Printer Friendly

Poverty and the welfare state in interwar London.

1. Introduction

The measurement of working class poverty and its relationship to social security was a growing preoccupation in the decades before the Second World War and it has been widely discussed by historians. This debate was greatly influenced by the results and interpretation of a variety of poverty surveys which, beginning with Charles Booth in the 1890s, and followed by Rowntree, Bowley, and others, measured working class incomes and compared the results with a scale of minimum needs. These studies provided objective evidence of the extent and causes of poverty and provided benchmarks for the evaluation of welfare state provision. Such comparisons were widely discussed during the interwar period and they influenced the welfare state provisions made after the Second World War following Beveridge's famous Report on Social Insurance and Allied Services of 1942.

There is a large historical literature on the changing concepts of poverty, the methods used to measure it, and the results obtained by contemporary investigators. A still larger literature is devoted to the evolution of the welfare state and it includes assessments of the place of poverty surveys in that development. Such studies have often, in the light of subsequent experience, explicitly or implicitly questioned the appropriateness of the definitions of poverty used by contemporary investigators. But, until recently, it has not been possible to assess comprehensively the effects on the scale and composition of poverty of changing these definitions. For most of the poverty surveys, such a re-evaluation is not possible except on the basis of published tabulations because the original records do not survive. The recent computerisation of the records from the largest of the interwar surveys, the New Survey of London Life and Labour (NSLLL), for which almost complete records survive, makes such a re-evaluation of poverty measurement possible.

In this paper we make use of this unique survey to address three questions. First, what difference does it make to the proportion and the composition of those judged to be in poverty when different assumptions are made about the poverty line? Our comparisons involve both variations in the absolute level of the poverty line and changes in the basis on which different types of households are compared. Second, to what extent did the interwar social security system effectively alleviate poverty? The interwar system has often been overshadowed by the post-Beveridge reforms and its failure to meet the needs of the poor has often been assumed rather than analysed in detail. This leads to the third question: what difference would it have made to the numbers in poverty if the post-Beveridge reforms had been introduced 20 years earlier? Our results suggest that such measures would have reduced the numbers in poverty by up to a half but would not have eliminated it.

2. Social surveys and the welfare system

Surveys of working class poverty reached their peak of activity and influence in the interwar period. Following the lead of the pre-war pioneers, Booth, Rowntree, and Bowley, numerous surveys were undertaken at different times, for different towns and using different criteria to measure the percentage of households or individuals who were classified as poor, and to examine the causes of their poverty. Some, such as the New Survey of London Life and Labour (1929-31) and Rowntree's second social survey of York (1936), sought to make comparisons with earlier surveys of the same area. Others compared their findings with contemporary findings for other towns. Such comparisons were, of course, affected by differences in the industrial character of the towns surveyed, the date of the survey, and the criterion used for determining poverty - the poverty line.

The rates of poverty found in some of the more important and influential interwar surveys are reported in Table 1. The surveys were confined to working class households but poverty rates can be calculated for all households under the assumption that middle class households would not have been poor. The results indicate that the percentages found to be in poverty increased somewhat in the 1930s when unemployment was rather higher than it had been in the 1920s.(1) But the poverty lines used in Tout's survey of Bristol and especially in Rowntree's survey of York were somewhat higher than those used for the five towns surveyed by Bowley and Hogg (1925) and the surveys for London and Merseyside.(2) Rowntree's 'human needs' poverty line was a deliberate upward revision of the line he [TABULAR DATA FOR TABLE 1 OMITTED] applied in 1899 (see Rowntree 1937, 1941). However, Rowntree's lower poverty line, Class A, also used in the 1936 survey, is rather closer to the others and it gives a proportion in poverty similar to those for London and Bristol. Recently, Linsley and Linsley (1993) re-examined poverty in London using a one in 20 sample of the records from the New Survey of London Life and Labour by applying Rowntree's human needs poverty line adjusted to prices in London in 1930. They found that on this basis the proportion in poverty was 21%, somewhat closer to, but still lower than, Rowntree's finding for York in 1936. It seems reasonable to conclude that the proportion in poverty was typically 10% on the lower poverty lines, 20% on the higher poverty lines, and somewhat greater than otherwise for the towns hardest hit by unemployment in the 1930s.

The results of these studies and others like them were widely cited in the interwar debate over income maintenance standards and the welfare system. Those found to be in poverty were chiefly children, old people, and the families of the unemployed. Such findings were used by different pressure groups to argue for family allowances and for increases in pensions and unemployment benefits. Pressure groups such as the Children's Minimum Council, the National Conference on Old Age Pensions, and the National Unemployed Workers' Movement campaigned at different times and with different degrees of success for increases (or against decreases) in benefits (Macnicol, 1980; Blaikie, 1990; Stevenson and Cook, 1979, Ch. 9). Bitter disputes over welfare, particularly in the early 1930s, followed by the Beveridge inspired reforms of the 1940s have left the impression that the interwar welfare state was a miserable failure.

In the light of these debates it is surprising how little attention was given in the reports from some of the surveys to the effects of the welfare system in relieving poverty - even though social security payments were included in the measures of household income which were then compared with the chosen poverty line. Thus scant attention was paid to it in Bowley and Hogg's study of five towns; neither was the impact of social security systematically analysed in the NSLLL. The authors of the NSLLL simply commented that 'It is satisfactory to know that the acutest suffering caused by destitution has been removed or at least blunted by the operation of the Social Services, ... ' (Llewellyn-Smith, 1930-5, Vol. 6, p.27). Yet virtually no attempt was made to evaluate the impact of the various income maintenance programmes.

However, the issue was given more prominence in some of the later surveys. In his study of Merseyside, Caradog-Jones (1934) found that about a third of social security recipients were below the poverty line while two thirds were above it. For specific types of benefits such as Unemployment or Health Insurance benefits, and especially public assistance, large proportions of recipients were found to be below the poverty line. Rowntree found that for households in his Class A (the poorest group) 54% of income was derived from social services while for Class B (those just below the human needs poverty line) the figure was 29%. He concluded that one reason for the decline in poverty in York since 1899 was 'the remarkable growth of social services during the period under review' (1941, p.454).

Rowntree found that of total welfare payments received by all working class households in York, 64% went to those below the human needs poverty line and 34% went to the very poor (1941, p.455). On the face of it these figures seem to suggest that (i) many welfare recipients remained in poverty despite the system, and (ii) that a significant share of welfare payments went to those who were not poor. However, such inferences could be misleading, since we do not know what proportion of the households surveyed who were above the poverty line would have fallen below the poverty line in the absence of social security, nor what proportion of benefit went to such households. Neither Rowntree nor any of the other survey reports attempted any such calculation) If they had, then perhaps the interwar welfare system would not have seemed such a failure.

As is well known, the findings of social surveys, especially those of Rowntree, had a significant influence on postwar reforms. In his pathbreaking and influential Report on Social Insurance and Allied Services, Beveridge observed that:

the provision for most of the many varieties of need through interruption of earnings and other causes that may arise in modern industrial communities has already been made in Britain on a scale not surpassed and hardly rivalled in any other country of the world (1942, p.5).

Citing the interwar poverty studies he went on to argue that:

In spite of these schemes, so many persons unemployed or sick or old or widowed are found to be without adequate income for subsistence according to the standards adopted in the social surveys, this means that the benefits amount to less than the needs, and that the assistance which supplements insurance is either insufficient in amount or available on terms which make men unwilling to have recourse to it (1942, p.7).

Despite Beveridge's comment about the development of the interwar welfare system, the Beveridge Report and the welfare state revolution that followed it cast a dark shadow over what went before. That the welfare state had virtually abolished poverty seemed to be confirmed by the early postwar findings. Most prominent among these were the results of Rowntree and Lavers' third social survey of York. Although the survey was less extensive than in 1899 and 1936 (a one in nine sample rather than the whole working class population), and although it was reported in far less detail, the results were strikingly clear. The proportion of working class households in poverty had fallen from 31.0% in 1936 to a mere 4.6% (and only 2.9% of individuals) in 1950. Rowntree and Lavers (1951) suggested that this was very largely due to the effects of the welfare state. The contrast seemed, even more than before, to put the interwar system in a poor light.

