Introduction
This background paper examines the main classes of labour market statistics, their uses and reliability. A case study is made of the interwar British labour market; this focuses on the causes and description of mass unemployment.

Development of labour market statistics
There are no continuous, reliable or comparable official statistics of unemployment in Britain before 1888, though there is sufficient evidence from which to calculate some sort of series from 1851. The absence of statistics is not particularly surprising since the prevailing ethos of the nineteenth century, which owed much to the teachings of the classical economists, was one which did not acknowledge a persistent imbalance between supply and demand in the labour market. Rather, unemployment was largely seen as a temporary and self-correcting phenomenon, the existence of which lay in the idleness and moral delinquency of the labour force. Indeed, the word unemployment, meaning involuntary idleness of the able-bodied, did not enter the currency of economists until Hobson’s definition of these terms in 1895 (Garside 1980, p. 1).

Thus it was not until the end of the nineteenth century that unemployment was being recognised as a major problem, and from this followed specific analysis and empirical investigation. Our first unemployment statistics, however, come from trade union returns (and derived from their activities in providing unemployment benefits for their members), not from official investigations, though trade unions did report their unemployment data to the Board of Trade. Indeed, broadly speaking, there are no government unemployment statistics before 1911, the year that saw the creation of the national insurance scheme covering health and unemployment. Thus, the first government unemployment statistics were, like those of the trade unions, the by-product of the administration of a social welfare programme; they were not the consequence of government interest in the pressure of demand in the labour market.

The unemployment insurance scheme initially only covered a limited number of industries. The scheme was then extended in 1916 and 1920, with further modifications during the interwar period. The detail need not concern you; the essential point is that the national insurance returns only really become useful from 1920, though there are still major problems associated with them.

The modern era of British labour statistics really begins in 1948 when the existing national insurance scheme, and thus its statistics, were greatly modified as part of the package of measures known as the ‘welfare state’ which was introduced by the postwar Attlee government. From 1948 until the early 1980s, when the current government began to make major changes in the statistics, we have a broadly comparable set of statistics.

The structure of the labour market
Labour statistics are best approached from the viewpoint of their being a market for labour, so that we can consider the supply of, and demand for, labour. We can also model the labour market in terms of stocks and flows.

First, the supply of labour. Our starting point is necessarily the population of the UK, from which we can derive the working population. Table BG.3A shows how this is constructed from Feinstein’s (1972) data for 1931. As you can see the population is roughly 46.1m, and is composed of 22.1m males and 24.0 females. The next stage is to calculate participation rates, the proportion of males and females who are available for paid employment outside of the home. This data is taken from Clark (1937, table 8), and is applied to the total numbers of males and females in the population, to produce the male and female labour forces. As can be seen the participation rates at this time were vastly different between males and females; since this time female rates have increased immensely. Finally, table BG.3A shows the distribution of the total working population between the employed and the unemployed.

 Table BG.3A UK working population and its distribution, 1931 (‘000s)
 

SOURCE: Clark (1937, table 8); Feinstein (1972, tables 55-6).

The demand for labour is another way of conceiving of the labour market. Clearly, the actual level of demand for labour is given by the total level of employment, with unemployment (the working population less employment) being a residual which can be considered as either excess supply or deficient demand.

We can also conceive of the labour market in terms of stocks and flows. By this we mean that if we take a base year with given levels of the working population, employment and unemployment, we can see how over time the working population is subject to flows (school leavers and retirees) which, depending upon the demand for labour, will affect the balance between those employed and those unemployed. Similarly, it is quite wrong to conceive of unemployment as a static mass; it too is subject to flows, so that the unemployment rate will be much affected by the duration of unemployment.

Labour market statistics

Main statistical series

The position so far can be summarised thus: economic historians need data from the following categories if they are to investigate the labour market:

• demographic: total population and distribution by age and sex, from which to calculate participation rates.
• census: data by occupation, whether employed or unemployed.

So far as employment is concerned, our main data classes cover total employment and its distribution by age, sex and the standard industrial classification (which details occupations and thus gives information on structural change).

