Interwar Unemployment in the UK and the US: Old and New Evidence

Abstract

Two contrasting views have dominated the research on unemployment during the interwar years. The conventional Keynesian view attributes the persistence of high unemployment in the UK and the US during the interwar period to sluggish adjustment of nominal wages to demand shocks. In contrast, equilibrium models of unemployment suggest that the natural rate is itself endogenous, determined by technological, institutional as well as demographic factors and is therefore not necessarily constant over time. According to this view, unemployment may remain elevated because some (or all) of the driving forces are persistent. How do we discriminate between these competing explanations? To this end, we estimate a time-varying parameter (TVP) model of the unemployment rate for the UK and the US. The Kalman filter estimates of the natural rate of unemployment suggest that most macroeconomic activity during the interwar period reflects persistent movements in steady state, not from steady state. We conclude that the observed persistence in unemployment appears to be consistent with multiple equilibria models and models with an endogenous natural rate.

JEL Classification: E24, J64

Keywords

Interwar unemployment persistence natural rate of unemployment Kalman filter

Introduction

Double digit unemployment devastated the UK economy for nearly 20 years and the US economy for a decade in the interwar period. The unemployment rate in both these countries reached levels that hitherto had never been experienced in either economy.³ As Figure 1 shows, in the turbulent years following World War I, the unemployment rate in the UK, which had averaged 4 per cent during the 20 years before World War I, rose dramatically averaging 10 per cent even during the expansions of the late 1920s and late 1930s.⁴ While the US also experienced a similar jump in the unemployment rate during the 1920s, it recovered rather better than the UK; the 1930s present a very different picture though. Between 1929 and 1933, the unemployment rate in the US soared to 25 per cent of the workforce. By 1939, employment and output remained well below their 1929 levels despite the best efforts of the New Deal.⁵

Why were unemployment rates so high for so long? In the traditional Keynesian model, persistence of high unemployment is the outcome of sluggish adjustment of nominal wages to aggregate demand shocks. If wages possess some degree of nominal inertia, then deflation will raise real wages and lower labour demand (Bernanke & Carey, 1996; Bordo, Erceg & Evans, 2000; Eichengreen & Sachs, 1985). Thus, according to this view, the high average unemployment experienced during the interwar period can be traced to a very slow speed of adjustment towards the steady state following an adverse shock rather than movements in the steady state itself. But why did the process of adjustment to nominal shocks appear to take so long in the interwar economies? As Cole and Ohanian (2002, 2004) have pointed out the slow pace of recovery is puzzling because economic fundamentals improved considerably in the UK in the mid-1920s and in the US after 1933. In the US, productivity growth was rapid, liquidity was plentiful, deflation was eliminated and the banking system was stabilized. With these fundamentals in place, the normal forces of supply, demand and competition should have produced a robust recovery.

Figure 1.

Source: Authors’ calculations.

The conventional view that the aggregate demand deficiency was the principal cause of unemployment in interwar Britain was first challenged by Benjamin and Kochin (1979) who argued that part of the explanation for the high level of unemployment was due to the generosity of unemployment benefits relative to wages. Their case rests on the assertion that both—the level of unemployment insurance benefits and the regulations governing eligibility for such benefits—were generous. Such benefits, therefore, reduced the costs of unemployment for the workers, making leisure less expensive as well as hardening the floor through which labour unions would not be prepared to allow wages for the employed to fall. According to this view, persistently high unemployment experienced during the interwar period can be traced to endogenous movement of the steady state brought about by changes in the labour market institutions.⁶

Identification of the causes of such unemployment persistence is crucial for public policy. If high unemployment is due to hysteresis (high-level of persistence) then shocks to demand would make a permanent dent in unemployment.⁷ On the other hand, emphasis should be placed on reforming institutions that alter labour market incentives for workers (i.e., minimum wage, the level and duration of unemployment benefits, taxes, social security payments etc.) if most of the rise in unemployment reflects movements in the natural rate.

This article focuses on the time series properties of unemployment during the interwar period to draw out some of the stylized facts any theory of the Great Depression would have to be taken into account. We do not estimate any particular model drawn from a well-defined economic theory. Instead, we focus on the time series properties of unemployment rates from a purely statistical point of view.⁸ Having said that, the results we find allow us to discriminate between these competing explanations. To anticipate our findings, our results suggest that most macroeconomic activity during the interwar period reflects movements in steady state, not from steady state. Indeed, movements in the natural rate account for most of the variation in the UK and US unemployment rates during the interwar period. We conclude that the observed persistence in unemployment appears to be consistent with multiple equilibria models and models with an endogenous natural rate.⁹

The second section of the article provides a brief overview of public policy in the UK and the US during the interwar period. The third section discusses the data, our econometric approach and our empirical findings. The fourth section concludes.