In time, however, a number of critics suggested that the conclusions of Rowntree and Lavers were over-optimistic. They argued that the latter's methods were flawed, that other data showed higher levels of poverty, and that using a different definition of the household unit would produce much higher poverty levels.(4) Nevertheless, as a comparison with the interwar period, the conclusions of Rowntree and Lavers (1951) have not been seriously challenged. Perhaps their most striking conclusion was that if the welfare measures had been the same in 1950 as in 1936 the percentage of households in poverty in 1950 would have been 24.7 rather than 4.6 (1951, p.36). There are reasons to be sceptical about this result,s but in any case it does not answer the question: how much would the post-Beveridge welfare measures have reduced poverty in the interwar period? Only an analysis of interwar data can answer this question.

To summarise, there remain many questions about how the use of different poverty lines influences the proportions in poverty during the interwar period. And it is unclear how many households would be found in poverty if social security poverty lines are used as they have been in the postwar period. Furthermore there was little analysis during the interwar period of how effective social security measures were in lifting households out of poverty. The contrast between this and the attention paid to such issues in the postwar period has served to reinforce the view that the interwar welfare system compares extremely unfavourably with that of the postwar. In order to shed further light on these issues we need to re-examine social survey data from the interwar period.

3. The New Survey of London Life and Labour

The New Survey of London Life and Labour was the largest of the interwar poverty surveys. based at the London School of Economics, the inquiry was undertaken under the direction of Sir Hubert Llewellyn-Smith with the advice and collaboration of several prominent social scientists including Lord Passfield (Sidney Webb), Sir William Beveridge, and A. L. Bowley (who directed the statistical part of the inquiry). The part of the survey dealing with poverty was designed specifically to measure the progress of London's working class since Charles Booth's pioneering survey 40 years earlier. In addition to Booth's method of classification, relying on street maps of London, a detailed House Sample Inquiry was undertaken during the years 1929-31 which essentially followed the survey design established by Rowntree (1901), Bowley and Burnett-Hurst (1915), and Bowley and Hogg (1925).

The survey area covered 38 London Boroughs and comprised the area covered by Booth plus an additional nine 'Outer Boroughs' to reflect the outward expansion of London in the previous 40 years. Within this area some 28,000 households were surveyed, approximately one in 50 of London's working class. The working class was defined on occupational criteria or in cases such as the self-employed it was based on normal earnings of less than [pounds]250 per annum. The households visited were chosen under strict random sampling criteria and those subsequently found to be middle class were subsequently discarded from the analysis. The poverty line drawn by the survey was based on a scale of estimated minimum requirements for food, clothing, fuel and light, and household and personal sundries. For each household this was then compared with income net of rent paid and work expenses, both in the survey week and in a 'full' week, to determine whether the household would be classified as in poverty.(6)

The NSLLL is the only interwar social survey for which a significant number of the original records remain,(7) and samples of the original records have been used in a number of recent studies.(8) We have now computerised the entire set of surviving records. The record cards are missing for only two boroughs, Tottenham and Walthamstow, and thus we have 26,915 cases all told. Full details of the information in the computer records and of the coding schemes used are available in Bailey and Leith (1997).

4. Alternative poverty lines

The interwar period saw continued experimentation and refinement in the specification and justification of scales of minimum needs to be used as yardsticks for measuring poverty. Table 2a offers a comparison for 12 different household types of four different poverty lines (in old pence), excluding rent. The first column is the minimum standard laid down by the NSLLL. It needs little further comment except to say that it was loosely based on the poverty standard implicit in the original Booth survey (Linsley and Linsley, 1993, p.91). The second column is a modification of the NSLLL poverty line using the only available econometric estimates of household equivalence scales for the interwar period. These were estimated by Prais and Houthakker (1955) and apply only to food and clothing (see Appendix 1). Such scales are based on the actual expenditure patterns of different household types and have been widely discussed in the recent literature (see McClements, 1978; Coulter et al., 1992). This scale is the same (by construction) for a single adult male but it makes slightly more generous allowances for women and children as is clear from the comparison of the family of five in row (7). In addition, it does not make the distinction, as the NSLLL did, between pensioner households and others [TABULAR DATA FOR TABLE 2 OMITTED] since equivalence scales typically concentrate on children and do not distinguish adults by age.(9)

The third column represents Rowntree's revised poverty standard, put forward in 1936, and applied in his second social survey of York (1941), adapted to London in 1930 by Linsley and Linsley (1993). In order to extend this to different household types, we have also adapted the scales given by Rowntree (1941) for pensioner households and for additional adults in the household in a similar manner (see Appendix 1). It is noticeable that while the scales are higher all round they are especially higher for individuals (non-pensioners) as compared with families containing more than one adult. This contrast is even greater when the Rowntree/Linsley scale is compared with a scale derived from the relevant calculations given in Beveridge's 1942 Report, which in turn were partly based on Rowntree (see Appendix 1).(10) The last row of Table 2a shows the average level of minimum needs according to each of these scales when applied to all the households in the NSLLL data. These are higher than most of the entries above, reflecting the fact that the average household in the NSLLL contains 3.53 individuals. It is notable that despite the differences in the equivalence scales, the NSLLL and the adjusted NSLLL give rather similar averages as does the comparison between Rowntree/Linsley and Beveridge.

It is useful to define an alternative set of poverty lines which include an allowance for rent, and to do this we need to have a measure of minimum housing requirements. This would provide a more comprehensive measure of poverty, and a measure more consistent with those used in the postwar period. In the interwar period, poverty and overcrowding were typically treated separately although they are clearly related. Combining these measures into one poverty index could conceivably lead to quite different estimates of the extent of poverty. Clearly the result will be influenced by the overcrowding standard used to define minimum housing needs. A number of alternative standards were offered during the interwar period and four of these were used in the analysis of overcrowding of the NSLLL (for details, see Appendix 1). We have chosen two of these to determine the minimum number of rooms required by the household. The more modest standard, no more than two persons per room, is used to augment the two NSLLL poverty lines and the somewhat higher standard, based on the required number of bedrooms given the age and sex composition of the household, is paired with the Rowntree/Linsley poverty line. To convert this to a monetary value the average rent for the number of required rooms was taken from the NSLLL. This gives poverty lines inclusive of housing costs which we think are consistent with the conceptions of poverty underlying the calculation of the other constituent components of expenditure. As Table 2b shows, adding a slightly more generous housing standard to the Rowntree/Linsley scale than to the NSLLL scales preserves (roughly) the proportions between them. Beveridge's calculations included an allowance for rent and we have simply adjusted this to London rents in 1929-30 to produce the scales illustrated in the fourth column. These again are rather similar to the Rowntree/Linsley scale overall although, as before, they differ markedly as between family types.

The four scales can now be compared with a poverty line based on the interwar social security system. This is derived from the unemployment benefit scale extant in 1929-30, and, for pensioner households, from the state pension rates. Although we might regard this as similar in spirit to postwar poverty lines based on National Assistance or Supplementary benefit scales, it is important to recognise that the interwar social security system was not conceived in the same way as a universal guarantee of subsistence income.(11) In particular, it was not used as an agreed standard for poverty measurement. As is clear from the table, this standard is [TABULAR DATA FOR TABLE 2b OMITTED] broadly comparable with the NSLLL for adult earners although it was less generous towards children and pensioners. When it is applied to the full set of London households, this social security line falls not far below that of the NSLLL as a scale of minimum needs. Thus, as represented by benefit rates and pensions, the interwar social security system was not too far out of line with poverty lines which were widely used at the time.

5. Measuring poverty

In order to calculate the extent of poverty we need to compute the income in the survey week for each household.(12) Such computations were made in the NSLLL itself but imputations were sometimes made where information was found to be, or was assumed to be, missing. With a few exceptions we have preferred to avoid such ad hoc imputations. Details of the definition of income used are given in Appendix 1. One important issue is the definition of the household. Lodgers and sub-tenants are not counted as members of the household if they were clearly documented as [TABULAR DATA FOR TABLE 3 OMITTED] lodgers or were paying rent.(13) Other individuals who were evidently not part of the primary family unit but who are not distinguished as lodgers are counted as members of the household. Net income received from lodgers or rent received from sub-tenants is counted as part of household income.(14) Income is measured alternatively as gross (i.e. including rent received but not deducting rent paid) and as net by deducting rent paid. Excluding households for which there is insufficient information we can measure income and needs for 22,016 households containing 77,700 individuals.(15)

The results from these calculations are reported in Table 3 for income net of rent in the upper panel and gross income in the lower panel. For income net of rent using the NSLLL poverty line the proportion of households in poverty is 10.5%, a little higher than the 9.8% computed in the NSLLL itself. Not surprisingly slightly more adult women than men are found to be in poor households but the poverty rates for children under the age of 14 and elderly people aged 65 or over are twice the rates for prime age adults.