Unemployment data, especially in the twentieth century, tends to yield more information. Thus we can investigate unemployment by:

• industry
• region
• age/sex
• duration

Problems of comparability

If we include the earliest trade union returns, we have unemployment data for Britain since the 1850s. This data, however, is not comparable over time. Consider table BG.3B, from which you might conclude that interwar unemployment was on average very nearly three times more severe than prewar and nearly eight times the magnitude of the early postwar period.

Table BG.3B UK average annual unemployment rates (%), 1881-1960
 

SOURCE: Booth and Glynn (1975, table 1).

In fact, the comparison is, to a large extent, misleading and gives an exaggerated impression of unemployment in the interwar period. For each of the three sub-periods the statistics quoted by Booth and Glynn (1975) are calculated on different bases.

As we noted earlier pre-1914 data was based on the returns of certain trade unions (principally in the engineering, wood-working and printing trades) which paid unemployment benefits to their members. There are three major limitations to the usefulness of this data.

1. The series represents a very small sample of the total working population: only about 300,000 workers in 1893 and 700,000 by 1910.
2. The figures are weighted on the basis of trade union membership covered in the returns, and not in proportion to the actual number of workers in the industries.
3. The sample is not representative; there are only a limited number of industries, only trade unionists are represented within these industries, and these men were more likely to be skilled and presumably less exposed to the danger of being unemployed.

On the basis of these problems, Beveridge (1944), one of the great unemployment statisticians, adjusted upwards the prewar data to make them more comparable with the interwar method of calculation, this yielding an average prewar figure of between 5 and 7%, which makes the unemployment problem of the interwar period appear less severe by comparison.

This is all very well, but the interwar data is itself flawed. In particular, the national insurance scheme, from which the data derives, only covered about 60% of the working population. Thus the scheme excluded those employed in agriculture, domestic service, national and local government, the railways and other public utilities, as well as all non-manual workers earning more than £250 per annum.

In such circumstances how representative is the interwar data, especially in comparison with the postwar period when the unemployment series cover the whole labour force, and which is our natural reference since for the most part it was a period of ‘full employment’. Table BG.3C shows the likely margins of error; that instead of 2.5m unemployed in Great Britain in April 1931 the real figure was closer to 3.2m. Feinstein was able to make such a calculation because of the existence of Census data for that month. On the basis of these differences, and much other detailed work, Feinstein has constructed series for employment and unemployment which cover the whole of the workforce.
 

 Table BG.3C Unemployment in 1931 (‘000s)
 

SOURCE: Feinstein (1972, p. 221).
 

 
It is now clear that relative to Feinstein’s series, the best data that we have, the national insurance data understates the total numbers unemployed but overstates the unemployment rate as those who were covered by the insurance scheme faced a greater risk of being unemployed than the labour force as whole. The two series produce the unemployment rates shown in figure BG.3A.

This leads to an important problem when we need to study the interwar labour market in detail; namely, that Feinstein’s series is only available on an annual basis; if we need monthly data, to look at turning-points, we have to use the national insurance series.

We can recalculate table BG.3B to take account of the problems of different coverage between sub-periods; the results are shown in table BG.3D. As can be seen the interwar period now appears in a rather more favourable light than prewar; the post-1945 figure also falls because Feinstein uses a different denominator from the official series record in table BG.3B.

 Table BG.3D Feinstein’s average unemployment rates, standardised bases, 1881-1960
 

SOURCE: Feinstein (1972, table 57).

Finally, we can chart the long-run series for unemployment and employment in figures BG.3B and BG.3C respectively.

 
Case study: interwar unemployment

Introduction

Having covered these basics about the range of labour market statistics available to us, and the differences in the coverage of data over the last century, let’s now look in more detail at an easily identifiable period in British economic history. The interwar period is particularly suitable, an age of mass unemployment, but also one which is very complicated to describe and analyse.

Interwar mass unemployment

Let’s begin by looking at the macro level, at the overall demand and supply of labour. We can then progress to a more micro level, a structural analysis, when we can investigate unemployment duration, regional distribution, etc. (all details from Middleton 1985a, ch. 2).