A Brief Overview of Public Policy in the UK and the US during the Interwar Period

In the period immediately following the World War I, the restoration of the gold standard was the primary objective of British public policy. It was widely believed that the dominance of British industry in the pre-war period and the pre-eminence of London as an international financial centre were attributable to the price stability engendered by the gold standard regime. The Cunliffe Committee recommended balanced government budgets and increase in the Bank Rate to check a foreign drain of gold in order to create the conditions necessary for the maintenance of an effective gold standard. The Bank Rate was raised from 5 per cent in November 1919 to 7 per cent in April 1920 and was held at that level for a year. Wholesale prices tumbled but unemployment soon rocketed from a low of 4 per cent in 1920 to over 20 per cent in 1921.

With the emergence of large-scale unemployment, the unemployment insurance scheme was substantially overhauled in the UK (Benjamin & Kochin, 1979; Cole & Ohanian, 2002; Loungani, 1991; Matthews, Minford & Naraidoo, 2008).¹⁰ In a series of process between 1920 and 1921, coverage was extended to most manual workers, weekly benefits for males were tripled, benefits were instituted for women, the number of contributions that had to be made before claiming benefits were reduced and the number of weeks for which benefits could be claimed were increased. The Unemployment Insurance Act of 1927 granted all insured workers, who had exhausted their standard benefits, the right to claim extended benefits for as long as they were unemployed. The only requirement was that claimants had to prove that they were genuinely seeking work but even this requirement was abolished by a 1930 Act. The relaxation of the insurance rules was also accompanied with a continuous increase in the average level of benefit.

Turning to the US, there was no centralized system of unemployment relief at the end of the World War I. Up until the 1930s, unemployment relief was local government based combined with private charitable organizations. The unprecedented unemployment and financial hardship that followed the ‘Great Depression’ brought forth a considerable expansion of relief activities under newly elected President Roosevelt. The Federal Emergency Relief Administration (FERA) was the centrepiece of the initial relief effort. Its principal aim was to put financial assistance directly into the hands of the unemployed through cash grants and work relief. Recipients of FERA were occasionally required to take part in public works programmes but it was far more commonplace for no work requirements were made. The ensuing expenditures on relief accounted for over two-thirds of the 300 per cent increase in federal spending from 1932 to 1940, with benefits going to a minimum of three million families each month (Wallis & Benjamin, 1981).

Other New Deal policies violated the most basic economic principles by suppressing competition and setting the prices and wages in many sectors well above their normal levels (see Cole & Ohanian, 2004 for further details). For example, the National Industrial Recovery Act (NIRA) permitted industries to collusively raise prices provided that they shared their new-found monopoly rents with the workers by substantially raising wages. While the artificially high wages created by the NIRA benefited the few that were fortunate to have a job in those industries, they significantly depressed production and employment, as the growth in wage costs far exceeded the productivity growth. When the NIRA was declared unconstitutional in 1935, these measures were continued (and strengthened) under the National Labor Relations Act (NLRA or the Wagner Act), which also gave more bargaining power to workers as compared to the benefits received under NIRA and encouraged faster growth in unions. The NLRA increased labour bargaining power further by permitting unions to use previously unaccepted tactics such as ‘sit-down’ strikes, in which strikers forcibly took over factories and prevented production. Using a Dynamic Stochastic General Equilibrium (DSGE) model, Cole and Ohanian (2004) found that the effect of New Deal sponsored ‘cartelization’ (including both the NIRA and the NLRA) prolonged the ‘depression’ by seven years.

Empirical Analysis

Data

The data for the UK (January 1887–October 1939) is based on two variables: Trade Union Members Unemployed (January 1887–December 1920) and Insured Workers Unemployed (January 1921–October 1939), the source being the National Bureau of Economic Research (NBER) Macrohistory Database. These data can be regarded as a single variable as they are based on returns collected by the Board of Trade and the Ministry of Labour from various trade unions which paid unemployment benefits (Matthews et al., 2008).