The second column shows what difference it makes when estimated equivalence scales are used for food and clothing. The total number of households in poverty rises to 12.1% but the proportion of children in poverty is virtually unchanged. Thus our calculations suggest that using an equivalence scale based on actual expenditure patterns, the results obtained with the original NSLLL poverty line are not too far wide of the mark.(16) They can be compared with the results of separate calculations made by Bowley (1936, p.365) in which raising the minimum needs scale for children to allow for a significantly improved diet raised the total number of households in poverty by 1.6 percentage points. But in contrast to our equivalence scale, Bowley's modification roughly doubled the number of children found to be in poverty.

The most important difference between columns (1) and (2) is the result for the elderly. Giving the same minimum needs allowance to the elderly as other adults results in a dramatic increase in the proportion in poverty, from 18% to 32%. The result arises because a substantial proportion of the elderly lived close to the poverty line. The justification for lower minimum needs for the elderly than for prime age adults is by no means strong. While old people might be expected to require less for food they might also be expected to require more for heating, clothing, and other specific needs associated with advancing age. Thus despite the prominence in contemporary debate about alleviating poverty among children, in terms of its impact on poverty, a stronger case might have been made for measures to alleviate poverty among the elderly.

Columns (3) and (4) indicate that the Rowntree/Linsley scale and that derived from Beveridge give similar results despite some differences in the structure of minimum needs. These poverty lines suggest poverty rates on the order of twice the levels suggested by the NSLLL, for households and for groups of individuals. The Rowntree/Linsley scale produces a poverty rate of 22.2% among households, confirming the calculations of Linsley and Linsley (1993, p.100).

By contrast, the contemporary social security poverty line results in only 6% of households falling into poverty. More notable than the absolute level however is the difference among individual groups. As compared with column (1), individuals under 65 are less often found to be in poverty but poverty among the elderly is close to the level found using the NSLLL poverty line. This may seem surprising at first sight given that minimum needs for elderly individuals are reduced by about as much as for couples with three children. However, it reflects the fact that elderly individuals often lived with other non-earners and that as a result, a larger proportion of them were significantly below the NSLLL poverty line. It reinforces the point that the elderly were more vulnerable to poverty than were children.

What difference does it make when both minimum needs and household income are calculated gross of rent? The calculation in column (1) of the lower panel indicates that, using more than two persons per room as the standard for overcrowding, fewer households are found to be in poverty than when income and minimum needs are calculated net of rent. Our calculation suggests that poverty would be lower by two percentage points on a gross basis than when calculated net of rent. This result can be compared with a the NSLLL itself where a similar calculation is reported using a slightly different minimum rent standard. There it was concluded that moving from a net to a gross basis made almost no difference to the proportion of households in poverty (Llewellyn-Smith, 1930-5, Vol. 6, p.90).

More notable is the fact that the difference in the basis of calculation has a large effect on one group - the elderly. This is clearly the result of the fact that many of the elderly, most likely cases in which other family members had left the home or died, were living in accommodation considerably above the overcrowding standard. As a result of 'excessive' rent a significant minority had net income less than minimum needs. If they had paid rent only for the necessary number of rooms, poverty among the elderly would have been dramatically reduced. This effect is less marked when the higher minimum rent standard applied to the Rowntree/Linsley scale is used and is absent altogether for the Beveridge scale. But it is much more marked for the social security scale. Hence excess rent could have a substantial effect on the numbers of elderly found to be in poverty when the poverty line itself is set at a low level.

It is useful to look at the intensity of poverty as well as the simple headcount ratio. The penultimate row of Table 3 shows the average income deficit for households in poverty as a percentage of their minimum needs. This is defined only for the gross of rent poverty lines.(17) The average income deficit is a little over 20% on the lower poverty lines and close to a quarter on the higher poverty lines. One measure of intensity used in recent characterisations of poverty is the product of the headcount ratio and the average income deficit (Jenkins and Lambert, 1997; Shorrocks, 1997). The final row of Table 3 shows that this measure enhances slightly the difference between the higher and lower poverty lines.

6. The impact of social security

The effects of the interwar system of social security was never a major focus of the interwar surveys, and particularly not of the NSLLL. In order to gauge the effect of [TABULAR DATA FOR TABLE 4 OMITTED] the nascent welfare state in relieving poverty we need to calculate how many households would have been in poverty in its absence. In order to do this we deduct from household income all state benefits. These include Unemployment Insurance and Health Insurance benefits, old age pensions, contributory retirement pensions, widows and orphan's pensions, workmen's compensation benefits, and payments from the poor law. We do not deduct any charitable gifts or financial assistance from relatives. Since all benefits are deducted we also add back to household income the amounts paid in National Insurance contributions.

It is likely that the removal of social security would have elicited behavioural responses in order to avert poverty. But we have too few estimates of such responses to be able to make plausible adjustments and thus our calculations of poverty in the absence of social security must be interpreted as an upper bound.(18)

Table 4 reports the percentages below the poverty line when all state benefits are excluded from income. On the NSLLL net of rent poverty line 17.2% of all households would have been in poverty, an increase of 6.8 percentage points. On the higher poverty lines, Rowntree/Linsley and Beveridge, poverty rates exceed 25%. There is some tendency for the lower poverty lines, the NSLLL and the social security line, to show larger gains in poverty relief than the higher poverty lines. This is because social security benefits were not typically high enough to carry many households up to and beyond the higher poverty lines. To a lesser extent it also reflects the fact that moderately poor households (which often contain earners) gain more from adding back social security contributions than do very poor households.

Most notable is the sharp contrast between the elderly and the other age groups. For those under 65 the effect of social security was generally to reduce poverty by between 2% and 4%, a rather modest proportion. In the absence of social security, rates of poverty for the elderly exceed 50% on all the poverty lines. Thus social security lifted between 15% and 40% of the elderly out of poverty, depending on which poverty line is chosen. Of course this still leaves a large number poor as was illustrated in Table 3, especially when income and needs are computed gross of rent. This reflects that fact that a substantial proportion of elderly people lived in households without earners and relied largely on pensions.

With the important exception of the elderly, the effects of the social security system seem surprisingly modest. But as the penultimate row shows in the absence of social security the average income deficit for those below the poverty line ranges between a half and two thirds. The average income deficit is smaller for the higher poverty lines because the increase in the income gap for very poor households is more than offset by the inclusion of households with greater non-social security income. This makes the intensity of poverty in the last row more similar across poverty lines than it was in Table 3. While the intensity of poverty is doubled for the Rowntree/Linsley and the Beveridge poverty lines, it is increased by a factor of five for the NSLLL poverty line and by a factor of ten for the social security poverty line. Thus the effects of the interwar social security system in relieving poverty appear more dramatic when the intensity of poverty is taken into account.

It is instructive also to examine total income deficits and the targeting of benefits along the lines of studies of the postwar (pre-Thatcher) years, particularly the study of Beckerman and Clark (1982). In Table 5, row (1) reports the total amount (in [pounds]s) by which the pre-benefit income of households in the survey fell short of minimum needs. Row (2) reports the total income gap after benefits and row (3) is the difference between (1) and (2). These comparisons show that social security payments made up for about half of the initial poverty gap for the higher poverty lines and as much as four-fifths for the interwar social security poverty line.(19)

Rows (4), (5), and (6) of Table 5 show, respectively, the total transfers to households who would have been poor in the absence of benefits, those who were poor even after benefits, and total benefits received by all households. These are used to [TABULAR DATA FOR TABLE 5 OMITTED] define three ratios. Row (7) is defined as 'vertical efficiency'; it is the proportion of total transfers which went to the pre-benefit poor. The proportions range from around three-fifths for the social security (gross) poverty line to nearly four-fifths for the Beveridge poverty line. These can be compared with Beckerman and Clark's (1982, p.53) figures based on the Family Expenditure Survey of 38% in 1961-3 and 57% in 1974-6. Thus even on the lowest poverty line the targeting of interwar benefits compares favourably with postwar results.(20)

'Spillover' in row (8) is that proportion of benefits going to the pre-benefit poor which raised their incomes in excess of the poverty line. These figures range from around a third to less than an eighth. They can be compared with 28% in 1961-3 and 30% in 1974-6 from Beckerman and Clark. Thus, except for the lowest poverty line, the interwar social security system pushed households up beyond the poverty line to a lesser degree than in the postwar.