Some relevant data is charted in figure BG.3A, which details the two unemployment series, and in figures BG.3D and BG.3E which chart total unemployment and employment respectively.

 
From these figures it is immediately obvious that the interwar period was a very disturbed one, characterized by marked cyclical fluctuations in the labour market. These mirrored to a great, though not identical, extent fluctuations in the real economy, as can be seen from figure BG.3F which charts actual values of real GDP against its trend growth.

 
Interwar trade cycle history can be characterized as follows. The period began with an inflationary boom 1919-20, followed by a severe slump 1920-1. The next stage was a recovery with a minor peak in 1924, then more unsettled conditions before another recovery which reached its peak in 1929. This was the beginning of the world depression which affected the British economy until 1932. The recovery beginning in that year was strong and lasted until 1937; it was then followed by a short sharp recession over 1937-8 before massive rearmament expenditures induced another recovery in the year preceding the outbreak of the Second World War.

Looking at the employment data we can identify two features of note.

1. That it was not until as late as 1935 that the level of employment prevailing in the boom year of 1920 was again obtained.
2. That over the whole period 1920-38 total civilian employment grew by 7.4% but the total working population grew by a far greater extent (by 14%).

In such circumstances, it is not surprising that on trend unemployment (Feinstein’s series) grew from 1.4m to 1.5m in 1929 and 1.8m in 1937 (unemployment rates moving in line, at 7.2, 7.3 and 7.8% respectively). Now remember that these are trade cycle peaks; the minimum and maximum figures were 1.3m in 1927 and 3.4m in 1932, the latter the worst year of the depression.

A comparison with prewar trends provides a partial explanation of this upward shift in trend unemployment. First, changes in labour supply need to be taken into account. The rate of growth of the working population is a function of the rate of growth of the total population and changes in its age distribution and age- and sex-specific participation rates. Before the First World War the rate of growth of the working population had approximately equalled that of the growth of the total population. During the interwar period population growth declined quite sharply, so that if prewar trends had continued a fall in the rate of growth of the working population would have resulted. Such trends, however, did not continue, mainly because a fall in the birth rate increased the proportion of the population of working age and there was a marked increase in female participation rates. Therefore, had prewar trend in these demographic and socio-cultural variables continued, the interwar unemployment problem would have been less severe.

Although serious when viewed at the macro level, the interwar unemployment problem was made more acute by its uneven incidence as between regions and industries. Indeed, this is increasingly seen as its most important feature. These disparities were a vivid manifestation of the serious structural problem confronting the interwar British economy: the secular decline of the staple industries. The regional problem resulted from their location, a strong concentration in only four regions: north-east and north-west England, south Wales and lowland Scotland.

Regional employment trends are summarised in tables BG.3E and BG.3F. The regional disparities are immediately apparent from both the figures for unemployment and employment growth, disparities which widened over time. These trends reflected both the high levels of unemployment in the declining staple industries (concentrated in the outer regions) and the pattern of location of the new, expanding industries of the period, one strongly directed towards regions with already lower than average unemployment (inner Britain).

 Table BG.3E Regional unemployment rates, insured persons aged 16-64, Ministry of Labour divisions, selected years 1923-37
 

NOTE:(a)The administrative divisions were rearranged in 1936, thereby precluding exact comparison between the 1937 figures and earlier years.
SOURCE: Middleton (1985a, table 2.1).
 

 Table BG3.F Regional employment growth, insured persons aged 16-64, Ministry of Labour divisions, selected years 1923-37
 

NOTE:(a)The administrative divisions were rearranged in 1936, thereby precluding exact comparison between the 1937 figures and earlier years.
SOURCE: Middleton (1985a, table 2.1).

The regional problem thus assumed two dimensions: the very high rates of unemployment of the industries centred on these depressed regions and their below average rates of employment growth. In addition, cyclical factors were also of importance in contributing to the regions’ problems. The outer regions were heavily committed to producer goods industries which, as with investment in GDP, experienced the most pronounced cyclical fluctuations in output and investment. By contrast, the more prosperous regions of inner Britain had more diversified industrial bases - in particular a growth of consumer goods industries - and this lessened their susceptibility to the worst effects of cyclical depression.

