s and 1930s. Indeed, during much of the ’20s, real incomes and wages were growing, and the situation was similar after 1932. Thus Keynes developed his theory of an underemployment equilibrium, which would form part of his General Theory.
But, there is an alternative explanation — one that several economists, including Jacques Rueff and Edwin Cannan, proposed, but was drowned out by the Keynesian revolution. In 1911, the Unemployment Insurance Act was passed in Britain, offering UI to about 15 percent of the workforce. In 1920, these benefits were expanded by about 40 percent (I’m assuming of their 1919 level). Despite the deflation of 1920–21, the nominal value of these benefits was not decreased, and was actually increased a number of times. Benjamin and Kochin write that, by 1931, weekly UI was 50 percent of weekly average wages. In other words, the real value of British UI rose significantly during the 1920s.
After running a regression, Benjamin and Kochin make the following estimations (taken from p. 467),
Apart from their original regression results, Benjamin and Kochin run two additional tests. First, they look at juvenile (years 16–17) unemployment rates during the same period. Juvenile workers were typically eligible for less benefits than older workers, so the incentive theory predicts that their unemployment rate should be lower than that of older workers. They find that the data supports this conjecture, and that alternative explanations are not strong enough to account for all of the evidence. Second, look at the unemployment rates between married women and men after 1931–32. During the 1920s, married women were eligible for benefits if they were let go or if they voluntarily left their jobs; thus, married women would complement their husband’s incomes by receiving UI. In 1931, the Anomalies Regulations were implemented, blocking many married woman from receiving these benefits. Following this piece of legislation, the female unemployment rate fell from 18 to 13.6 percent, while the male unemployment rate increases from 21 to 25.4 percent.
What does this evidence say in the context of Paul Krugman’s recent defense of UI during demand shortages? (I question his theory, here.) For those of us who are skeptical of demand-explanations and like to point out supply-side distortions, the evidence probably says less than what one might think. For those of us on the opposite side of the spectrum, I think the evidence says quite a bit. Before anything else is said, there are important differences in the details of what exactly UI entails. During interwar Britain, an unemployed person could draw on UI indefinitely. That is not the case in contemporary America. Further, one should question whether the level of real benefits between interwar Britain and the current U.S. are comparable. What this means is that the estimates Benjamin and Kochin come up for interwar Britain are not applicable to our current situation.
Also, I wonder if their estimates for the Great Depression are biased upwards. Multiple regressions are supposed to reduce bias by increasing the number of explanatory variables, where the coefficient for each variable is taken net of shared variation. This idea might be better explained by means of a diagram,
UI is short for unemployment insurance; DS is short for demand shortage. The unnamed circle is the Y variable, or unemployment. What a multiple linear regression does is reduce the amount of bias in a model by considering, and eliminating (from the coefficient estimate), shared variation — on the right hand side, that is the space jointly shared by UI and DS that overlaps with the dependent (Y) variable. When the real economy is growing and there likely isn’t a demand shortage, the amount of shared variation should fall. This is represented by the diagram on the left, where UI and DS have no shared variation at all. However, during the Great Depression, and whatever other period of demand shortages, the amount of shared variation will be relatively high. The estimates provided by the regression are an average over the period of years covered by the time series data. There were many more years of growth than there were of contraction, so the average will better capture these periods than they will depression years. In other words, there might be a case for a downward revision of the estimates of changes in the unemployment rate for depression years.
Nevertheless, Benjamin and Kochin provide good evidence that could be used to argue against extending UI, even during depressions (and recessions). First, even assuming that we should revise their estimates down (during specific years), Krugman’s theory predicts a net positive employment effect; the evidence provided here argues against that. Second, there is the consideration that governments will not necessarily change UI benefit schemes during periods of growth, meaning that we should also consider future labor market inefficiencies when judging the appeal of increasing UI during recessions. Admittedly, however, we live in a period characterized by steady inflation, rather than price level stability or slight deflation — this means, sans nominal increases in the value of UI benefits, the real value of these benefits will fall over time. Third, Benjamin’s and Kochin’s study should warn us against UI increases in the context of “secular stagnation.” Unemployment may remain high in periods of growth not because of “secular stagnation,” but because of high real unemployment benefits. Fourth, it has been ~6 years since the beginning of the “Great Recession,” and demand-side theories of unemployment are losing some of their attractiveness. This gives credence to supply-side theories of current unemployment.
My opinion is that there are both supply- and demand-side factors that explain current unemployment levels, and that both are important. I don’t think increasing or extending UI is good economic policy, although it is reasonable to justify it under moral or ethical considerations. There are better alternative solutions to demand-shortages. But, the point I really want to make is, just like supply-siders shouldn’t underestimate demand-side factors, demand-siders shouldn’t underestimate supply-side factors. Potentially, supply-side explanations for unemployment can explain a larger chunk than demand-side economists may initially suspect.