Finally the 'poverty reduction efficiency' ratio in row (9) is the income gap reduction divided by total transfers to all households. This indicates that between 40% and two thirds of total payments went to reduce the poverty gap. Clearly the lower the poverty line, the smaller the share of payments that went to relieve poverty. These too can be compared with Beckerman and Clark's figures of 27% for 1961-3 and 40% for 1974-5.

Perhaps the most appropriate comparisons between the interwar and postwar periods is that which uses the interwar social security scale since the postwar estimates of poverty use supplementary benefit scales. And perhaps the most appropriate comparison is with 1974-6 a time of moderately high unemployment as was the case in London in 1929-31. On this basis the 'efficiency' of the interwar social security system and that of the postwar (pre-Thatcher) look rather similar. It is certainly not the case that the interwar system performed significantly worse on these criteria.

7. Postwar social security from an interwar perspective

Beveridge observed that, before the War, 'want was a needless scandal due to not taking the trouble to prevent it' (1942, para.445). He was greatly influenced by the spectre of poverty during the interwar period and most of the Beveridge Report was devoted to slaying this, the greatest of the 'five giants'. Rates of poverty were much lower in the postwar period and this has been attributed, in large part, to the Beveridge welfare reforms. As we have observed, Rowntree and Lavers (1951) found that, for York, the effect of welfare legislation in 1950 was to lift a very substantial proportion of households out of poverty. But postwar conditions were very different from those of the interwar period and hence the results of Rowntree and Lavers are not necessarily indicative of what poverty would have looked like in the interwar period had the reforms enacted in the wake of the Beveridge Report been in place 20 years earlier.

An alternative approach is to see what difference it would have made to the numbers in poverty if postwar social security (specifically, 1950) legislation had been in place in London in 1929-31. In order to do this we must consider differences in both the scales and the incidence of various benefits. In order to construct such a counterfactual we have made a series of assumptions which are detailed more fully in Appendix 2. The main benefits changed are Unemployment Insurance benefits (where the incidence is also widened), Health Insurance benefits, workman's compensation, and old age and widow's pensions. National Insurance contributions are also adjusted to the 1950 level (with widened incidence) and Family Allowances, which were absent in the interwar period, are added. The main benefits left unchanged are war pensions, service pensions, and non-state pensions. War pensions and service pensions appear to have been slightly more generous (in real terms) in the interwar period than after 1945. In view of this and the difficulty of determining the exact rates without the precise information about individual circumstances, these were left unchanged. Non-state pensions including those from local authorities and public utilities were also left unchanged.

The most difficult benefit to estimate is means-tested National Assistance. As is well known, already by 1950 this had become a major component of social security expenditure. Even though its incidence was much more widespread than payments under the Poor Law had been in 1930, take up was still less than complete.(21) Furthermore, the unit of assessment was not the household as reflected in the NSLLL and in other interwar surveys but the individual and his or her immediate dependants. In order to apply National Assistance with incomplete take up, we make two assumptions. First, we restrict it only to heads of household and their spouses and children under 16 (the core household). Second we apply National Assistance only when the income of the core household fell below 80% of that needed to qualify for National Assistance (see Appendix 2).(22)

In order to preserve comparison in real terms we need to adjust all the postwar benefits by the difference in prices between 1930 and 1950. According to Feinstein's estimates the consumer price index increased by 92% between 1930 and 1950 while the cost of living index (based on Ministry of Labour data) increased by 83% (Feinstein, 1972, p.T140). The latter is more appropriate since it is based on working class budget weights.(23) We therefore multiply postwar benefits by 55% before applying them to the NSLLL households in 1929-31.

It must be stressed that constructing a counterfactual such as this involves many assumptions some of which are debatable. The results in Table 6 should therefore be viewed as suggestive rather than definitive. They indicate that with the 1950 [TABULAR DATA FOR TABLE 6 OMITTED] social security system the proportion of households in poverty would have been more than halved on the lower poverty lines and somewhat less than halved on the higher poverty lines. The proportion of households in poverty falls to 4.4% on the NSLLL net of rent poverty line and to 2.5% using the interwar social security benchmark. But on the higher poverty lines there would still have been substantial numbers in poverty. If the reforms inspired by Beveridge, had been implemented 20 years earlier, then it appears that, on Beveridge's own standard, 17% of households in London would still have remained below the poverty line.

The incidence of poverty also deserves comment. Under the 1950 system children still suffer a higher incidence of poverty than prime age adults (except on the interwar social security standard) despite the introduction of family allowances for the second and subsequent children in a family.(24) Part of the reason is that the other benefits in 1950 only provided allowances for the first child whereas in the interwar period unemployment benefits and widow's pensions provided allowances for all dependent children. The elderly would have fared somewhat better under the 1950 system although their poverty rates are still substantial on the higher poverty lines. The improvement for the elderly reflects the phenomenon observed in the postwar period - that pensioners were the major recipients of National Assistance. Because National Assistance covered basic needs plus an allowance for rent paid, it especially helps the elderly who, as we have seen, often paid more in rent than the minimum standard of accommodation would require.

The large role in alleviating poverty played by means tested National Assistance, often noted by postwar observers, would probably have been even greater had it been available in the interwar period. Indeed this is the single largest reason for the substantial reduction of poverty rates under our counterfactual. In real terms, standard rates of state insurance benefits were not much higher in 1950 than they were in 1930. While pensions were a little higher, especially for single individuals, unemployment benefit rates and workman's compensation were, on the whole, lower (see Appendix 2). Although the incidence of these benefits is increased somewhat, the effects on poverty owe most to National Assistance. On our calculation, National Assistance, if applied to London in 1929-31, would have amounted to 27% of total social security payments.

It is interesting to note that even though the headcount ratio of poverty is significantly reduced as compared with Table 3, the income gap for those remaining below the poverty line falls less dramatically. With the 1950 social security system the average income deficit is still between 15 and 23%. This somewhat mitigates the decline in the index of intensity. The 1950 social security system would have cut the intensity of poverty by somewhat more than half as compared with the situation under the interwar social security system in Table 3. This is a rather less dramatic improvement, at least on the lower poverty lines, than when poverty under the interwar social security system was compared with a total absence of the system.

Under our counterfactual, total social security payments to the households in our sample would have increased by 88%. Table 7 gives the same breakdown for the counterfactual as Table 5 did for the actual benefits received. It might have been expected that with such a substantial increase in total payments, the extent to which they were targeted on the poor would decline. As column (7) shows, vertical efficiency declines, but only by about 10 to 15 percentage points as compared with the equivalent figures in Table 5. 'Spillover' is increased by about 10 to 15 percentage points. Finally, 'poverty reduction efficiency' also falls by 10 to 15 percentage points. As compared with Beckerman and Clark's figures for poverty reduction efficiency of 27% for 1961-3 and 40% for 1974-5, these results still compare favourably for the higher poverty lines but less so for the lower poverty lines.

To summarise, applying the 1950 social security system to London in 1929-30 would have substantially reduced poverty as might have been expected. But on the higher poverty lines substantial poverty would have remained. The lion's share of the increase in total social security payments would have been due to National Assistance. It is largely due to the heightened importance of this means tested benefit that payments would have remained relatively well-targeted on those in need and on closing the poverty gap.


8. Conclusion

There are essentially two views about the development of the welfare state from the 1930s to the 1950s, the period punctuated by Beveridge's much vaunted Report and the reforms which followed in its wake. One view is that the Beveridge Report and the subsequent implementation of its main recommendations was a dramatic break with the past and marked the foundation of the modern welfare state. Supporters of this view could point to the rhetoric of the Report itself, the popular acclaim with which it was received and, above all, the apparent disappearance of poverty on the scale witnessed in the interwar period. The 'rediscovery' of poverty in the 1950s notwithstanding, there remains a great divide between the experience of poverty in the 1930s and the 1950s if poverty is measured on a broadly comparable basis. But other things had changed too and so the simple comparison of ex post levels of poverty and its incidence do not provide a reliable guide to the effects of improved welfare state provision.