The data for the US (June 1906–June 1942) is based on two different variables (June 1906–January 1933; Unemployment of Trade Union Members UT/T, where UT denotes the number of trade union members unemployed and T is the number of trade union members) and (April 1929–June 1942); official unemployment rate U/L, where U is the number of unemployed and L is the size of the labour force). All figures are given in percentage terms. The source is from the NBER Macrohistory Database. To convert the first time series into the official unemployment rate U/L, Matthews et al. (2008) make use of the splicing factor U/UT, which is calculated using the data where the two time series overlap, namely, the period April 1929–January 1933. First, we convert the monthly time series UT/T into UT/L by multiplying it by T/L (Source: Table-I, Union Membership, 1897–1953; Bernstein, 1954), where only yearly data is available. We, therefore, make the assumption that union density is constant over the year. Then to obtain the splicing factor U/UT they take U/L from the second time series (over the period April 1929–January 1933) and divide it by UT/L. Finally, to obtain U/L they multiply UT/L by the splicing factor and hence convert the first time series into the second one.

Unit Roots in Unemployment

We begin, following standard practice, by testing both the series for stationarity. Augmented Dickey-Fuller (ADF) and Phillips and Perron (1988) unit root tests do not reject the null hypothesis of a unit root for both countries at 5 per cent significance level. However, it is well known that standard unit root test is biased towards non-rejection of the unit root hypothesis if the true data generating process includes breaks in its deterministic components (Perron, 1990). In response, a number of studies have developed different methodologies for endogenizing the break dates in the analysis of unit root (e.g., Zivot & Andrews, 1992). The Zivot and Andrews test (allowing for both a change in the intercept and trend) does not contradict the results obtained from conventional unit root tests, thereby providing evidence for the unit root version of hysteresis in unemployment.

Having said that, there is no economic reason for restricting the analysis to one break. Bai and Perron (1998) developed methods to test for and estimate multiple structural changes. We apply their methods to our non-stationary data.¹¹ Figures 2a and 2b plot the unemployment rates along with the mean unemployment rates before the first break and after each subsequent break for both countries. The UK unemployment data is characterized by five structural breaks (August 1895, July 1903, April 1912, December 1920 and March 1930) while the US data on the other hand has three breaks (August 1923, January 1931 and May 1936). This evidence is consistent with the results reported by Matthews et al. (2008) which suggest that the dynamics of unemployment in both countries during the interwar period exhibit multiple equilibria: one stable low unemployment equilibrium, one stable high unemployment equilibrium and one non-stable intermediate unemployment equilibrium that lies between the two.

Estimating the Time-varying Natural Rate of Unemployment

Our empirical strategy is to estimate the path of the time-varying natural rate (the mean unemployment rate). To this end, we estimate a reduced-form model for unemployment, treating unemployment rate as an observable variable and the intercept (δ) and persistence parameter (ρ) as unobserved time-varying state variables. We estimate the following ARMA(p,q) model:

Figure 2.

Source: Authors’ calculations.

u ​_{t} = δ ​_{t} + \sum_{i = 1}^{p} ρ ​_{i t} u ​_{t}^{}_{- i}^{} + ε ​_{t} + \sum_{j = 1}^{q} θ ​_{j} ε ​_{t}^{}_{- j}^{},

(1)

δ ​_{t} = δ ​_{t}^{}_{- 1}^{} + ξ ​_{t},

(2)

ρ_{i t} = ρ_{i t - 1} + η_{i t},

(3)

where u_t is the unemployment rate and the order-q moving average (MA) error term is motivated by time aggregation as well as other measurement error that could well introduce such a component.¹²

The first equation represents the measurement equation and the remaining two equations are transition equations. The disturbances ξ_t and η_it are serially uncorrelated disturbances with zero mean and constant variances and are assumed uncorrelated with each other in all time periods. These equations represent a state space form, in which the unknown parameters of the model, including the variance of ξ_t and η_it, can be estimated jointly by maximum likelihood estimation (MLE) using the Kalman filter algorithm. Provided with an estimate of the variance of ξ_t and η_it the time series of the parameters δ_t and ρ_it, can be obtained using the Kalman filter.

We use Kalman filter estimates to decide the best-fitting ARMA(p,q) model for each country. Starting with ARMA(1,0) for the measurement equation, we first raised the order of MA by one and then that of the AR by one, and so on upwards. We went up to order 2 for AR and order 12 for MA as the information criteria values increased steadily for higher order AR terms.¹³ We discarded those models where we found evidence of serial correlation in the error terms. Among all the specifications for which errors were not serially correlated the best-fitting ARMA model was chosen. As per this criterion, the best-fitting model for the UK and US turned out to be ARMA(1,4) and ARMA(2,2), respectively.