The next stage in cataloguing the nature of the interwar unemployment problem is to conceive of it in terms of stocks and flows. It is axiomatic that the actual numbers unemployed rises or falls as the flow of newly unemployed people coming onto the register exceeds or falls short of the flow of people leaving the register to take up employment; thus the duration of unemployment is as important an indicator of labour market conditions as the absolute numbers unemployed.

Investigated in this light, it appears that the true significance of Britain’s interwar unemployment problem lay not only in the high absolute levels of unemployment in certain depressed regions but in the prolonged nature of this unemployment and its worsening as the period progressed.

Although no data are available before 1929, Ministry of Labour figures collected thereafter show that the average duration of unemployment rose significantly over the 1930s; while other evidence suggests that this deterioration was in fact evident from the early 1920s onwards. In September 1929, 89.1% of the total had been unemployed for less than 6 months (short-period unemployment) and only 4.7% for more than 12 months (long-period unemployment). By August 1937, however, short-period unemployment stood at 65.7% of the total while long-period unemployment had grown to 24.3%.

The policy implications of this long-period unemployment problem stemmed from its regional nature. This is very clear from table BG.3G, where, for example, long-period unemployment was almost six times higher in Wales than it was in London in 1937. The astronomical rates of unemployment recorded in certain towns at the depth of the depression (January 1933) are well known - for example, 91% in Saltburn and 77% in Jarrow. Less well known, and infrequently cited, was the accompanying long-period unemployment problem for many depressed area towns. This was vividly illustrated in a survey of November 1936 which found that, even after four years of strong national recovery, 71% of the unemployed in Crook and 45% in the Rhondda had been unemployed for five years or more, as compared with only 3% in the prosperous southern town of Deptford.

 Table BG.3G Duration of unemployment among applicants for benefits and allowances, all persons aged 16-64, Ministry of Labour divisions, 21 June 1937
 

SOURCE: Beveridge (1944, table 10).

Causes of interwar unemployment

Even from the limited statistical evidence we have discussed here (limited in that we have not looked at, for example, unemployment by age or sex or the differing experience of other economies), it should by now be clear that interwar unemployment was neither monolithic, nor monocausal. With a labour market subject to such extremes, between north and south, between different industries, there was no uniform experience.

Nonetheless, we can some further progress by applying a taxonomy of unemployment. A typical Keynesian method would be to distinguish between certain types of unemployment (see Sinclair 1987):

• frictional
• seasonal
• cyclical
• structural

The first two categories are deemed to be temporary unemployment due to frictions in the labour market (asymmetric information, seasonality), which are not much connected to the overall level of demand for labour. Cyclical unemployment, however, is deemed to be a function of the pressure of demand; while structural unemployment relates to shifts in the direction of demand (for example, the high unemployment rates of the interwar coal industry can be related to the shift in demand from coal to oil).

Such a taxonomy obviously gives insights into the causes of unemployment, and the possible efficacy of the Keynesian solution. There are four main contenders for the cause of interwar unemployment:

• deficient aggregate demand
• too high real wages
• benefit-induced unemployment
• technological progress

Conclusions

The golden rules of labour market statistics can be summarised thus:

1. Whenever you see an unemployment figure, or a time series, don’t take it at face value. Consider the constituent elements of the numerator and denominator; have they changed over time and are they satisfactory.
2. Do changes in unemployment seem to be coming from the demand or supply side of labour, and what are their likely determinants.

The general problems are well-expressed by Sinclair (1988, p. 1) in his excellent textbook on unemployment:

Unemployment is like an elephant: easier to recognise than to define. Definitions abound. Practices differ between countries. They are apt to change within countries, too; politicians beset by a sharp rise in unemployment sometimes yield to the temptation to redefine their problem away.

 
 
 
 
To IT-MA home page
To Department of Historical Studies home page.

(c)R. Middleton 1997. Last modified Monday 30 June 1998.