The other view tends to see the Beveridge Report as at best a set of incremental improvements to a system the essentials of which already existed, and at worst a flawed, misguided and ultimately unsuccessful set of reforms.(25) Adherents to this view could point to Beveridge's essentially backward looking approach, his compromises on the subsistence minimum, and the fact that when implemented many of the benefits differed little in scope and value than those that had existed before. Insofar as National Assistance became the main safety net, this was a means tested benefit somewhat reminiscent of (but less stigmatising than) Public Assistance in the 1930s. And in any case its origins owe more to Rowntree than to Beveridge who, 'by casting assistance as a minimalist agency of lower moral standing... only made the task more difficult' (Glennerster and Evans, 1994, p.70).

One reason why views differ so widely is that they use different criteria for judgement. While some focus on the development of concepts of poverty and the evolution of policy, others examine the evolution of the institutional structure and the administration of the system.(26) Insofar as the outcomes for poverty are considered the changing impact of social security provision is not generally separated from other influences. Such evaluations can only be achieved by comparing the level and incidence of poverty for the same population under different systems. And the evaluation of the impact of changes in the social security system cannot be examined in isolation from what went before.

Our findings for the interwar period have a direct bearing on these issues. First our evaluation of poverty in interwar London indicates that using the higher poverty lines which lay at the heart of the Beveridge Report, the numbers found to be in poverty is almost doubled as compared with the poverty line used in the NSLLL itself. Altering the equivalence scales makes relatively little difference as compared with varying the overall level of the poverty line. Although different levels of the poverty line give very different impressions of the degree of poverty they make rather less difference to the evaluation of the effect of social security in helping to relieve poverty. In terms of its incremental effects on relieving poverty, and in particular its poverty reduction efficiency, the 1930s social security system compares favourably even with similar measures estimated for the 1970s.

While high rates of poverty are found, even using relatively low poverty lines, rates of poverty would have been substantially higher (especially on the lower poverty lines) in the absence of the system as it existed in 1930. But although it operated at lower subsistence levels than were acceptable in the postwar period the interwar system effectively filled much of the poverty gap. Any evaluation of reforms to the system following Beveridge should recognise that fact. based on our assumptions about how the post-Beveridge social security system might have applied in the interwar period we find that the poverty count would have been halved on the lower poverty lines or somewhat less than halved on the scales used by Beveridge and Rowntree. These calculations, tentative though they are, suggest that Beveridge would not have solved the problem of interwar poverty which so greatly influenced his Report, although the reforms he spawned would have made a very significant dent in it. Furthermore, much of the poverty reducing effect would have been due to National Assistance rather than the insurance-based benefits championed by Beveridge. Thus a perspective from the interwar period of the two views outlined above and, above all, one based on quantitative comparisons, would suggest an intermediate position as perhaps the most appropriate. In terms of their impact, the effects of changes in the social security system between 1930 and 1950 were neither trivial nor were they overwhelming.


We would like to record our thanks to the ESRC for financial support for the computerisation and analysis of the New Survey of London Life and Labour under grant No. R000235697. We would also like to thank the other members of the project team, Paul Johnson, Dudley Baines, and especially Anna Leith for her extensive work on the data and illuminating advice on its interpretation. For valuable comments we are grateful to George Boyer, Peter Wardley, Matthew Woollard, participants at a conference on the Microeconomic Analysis of the Household and the Labour Market in History at the Fondation des Trielles, France, in April 1997, and to two anonymous referees.

Appendix 1

The definition of alternative poverty scales

The purpose of this Appendix is to provide details of the way the alternative poverty lines used above were defined and implemented. Significant elements of judgement are involved. The sources from which the scales are derived often do not give full details of how the basic scales were applied to different family types. In order to gain some consistency in implementing them we have applied certain rules which may deviate slightly from the way the scales were applied in the original sources. One important example is in the treatment of what we call 'implicit household overhead'. Thus where (for example) a married couple is allowed a sum larger than a single male plus an additional female (or a single female plus an additional male) then the household overhead is the (positive) difference between the two sums. This is added on for all households containing and adult male and an adult female regardless of whether they can be positively identified as a married couple. All amounts are weekly in old pence (d).

In some cases it is necessary to distinguish between pensioner households and others, either because individual allowances differ or because the amount of implicit household overhead differs. In such cases we have defined pensioner households as households containing no earners and at least one person aged 65 or over.

The NSLLL poverty standard

The basic allowances applied by the Llewllyn-Smith (1930-35, Vol. 3, p.422) excluding rent are as follows:

For the NSLLL scale including rent the minimum required number of rooms is based on the overcrowding scale R1 below.

The adjusted NSLLL scale

Our adjustments to the NSLLL scale are based on the equivalence scales for food and clothing estimated by Prais and Houthakker (1955, p.141, 152). We use the versions estimated from constant elasticity Engel curves using the Ministry of Labour's Budget survey of working class households in 1938. In order to adapt these to the NSLLL poverty line we make the same allowance as the NSLLL for an adult male but allow the relativities for food and clothing to follow the alternative equivalence scale. The other expenditure components are the same as in the NSLLL scale above but those aged 65 and over are treated the same as adults aged 18-64. Thus the adjusted scale is as follows:

male 18+                          102
female 18+                         94
boy 14-17                          91
girl 14-17                         78
child 10-13                        81
child 5-9                          60
child 1-4                          56
child [less than] 1                38
plus fuel (per household)          36

(add 2d for additional fuel costs in south London boroughs)

For the NSLLL adjusted scale including rent the minimum number of rooms required is based on the overcrowding scale R1 below.

The Rowntree/Linsley scale

Linsley and Linsley (1993) adjusted Rowntree's 'human needs' scale for York in 1936 to London in 1930 using price indices for each component of expenditure (omitting Rowntree's allowances for Unemployment Insurance and Health Insurance contributions,(27) contributions to sick and burial clubs, trade union subscriptions, and wireless). We have augmented their scale to allow for additional adults in the household and for pensioner households based on the scales reported in Rowntree (1941, p.30).(28) This gives the following scales (exclusive of rent):

For the Rowntree/Linsley scale including rent the minimum number of rooms required was calculated from the overcrowding scale R4 (below).

The Beveridge scale

In his Report on Social Insurance and Allied Services (1942, p.84-89) Beveridge drew on Rowntree, the Ministry of Labour's survey of working class budgets and other information to establish a Social Security Scale of minimum needs. This was based on 1938 prices and was inclusive of rent. From the components it is possible to derive a scale exclusive of rent. We have made no price adjustments since prices in 1938 were close to those in 1930 but we add 3s 6d to the rent allowance to allow for higher rents in London and we add an allowance for additional adults in the household (Beveridge, 1942, p.83).(29) This gives the following scales:(30)

The interwar social security scale

We derive the interwar social security scale from the rates of unemployment benefit existing before the National Economy Order of 1931 came into effect on October 8 1931. These are higher, and offer a more complete set of rates, than could be derived from the Health Insurance system (see Cohen 1932, p.27, 158) For pensioner households the rates are those prevailing under the contributory pensions scheme established under the Pensions Act of 1925 (Cohen, 1932, p.60, 71). The scale (which naturally includes rent) is as follows:

To compute the Social Security Scale excluding rent the minimum required number of rooms was calculated using the overcrowding standard R1 below.

Overcrowding standards

The NSLLL offered four different standards on which to measure the extent of overcrowding. These standards define a minimum required number of rooms for each household, which can then be compared with the number of rooms occupied. They are as follows:

R1: a minimum of one room (of any kind) for every two persons. Thus three persons require two rooms.

R2: a maximum of one (weighted) person per room where the weights are defined as follows: adults (female 16+ or male 18+) = 1; boy 14-17 = 0.75; girl 14-15 = 0.75; child 5-14 = 0.5; child 0-4 = 0.25. Thus two adults, a boy aged 15 and a child aged 10 (weighted total = 3.25) would require a minimum of four rooms.

R3: the (Manchester) bedroom standard; a minimum of one bedroom per 2.5 (weighted) persons, where the weights are: adults 10+ = 1; children 0-9 = 0.5. Subject to the requirements that married couples share but other adults must he separated by sex with a minimum of one room for three adults of the same sex.(31)

R4: the bedroom standard as above but overcrowding is defined where the number of required bedrooms plus one additional room exceeds the number of rooms occupied (except where the number of adult equivalents is less than two in which case no additional room is required).(32)

Implementing these four overcrowding standards on our data from the NSLLL gave the following results which are compared with those reported in Llewllyn-Smith (1930-5, Vol. 6, p.65):
percentage of households overcrowded

overcrowding standard               R1       R2       R3       R4
our calculation                     9.8      36.5     29.2     21.7
NSLLL                               9        34       28       16

Our results differ slightly from those used in the NSLLL due probably to slight differences in the way the rules were implemented and in the number of households included. For R4 the result differs significantly because we have modified the standard so that it does not depend on rooms actually occupied (see footnote 32). For the purposes of using these standards to augment the poverty line we eliminated R2 and R3 both of which give very high levels of overcrowding. R2 was defined by the NSLLL as crowded rather than overcrowded and hence would be inappropriate as an indicator of primary poverty. We regard R3 as unsatisfactory because, apart from bedrooms, the number of rooms required is the number actually occupied. We associate R1 with the poverty standard used by the NSLLL (and our equivalence scale variant of it) and R4 with the Rowntree/Linsley standard. The fact that the proportion of households found to be overcrowded on each of these measures is broadly similar to the proportions in poverty when measured exclusive of rent adds support to our choice.