The estimated series ( ${\hat{δ}}_{t}$ ) and ( ${\hat{ρ}}_{i t}$ ) for UK and US is plotted as the solid line in Figures 3(a) and 3(b) respectively, along with two root mean-square error bands (95 per cent confidence interval). There are two general points to note. First, the estimates of the intercept and persistence parameter (sum of AR coefficients in the case of the US) show distinctive jumps: in the case of the UK in the beginning of the 1920s and the beginning of the 1930s, the US in the early 1930s. The identified break date for the UK appears to coincide with the evolution of labour market institutions; both the level of unemployment insurance benefits and the regulations governing eligibility for such benefits became much more generous during this period. These developments were either more muted or totally absent in the US in the 1920s. But policies under the New Deal such as FERA brought forth a considerable expansion of relief activities in the 1930s. Other New Deal policies such as NIRA which permitted industries to collude and NLRA which gave more bargaining power to the workers and encouraged faster growth in unions led to the growing labour market rigidity in the US. Second, and perhaps more importantly, the persistence estimates obtained conditionally on an intercept shift are clearly below one for both countries. Thus, the unit root hysteresis hypothesis can be resoundingly rejected in the case of both countries.

Figure 3.

Source: Authors’ calculations.

Figures 4(a) and 4(b) plot our estimates of the natural rate, $E (u) = (\frac{δ_{t}}{1 - \sum_{i = 1}^{p} ρ_{i t}}),$ alongside the actual unemployment rate for both countries.¹⁴ There are two general points to note. First, the moving natural rate model tracks the actual unemployment path quite well. The rise in the natural rate occurred in the UK in the beginning of the 1920s and the beginning of the 1930s and in the US in the early 1930s, about the same time as the possible regime change discussed above. However, the estimate of the top equilibrium rate of unemployment for the UK of 37 per cent in 1921 is harder to explain. In this regard, we would like to point out that results for the UK do differ qualitatively if one allows for additional MA components. Figure 5 plots our estimates of the natural rate (alongside the actual unemployment rate) based on ARMA(1,7) model. The model does a much better job in tracking the evolution of UK unemployment during the interwar period. The estimate of the equilibrium rate of unemployment in 1921 dropped to 27 per cent. Second, the estimated natural rate leads the actual unemployment rate over the business cycle, which is inconsistent with the hysteresis view. On the basis of our results, it appears that possibly a large part of the rise in unemployment in the UK and the US can be attributed to a rise in the natural rate.

Figure 4.

Source: Authors’ calculations.

Figure 5.

Source: Authors’ calculations.

Our findings are consistent with Benjamin and Kochin (1979) and Matthews et al. (2008) who also attributed the interwar unemployment experience to a steady rise in the natural rate. They also accord well with King and Morley (2007) who found that fluctuations in the natural rate explain the bulk of variation in the post-war US unemployment rate. Their estimates of the natural rate ranges between 1.8 per cent and 9.5 per cent over a 50-year period which suggest that the natural rate is quite volatile and support the idea that most macroeconomic activity reflects movements in long-run equilibrium, not from equilibrium. The findings seem to suggest that a further look at the determinants of the natural rate in interwar economies might be worthwhile. Part of the solution to the persistence puzzle may lie in these.

Conclusion

The persistently high level of unemployment experienced by the UK and the US during the interwar period greatly undermined the respectability of the classical theory and was an impelling force behind the development of Keynesian economics. Traditionally, economists after Keynes have attributed the persistently high unemployment to the sluggish adjustment of nominal wages to aggregate demand shocks. Although output prices fell, various economic, legal and institutional factors kept nominal wages from falling in tandem to restore full employment. The upshot was high involuntary unemployment, which remained high until reduced by the insatiable demands of a wartime economy in the early 1940s. In contrast, equilibrium theories attribute the persistence to movements in the natural rate. They suggest that several real shocks and the ensuing adjustments to them (a rise in overseas real interest rates, productivity shock, a rise in welfare entitlements, etc.) shift up the path of the moving natural rate. According to this view, unemployment may remain elevated because some (or all) of the driving forces are persisting.