Minimum expenditure on rent

In order to attach a value to minimum housing needs, we use rents reported in the NSLLL for dwellings according to the number of required rooms (as defined above):
rooms                               rent

1                                    60
2                                   102
3                                   126
4                                   156
5                                   186

(add 30d for each additional room over five)

Appendix 2

Applying the social security system in 1950

In order to measure the effects on poverty had the social security provisions of 1950 been in place in 1930, we apply the 1950 rates of benefit and contribution to the NSLLL households. To allow for inflation, the 1950 social security rates are first multiplied by 0.55 (all 1950 rates quoted below are 0.55 of the nominal rates prevailing in 1950). Compared with what households actually received in 1929-31, there are differences both in the rates and in the incidence of benefit or contribution. Each component is dealt with separately.

National Insurance contributions

National Insurance contributions in 1930 were typically 1s 4d for an adult male and 1s for an adult female with lower rates for younger workers. About a third of earners were not insured against unemployment and so paid at a lower rate, and some paid no National Insurance at all. The 1950 rates applied are 2s 10d for adult males and 2s 3d for adult females (including workman's compensation), with lower rates for those aged 16 and 17 and those earning less than 16s 6d per week, and higher rates for the self employed. These are applied to all earners so both the rates and the incidence are increased.

Unemployment benefits

Unemployment benefit rates in 1930 were 17s for an adult male, 15s for an adult female, 9s for an adult dependant, and 2s for each child dependant. The equivalent 1950 rates are 14s 4d for an adult male or female, 8s 9d for an adult dependant, and 4s 2d for the first child dependant only. In 1929-31 a significant number of unemployed earners failed to qualify for benefit, either because they had not paid sufficient contributions or had exhausted their entitlement, or because they were in non-insured occupations. The relevant 1950 rates were applied to all unemployed earners aged 16-65. Thus, although benefit rates are typically lower, the incidence is significantly increased.

Health Insurance benefits

Standard sickness benefits in 1930 were 15s per week for males and 12s per week for females with 7s 6d disablement benefit after 26 weeks of incapacity, although the actual payments received through Approved Societies varied considerably and in some cases dependants' allowances were paid. The equivalent 1950 rates are the same as for unemployment benefit including dependants allowances for one adult dependant and one child. These rates were applied, including dependants allowances wherever appropriate, but with no change in incidence. Thus rates are, on the whole, slightly higher but incidence is unchanged.

Workman's compensation

In 1930 rates of compensation for accidents at work were typically 30s with a sliding scale related to normal earnings for those earning less than 60s per week and with reduced rates for partial disablement. The equivalent 1950 rate for total disablement would be 24s 9d, again with reduced rates for partial disablement. Thus incidence is unchanged but rates are generally reduced.

Old age pensions

Under the 1908 Act and subsequent amendments to it the standard old age pension was 10s for males and females aged 70 and over, subject to a means test. From 1928 persons aged 65 and over became eligible for a 10s pension under the contributory scheme, subject to certain contributory conditions but not subject to a means test. The equivalent 1950 pension rates were 14s 4d per week for a single individual and 23s 1d for a married couple with 4s 2d for one dependent child. The relevant rates were applied to all old age pension recipients in the NSLLL. Thus rates are increased but incidence is unchanged (but see National Assistance below).

Widow's pensions

Under the contributory scheme widows of insured men were paid a pension of 10s per week with 5s for the first child and 3s for each additional child subject to certain conditions regarding the husband's insurance contributions. The equivalent 1950 rate is 14s 4d plus 4s 2d dependant's allowance for the first child only. Thus rates are raised on balance but incidence is unchanged.

War pensions, service pensions, and other pensions.

Pensions arising from the First World War were paid either for the loss of a breadwinner or for the total or partial disablement of a surviving servicemen. Pensions were also paid to ex-servicemen depending on rank and length of service. Similar arrangements applied in 1950 although the allowances appear to be less generous in real terms. In view of this and of the difficulty of determining the appropriate rates without precise information about individual circumstances, these have been left unchanged. Other industrial pensions, e.g. from railway companies, the GPO or the LCC and other private pensions, have also been left unchanged.

Family allowances

Family allowances were not paid in the interwar period although some allowances for child dependants were available for the unemployed and for pensioners. In 1950 child allowances were paid to virtually all families at the rate of 2s 9d for the second and subsequent children but not the first. This has been applied to all families in the NSLLL. Thus as compared with dependants' allowances available in the interwar period, incidence is dramatically increased.

Poor Law and National Assistance

The Poor Law was still in existence in the interwar period and was dispensed to applicants at the discretion of Boards of Guardians (later Public Assistance Committees). It is generally agreed that applying to the guardians was stigmatising and was often avoided even by those in desperate poverty. Under the National Assistance Act of 1948 benefits were made available as of right and the receipt of benefits was not generally regarded as stigmatising although the benefits were subject to a means test. There are two major difficulties in applying the 1950 National Assistance rates to the NSLLL: (i) determining the appropriate family unit, and (ii) determining its incidence.

National Assistance was applied to family units or tax units, rather than to households, which might comprise more than one unit for the purposes of National Assistance. In order to isolate the core family unit in each household we identify the household head, his or her spouse, and children. Where the household composition was more complex, including for instance three generations, more distant relatives, or persons unrelated to the head of household, each case was examined separately to determine the core family. Children under 16 in households where no parent could be identified were treated as the dependants of the household head. Since we know that even in the postwar period the take up of National Insurance was never complete our assumptions about incidence must reflect this. First we eliminate from consideration as separate qualifying units children of the head of household aged 16 and over and all other independent individuals or family units not containing a household head. Second, we make an allowance for National Assistance where the income of a core family is less than 80% of the National Assistance minimum, allowing for an earnings disregard of 11s (the equivalent of [pounds]1 in 1950) and assuming a contribution to the core household of 2s 9d from the earnings of each earner not counted in the core family. In such cases we allow full National Assistance plus rent paid. To give an example, a retired couple paying 8s rent would be given National Assistance if their weekly income was lower than [0.8.sup.*](22s + 8s) = 24s. If they received only two pensions of 10s each then they would be given an extra (22s + 8s) - (10s + 10s) = 10s National Assistance. But if they had possessed an additional 4s of unearned income they would not be given National Assistance on our rules even though they would, in principle, qualify. Applying National Assistance in this manner allows us to compensate for not having increased the incidence of pensions (see above) and it is consistent with the observation that, when National Assistance was introduced, a large share went to elderly people, most of whom were receiving pensions.

When our criteria are met we calculate National Assistance, subject to the disregards, using the following basic rates:

adult aged 16 and over           158
couple                           264
child 11-15                       69
child 5-10                        59
child 0-4                         50

Plus net rent less income of core household and less 33d per other adult household member

There are a number of elements of National Assistance which we were not able to apply and which will impose biases in both directions. First we have not taken into account discretionary additions to National Assistance for a variety of needs. These applied to about a third of all claims in 1950 (Atkinson, Maynard, and Trinder, 1981, p.23) and amounted to about 10% of total payments. Second, we have not allowed Assistance to be reduced where rent paid was in excess of what the National Assistance Board's officers regarded as reasonable. After new rules were introduced in 1950, the proportion of rents not fully met by the board was 7% (George, 1968, p.217). Third, we have not imposed the 'wages stop' - the reduction in the allowance made if the allowance would otherwise exceed earnings in full-time employment. Finally, we have not taken into account the disqualification for Assistance of applicants owning capital with a value of over [pounds]400 (disregarding [pounds]375 of War savings and the value of owner-occupied housing), but it is doubtful that this would have applied in many cases.

1 Most of the local social surveys undertaken up to 1934 are summarised in Wells (1935).

2 Applying the minimum needs scale used for London to Bristol gave 6.9% of households in poverty (Tout, 1938, p.51).

3 A partial exception to this is the report for Sheffield: 'The part played by social income in maintaining the standard of living is, of course, very important. It has not been possible to prepare a detailed statement of the effect of the receipt of social income on the standard of living of the families considered in time for publication of this report, but the results of a rough calculation suggest that, had it not been for the receipt of social income (leaving public assistance relief out of account) rather more than sixty per cent of the families would have been below the poverty line at the time when the investigation was made.' (Owen, 1933, p.43). For Bristol, Tout estimated that in the absence of Public Assistance alone the proportion of households in poverty would have increased from 10.7% to 12.1% (1938, p.29).

4 See Townsend (1952), Abel-Smith and Townsend (1965), and Atkinson, Corlyon, Maynard, Sutherland, and Trinder (1981). Abel-Smith and Townsend used the Family Expenditure Survey for 1953/4. They found that 5.4% of households were below the Rowntree and Lavers poverty line (updated to 1953/4), although they acknowledged the difficulty of inferring income from expenditure data. Atkinson, Corlyon, Maynard, Sutherland, and Trinder (1981) re-analysed the surviving records from Rowntree and Lavers' survey. They found that, basing the measurement of income and needs on the primary, family (following National Assistance definitions) rather than on the household (as in the social surveys), 14.4% of working class families fell below the National Assistance line (a somewhat lower poverty line that that used by Rowntree and Lavers).

5 Rowntree and Lavers' counterfactual is described as the 'situation in 1950 if welfare measures had been identical with those in force in 1936' (1951, p.39; our italics). But there is insufficient detail to judge exactly what assumptions were made: whether 1936 benefits were taken at real or nominal values; whether dependants allowances were included; and whether the incidence of benefits was changed. Furthermore the overall difference in incomes is taken to include the effects of food subsidies in 1950 which (arguably) were not part of the welfare state.

6 It might be suspected that, given those directing the NSLLL were closely involved in the pre-war welfare reforms, they might have had an interest in presenting as favourable a picture as possible, If so then it is even more surprising that there was so little comment on the effects of social security in the published volumes. There is no evidence at all in the NSLLL records of any deliberate bias and we can closely replicate the published results. As Linsley and Linsley (1993, p.91) point out, the poverty line used was somewhat higher than that implicit in Booth's survey and hence it revealed rather more poverty than a strict comparison with Booth would have done.

7 Some original schedules survive from Rowntree's second social survey of York, but there are only 1,356 out of the original 16,362, and these do not appear to be a random selection of the originals (see Hatton and Bailey, 1993, p.238). The records for Tout's survey of Bristol in 1937 survive only in the form of coded output from punched-card originals (see Wardley and Woollard, 1994).

8 See Eichengreen (1986, 1987); Gordon (1988); Hatton and Bailey (1988, 1993); Linsley (1986); Linsley and Linsley (1993).

9 It is interesting to note two important differences in the structure of interwar poverty lines as compared with those used in the postwar period, typically based on National Assistance or Supplementary Benefit scales. First, in the postwar period the allowance for pensioners is higher than for those of working age whereas in the interwar period it was typically lower. Second, in contrast to the interwar period, males and females received the same allowance in the postwar scales (Atkinson, 1989, p.56).

10 Townsend (1952, p.26) derived the same scales from the Beveridge Report; see also McClements (1978, p. 116) for a comparison with other (postwar) equivalence scares.

11 As Gilbert (1966, p.258) observes, unemployment benefits and pensions were not viewed as sufficient for subsistence but were expected to be supplemented by private savings. As compared with National Assistance rates in 1950, the 1930 unemployment benefit rate was somewhat higher and pension rate somewhat lower in real terms (see Appendix 2).

12 Two different measures of earnings were reported in the NSLLL for each earner in the household: actual earnings last week, and earnings that would normally be received in a 'full week'. We calculate household income as the total of last week's earnings for all members of the household (excluding lodgers) plus all other sources of income minus travel expenses and minus state insurance deductions.

13 Lodgers (but not subtenants who are counted as separate households) are included in the calculation of minimum housing needs.

14 Even if we wished to include lodgers as full members of the household we would not be able to do so because in most cases only their contribution is recorded and not their income. Hence they are excluded from consideration completely. However it is unlikely that integrating lodgers into the households would make much difference to the proportions in poverty. There are only 324 lodgers in the set of households used in the analysis; of these 20 (eight of whom were employed) were in households classified as poor on the NSLLL poverty line and 56 (of whom 32 were employed) were in households classified as poor on the Rowntree/Linsley poverty line. Thus pooling their incomes would have had very small effects either in reducing or increasing the numbers in poverty.

15 We cannot compare these numbers with those counted in the calculations made by the NSLLL since no totals were given.

16 It should be noted however that the results obtained here will be influenced not just by the different equivalence scales applied to children but to adults as well since poverty is calculated on a household rather than an individual basis.

17 This measure is less meaningful when applied to net of rent poverty lines because households with income less than their rent would have a deficit of more than 100%. This becomes a more serious problem when we examine poverty in the absence of social security (see Table 4).

18 An estimate (using samples from the NSLLL) of the effects of unemployment benefit rates on the incidence of unemployment (Eichengreen, 1987) suggests that the induced effects were modest. Another study suggests that a fall in household income would have caused a modest increase in the labour market participation of females (Hatton and Bailey, 1988). It seems likely that the most important (and most imponderable) effect of removing benefits would have been a reorganisation of households. In particular, it seems likely that more of the elderly would have been found living with relatives.

19 These calculations (unlike those underlying Table 4) do not net out National Insurance contributions.

20 It should be noted that Beckerman and Clark's (1982) calculations are not restricted to the working class alone. But this will not affect the comparisons between their results and those based on the NSLLL as long as middle class households in the interwar period were not poor and did not receive social security benefits. While there would have been a few middle class pension recipients, they are unlikely to affect the results greatly.

21 Using the remaining records from Rowntree's third survey, Atkinson, Maynard, and Trinder (1981, p.28) estimated a take-up rate of little more than a half for York in 1950, and using a Ministry of Labour survey, Abel-Smith and Townsend (1965, p.48) estimated a take-up rate of about 61% in 1960. For pensioners, the Ministry of Pensions and National Insurance (1966, p.19) estimated a take-up rate of 62% in 1965.

22 It is not possible to calculate exactly the take-up rate implied by these assumptions because we have not assessed separately the individuals and family units which do not include a household head. But for core families our assumptions imply a take-up rate of two-thirds.

23 Until June 1947 the budget weights in Ministry of Labour's cost of living index were based on the budget survey of 1904; the interim index from 1947 used the weights of the 1938 survey. Feinstein (1972) uses the 1904 weights up to 1938 and the geometric mean of 1938 and 1950 for 1938 to 1952. Calculations over the wartime period are sensitive to the weights used. The Ministry of Labour's original index increased by 30% between 1938 and June 1947, but as Seers (1949) and Allen (1952) demonstrated, using 1938 rather than 1904 weights the increase is 62%. This is largely because food prices and rents were kept down during (and after) the War. Feinstein's index increases by 78% over the same period. Using this index represents a compromise since (i) it uses weights for all working class households, and not the poorest whose budget weights would give a smaller price rise, and (ii) it includes rent, the exclusion of which would give a larger price rise. Alternative price adjustments for poverty lines across the War are discussed by Lynes (1962, p.34) and Atkinson (1989, pp.57-60).

24 Family allowances were paid whether or not a member of the household was unemployed, a feature which is incorporated into the calculations reported here.

25 A number of the contributions in Hills et al. (1994) offer views along these lines. For a discussion of the institutional background to the development of the welfare state, see Lowe (1990).

26 As Melling points out in his sketch of the literature up to 1990, "Much of the debate on Beveridge and welfare reform has taken place around the issues of his intellectual consistency and the political intentions of both himself and contemporary bureaucrats. . . The important issue is the practical implications of his proposals" (Melling, 1991, p.78).