The question arises how one can empirically discriminate between these competing explanations. To this end, we estimate a time-varying parameter (TVP) model of the unemployment rate for the UK and the US. The parameters of the model are estimated jointly by maximum likelihood estimation using the Kalman filter algorithm. When the moving natural rate model is tested against the alternative of a unit root process, the unit root hypothesis is resoundingly rejected. Specifically, our estimates of the natural rate suggest that most macroeconomic activity during the interwar period reflects persistent movements in steady state, not from steady state. The identified break dates appears to coincide with shifts in the natural rate fundamentals. The combination of these two factors is more consistent with endogenous natural rate models and models with multiple unemployment equilibria than with the traditional Keynesian explanation (where the natural rate is constant) or with the unit root hysteresis theories.

Footnotes

Notes

References

Alogoskoufis

, & Manning

(1988). Wage setting and unemployment persistence in Europe, Japan and the USA. European Economic Review, 32(23), 698–706.

Bai

, & Perron

(1998). Estimating and testing linear models with multiple structural changes. Econometrica, 66(1), 47–78.

Barro

R.J.

(1988). The persistence of unemployment. American Economic Review, 78(2), 32–37.

Benjamin

D.K.

, & Kochin

L.A.

(1979). Searching for an explanation of unemployment in interwar Britain. Journal of Political Economy, 87(3), 441–478.

Bernanke

Ben S.

, & Carey

(1996). Nominal wage stickiness and aggregate supply in the great depression. The Quarterly Journal of Economics, 111(3), 853–883.

Bernstein

(1954). The growth of American unions. American Economic Review, 44(3), 301–318.

Bianchi

, & Zoega

(1998). Unemployment persistence: Does the size of the shock matter? Journal of Applied Econometrics, 13(3), 283–304.

Blanchard

O. J.

, & Summers

L. H.

(1986). Hysteresis and the European unemployment problem. In Fischer

(Ed.), NBER macroeconomics annual, 1986 (pp. 15–78). Cambridge, MA: MIT Press.

Bordo

M. D.

, Erceg

C. J.

, & Evans

C. L.

(2000). Money, sticky wages, and the great depression. American Economic Review, 90(5), 1447–1463.

10.

Broadberry

(1986). Aggregate supply in interwar Britain. Economic Journal, 96(382), 467–481.

11.

Cole

H. L.

, & Ohanian

L. E.

(2002). The great U.K. depression: A puzzle and possible resolution. Review of Economic Dynamics, 5(1), 19–44.

12.

Cole

H. L.

, & Ohanian

L. E.

(2004). New deal policies and the persistence of the great depression: A general equilibrium analysis. Journal of Political Economy, 112(4), 779–816.

13.

Eichengreen

(1987). Unemployment in interwar Britain: Dole or doldrums? Oxford Economic Papers, 39(4), 597–623.

14.

Eichengreen

, & Sachs

(1985). Exchange rates and economic recovery in the 1930s. Journal of Economic History, 45, 925–946.

15.

King

T. B.

, & Morley

(2007). In search of the natural rate of unemployment. Journal of Monetary Economics, 54(2), 550–564.

16.

Layard

, & Nickell

S. J.

(1986). Unemployment in Britain. Economica, 53(Supplement), 121–170.

17.

Loungani

(1991). Structural unemployment and public policy in interwar Britain: A review essay. Journal of Monetary Economics, 28(1), 149–159.

18.

Matthews

, Minford

, & Naraidoo

(2008). Vicious and virtuous circles of unemployment: Explaining UK and US interwar unemployment. European Journal of Political Economy, 24(3), 605–614.

19.

Minford

A. P. L.

(1983). Labour market equilibrium in an open economy. Oxford Economic Papers, 35(Supplement): 207–244.

20.

Papell

D. H.

, Murray

C. J.

, & Ghiblawi

(2000). The structure of unemployment. Review of Economics and Statistics, 82(2), 309–315.

21.

Perron

(1990). Testing for a unit root in a time series with changing mean. Journal of Business and Economic Statistics, 8(2), 153–162.

22.

Phelps

Edmund S.

, & Zoega

Gylfi

. (1998). Natural-rate theory and OECD unemployment. Economic Journal, 108(448), 782–801.

23.

Phillips

P. C. B.

, & Perron

(1988). Testing for a unit root in time series regression. Biometrika, 75, 335–346.

24.

Wallis

John J.

, & Benjamin

Daniel K.

(1981). Public relief and private employment in the great depression. Journal of Economic History, 41(1), 97–102.

25.

Zivot

, & Andrews

D. W. K.

(1992). Further evidence on the great crash, the oil price shock and the unit-root hypothesis. Journal of Business and Economics Statistics, 10(3), 251–270.