27 We have netted these out of household income (see below).

28 For non-pensioner households the 'implicit household overhead' is negative, that is, 321d (the amount for a married couple) minus 345d (the amount for a single male plus an additional female) = -24. For a pensioner household it is 259d - 264d = -5d. The allowances for additional adults and for pensioner households (not reported by Linsley and Linsley, 1993) were derived by applying the relevant ratios to the figures given by Rowntree (1941, p.30). For additional females in both types of households the allowance was fixed so that a male plus an additional female gives the same total as a female plus an additional male. This gives the result (consistent with Rowntree, 1941, p.30, footnote 4) that additional females get the same allowance in pensioner and in non-pensioner households and that male additional members in pensioner households get an allowance 12d less than in non-pensioner households.

29 The rents given by Beveridge are raised by 54% for individuals and 60% for multiple person households.

30 For non-pensioner households the 'implicit household overhead' is (by comparing a couple with a male plus an additional female) 14d including rent and 60d excluding rent. For pensioner households the implicit overhead is 30d and 36d respectively.

31 One variant of the standard (which we use below) compares the minimum required rooms with the number of rooms in the dwelling designated as bedrooms. An alternative compares the minimum required rooms with the number of rooms actually used as bedrooms.

32 This differs from the way the 'necessary bedroom standard' was used in the LLewellyn-Smith (1930-5, Vol. 3, pp.230, 248). The rule implemented there was that in one room tenements no room other than a bedroom was required. This means that the number of rooms required can be influenced by the number actually occupied. However, if we try and avoid this by specifying that, when the required number of bedrooms is one, no additional room is required, we run into an anomaly. In this case a couple plus a child (2.5 equivalent adults) would require only one room but an extra child would raise the requirement to three rooms. There would be no case where the required total number of rooms is two. Under our modification (which does not depend on the number of rooms actually occupied) an adult and a child would require only one room, a couple or a couple and a child would require two rooms, while a couple and two children would require three rooms.


Abel-Smith, B. and Townsend, P. (1965). The Poor and the Poorest, G. Bell and Sons, London.

Allen, R.D.G. (1952). 'Index Numbers of Retail Prices, 1938-51', Applied Statistics, 1, 106-17.

Atkinson, A.B. (1989). Poverty and Social Security, Harvester Wheatsheaf, London.

Atkinson, A.B., Maynard, A.K. and Trinder, C.G. (1981). 'National Assistance and Low Incomes in 1950', Social Policy and Administration, 15, 19-31.

Atkinson, A.B., Corlyon, J., Maynard, A.K., Sutherland, H. and Trinder, C.G. (1981). 'Poverty in York: a Re-analysis of Rowntree's 1950 Survey', Bulletin of Economic Research, 33, 59-71.

Bailey, R.E. and Leith, A. (1997). 'New Survey of London Life and Labour: Codebook for the Data Files (Release 1.0)', University of Essex, unpublished.

Beckerman, W. and Clark, S. (1982). Poverty and Social Security in Britain since 1961, Oxford University Press, New York, NY.

Beveridge, W.H. (1942). Report on Social and Allied Services, (Cmd. 6404), HMSO, London.

Blaikie, A. (1990). 'The Emerging Political Power of the Elderly in Britain, 1908-1948', Ageing and Society, 10, 17-39.

Bowley, A.L. (1936). 'The New Survey of London Life and Labour: Effect of Modifying the Poverty Line', Journal of the Royal Statistical Society, 99, 364-6.

Bowley, A.L. and Burnett-Hurst, A.R. (1915), Livelihood and Poverty, G. Bell and Sons, London.

Bowley, A.L. and Hogg, M.H. (1925). Has Poverty Diminished? P. S. King, London.

Cohen, P. (1932). The British System of Social Insurance, P. Allan, London.

Caradog-Jones, D. (1934). The Social Survey of Merseyside, (3 Vols), Hodder and Stoughton, London.

Coulter, F.A.E., Cowell, F.A. and Jenkins, S.P. (1992). 'Differences in Needs and Assessment of Income Distributions', Bulletin of Economic Research, 44, 77-124.

Eichengreen B.J. (1986). 'Unemployment in Interwar London: a Microeconomic Perspective', Journal of Interdisciplinary History, 17, 335-48.

Eichengreen, B.J. (1987). 'Unemployment in Interwar Britain: Dole or Doldrums?', Oxford Economic Papers, 39, 597-623.

Feinstein, C.H. (1972). National Income, Expenditure and Output of the United Kingdom, 1855-1965, Cambridge University Press, Cambridge.

Ford, P. (1934). Work and Wealth in a Modern Port: an Economic Survey of Southampton, George Allen and Unwin, London.

George, V. (1968), Social Security: Beveridge and After, Routledge and Kegan Paul, London.

Glennerster, H. and Evans, M. (1994). 'Beveridge and His Assumptive Worlds', in J. Hills, J. Ditch, and H. Glennerster (eds), Beveridge and Social Security, Clarendon Press, Oxford.

Gordon, C. (1988). 'Familial Support for the Elderly in the Past: The Case of London's Working Class in the Early 1930s', Ageing and Society, 8, 287-320.

Hatton, T.J. and Bailey R.E. (1988). 'Female Labour Force Participation in Interwar Britain', Oxford Economic Papers, 40, 697-718.

Hatton, T.J. and Bailey, R.E. (1993). 'Household Labour Supply and Women's Work in Interwar Britain', Explorations in Economic History, 30, 229-56.

Hills, J., Ditch, J. and Glennerster, H. (eds.) (1994) Beveridge and Social Security: an international retrospective, Clarendon Press, Oxford.

Jenkins, S.P. and Lambert, P.J. (1997). 'Three 'I's of Poverty Curves, with an Analysis of UK Poverty Trends', Oxford Economic Papers, 49, 317-27.

Linsley, C.A. (1986). 'The Housing Market, Deprivation and Household Behaviour in Interwar Britain: a Microeconometric Analysis of the New Survey of London Life and Labour', Ph.D. thesis, University of Essex.

Linsley, C.A. and Linsley, C.L. (1993). 'Booth, Rowntree and Llewellyn-Smith: a Reassessment of Interwar Poverty', Economic History Review, 46, 88-104.

Llewellyn-Smith, H. (ed.) (1930-5). The New Survey of London Life and Labour, (9 vols.), P.S. King, London.

Lynes, T. (1962). National Assistance and National Prosperity, Codicote Press, Welwyn.

McClements, L.D. (1978). The Economics of Social Security, Heinemann, London.

Macnicol, J. (1980). The Movement for Family Allowances, Heinemann, London.

Melling, J. (1991). 'Reading Beveridge: Recent Research on Pre-war Social Policy', Social Policy and Administration, 25, 73-8.

Ministry of Pensions and National Insurance (1966). Financial and other Circumstances of Retirement Pensioners, HMSO, London.

Owen, A.D.K. (1933). A Survey of the Standard of Living in Sheffield, Sheffield Social Survey Committee, Sheffield.

Prais, S.J. and Houthakker, H.S. (1955). The Analysis of Family Budgets, Cambridge University Press, Cambridge.

Rowntree, B.S. (1901). Poverty: A Study of Town Life, Macmillan, London.

Rowntree, B.S. (1937). The Human Needs of Labour, Longmans Green, London.

Rowntree, B.S. (1941). Poverty and Progress, Longmans Green, London.

Rowntree, B.S. and Lavers, G.R. (1951). Poverty and the Welfare State, Longmans Green, London.

Seers, D. (1949). Changes in the Cost of Living and the Distribution of Income since 1938, Blackwell, Oxford.

Shorrocks, A.F. (1997). 'Deprivation Profiles and Deprivation Indices', in S.P. Jenkins, A. Kapetyn, and B. van Praag (eds), The Distribution of Welfare and Household Production: International Perspectives, Cambridge University Press, Cambridge.

Stevenson, J. and Cook, C. (1979). The Slump: Society and Politics during the Depression, Quartet, London.

Townsend, P. (1952). 'Poverty: Ten Years After Beveridge', Planning, 19, 21-40.

Tout, H. (1938). The Standard of Living in Bristol, Arrowsmith, Bristol.

Wardley, P. and Woollard, M. (1994). 'Retrieving the Past: a Reclamation and Reconstruction of the Social Survey of Bristol, 1937', History and Computing, 6, 85-105.

Wells, A.F. (1935). The Local Social Survey in Great Britain, George Allen and Unwin, London.
COPYRIGHT 1998 Oxford University Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1998 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Hatton, Timothy J.; Bailey, Roy E.
Publication:Oxford Economic Papers
Date:Oct 1, 1998
Previous Article:On economic causes of civil war.
Next Article:Who should be on workfare? The use of work requirements as part of an optimal tax mix.

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters