Intergenerational Mobility In Europe And North America

Intergenerational Mobility in Europe and North America

A Report Supported by the Sutton Trust

Jo Blanden, Paul Gregg and Stephen Machin1

April 2005

Embargo: 00.01 HOURS Monday April 25th

Contacts: Jo Blanden [email protected] Tel: 020 78523524 Paul Gregg [email protected] Tel: 0117 9289083

1

Jo Blanden and Stephen Machin are in the Department of Economics, University College London and Centre for Economic Performance, London School of Economics, Paul Gregg is in the Department of Economics, University of Bristol and Centre for Economic Performance, London School of Economics.

Executive Summary

The level of intergenerational mobility in society is seen by many as a measure of the extent of equality of economic and social opportunity. It captures the degree of equality in life chances - the extent to which a person’s circumstances during childhood are reflected in their success in later life, or on the flip-side, the extent to which individuals can make it by virtue of their own talents, motivation and luck.

Under a project supported by the Sutton Trust we have sought to understand more about how intergenerational mobility compares across countries in Europe and North America. In addition, work has been carried out to understand more about mobility in Britain: how mobility has changed over time; and the role of education in shaping opportunity.

The key findings are: •

International comparisons indicate that intergenerational mobility in Britain is of the same order of magnitude as in the US, but that these countries are substantially less mobile than Canada and the Nordic countries. Germany also looks to be more mobile than the UK and US, but a small sample size prevents us drawing a firm conclusion.

•

Intergenerational mobility fell markedly over time in Britain, with there being less mobility for a cohort of people born in 1970 compared to a cohort born in 1958. No similar change is observed in the US.

•

Part of the reason for the decline in mobility has been the increasing relationship between family income and educational attainment between these cohorts. This was because additional opportunities to stay in education at both age 16 and age 18 disproportionately benefited those from better-off backgrounds.

•

For a more recent birth cohort (born in the late 1970s and early 1980s), there is a more mixed picture on changes in educational inequality.

Their education

participation in the 1990s was characterized by a narrowing in the gap between the staying on rates at 16 between rich and poor children, but a further widening in the inequality of access to higher education.

2

•

The expansion of higher education since the late 1980s has so far disproportionately benefited those from more affluent families.

•

The research shows clearly, using a variety of identification techniques, that family income in the childhood years does make a genuine difference to educational outcomes, rather than reflecting other aspects which differ across families. However, the estimates are not able to say definitively whether this causal effect has increased in strength over time.

Implications International comparisons of intergenerational mobility show that Britain, like the United States, is at the lower end of international comparisons of mobility. Also intergenerational mobility has declined in Britain at a time of rising income inequality. The strength of the relationship between educational attainment and family income, especially for access to higher education, is at the heart of Britain’s low mobility culture. If improving intergenerational mobility is viewed as desirable, this clearly suggests that from early ages, including prior to school entry, Britain needs to adopt a strategy to equalize opportunities. This should apply at all stages of the education process, and include support during the early years, for both parents and children; policies to improve the performance of deprived children in schools; and steps to promote participation at the post-compulsory level. Such policies have the potential to enhance intergenerational mobility by ensuring greater equality of educational opportunity.

3

Intergenerational Mobility

The level of intergenerational mobility in society is seen by many as a measure of the extent of equality of economic opportunity or life chances. It captures the extent to which a person’s circumstances during childhood are reflected in their success in later life, or on the flip-side, the extent to which individuals can make it by virtue of their own talents, motivation and luck. The most intuitive way to see the extent of intergenerational mobility is to see where children from the most or least affluent families end up in the earnings or income distribution as adults. This can be shown by a transition matrix showing movements in the income distribution across generations. Table 1 reports an example of such a transition matrix for Britain for children born in 1970. It splits each generation’s distribution up into four equal sized quartiles (each containing 25 percent of people) and sees how much movement there is between quartiles across generations. In a fully mobile society a quarter of the children from each income group would then end up in each quarter of the adult earnings distribution, so every cell would contain a .25. In the case of no mobility, all children would be in the same quartile as their parents and the lack of movement between quartiles would be shown by 1’s on the diagonal and 0’s elsewhere. The Table makes it clear that, for the cohort born in 1970, 37% remained in the poorest quarter as adults, whilst only 16% made it to be among the most affluent as adults. Likewise, far more of the most affluent quarter remains in the top quarter in the next generation than would occur with perfect mobility. Table 1: Transition Matrix for Britain, Sons Born in 1970 Sons’ earnings quartile aged 30 in 2000 Bottom 2nd Parental average 3rd Top income quartile (average of incomes measured when son aged 10 and 16) Bottom .37 .23 .23 .16 nd 2 .30 .30 .24 .16 .20 .24 .29 .27 3rd Top .13 .23 .24 .40 Data drawn from the British Cohort Study of 1970 as described in the text.

4

Among economists, intergenerational mobility is most commonly measured by an intergenerational elasticity (β) measuring the strength of the statistical association between parent and child outcomes. A higher elasticity indicates a stronger impact of parental outcomes on children’s economic success, meaning higher intergenerational inequality and less intergenerational mobility. If β equals1 this corresponds to complete intergenerational immobility. If β equals 0, and there is no relationship between incomes across generations, this corresponds to complete mobility. On this basis our research reports an intergenerational elasticity of son’s earnings with respect to family income of .29 in Britain for those born in 1970. From one such observation of intergenerational mobility in one country, it is not instantly obvious what constitutes a high or low level of mobility. So here we adopt two approaches to give benchmarks for mobility in countries. Firstly we compare mobility across a set of other major industrial countries and secondly we use a historical comparison to suggest whether the extent of mobility in Britain has changed over time.

International Comparisons of Mobility Most evidence on intergenerational mobility across countries is generally from studies considering one or two countries in isolation. However, drawing strong conclusions about relative levels of mobility in different countries is hampered by the fact that few studies are carried out with an explicit comparative aim. Different researchers take their own decisions about variable choice, sample selection and estimation methods, meaning that it is impossible to know whether differences are a consequence of fundamentals or a lack of comparability across studies. The research outlined here seeks to fill this gap. Here we combine our own analysis for mobility in Britain, the US, West Germany and Canada, with research by Bjorklund et al (2005) who consider Britain, the US, Sweden, Norway, Finland and Denmark. Both studies are strongly focused on using a consistent approach across the studies. Combining them enables a comparison to be made over eight countries. Table 2 provides estimates of intergenerational mobility on the available data across the two studies. Data limitations mean that not all countries are available for

5

the same broad periods of time. Also, for some countries data is available on father’s earnings whereas for others we only have combined parental income.2

Table 2: Internationally Comparable Estimates of Intergenerational Mobility Country

Dataset

Sons Born

Britain

British Cohort Study

1970

US

Panel Study of Income Dynamics

19541970

Age 30

West Socio-Economic Panel Germany

19601973

2000

Canada

19671970

1998

1958

1992 and 1999 (average) 1998 and 2000 (average) 1996 and 1999 (average) 1995 and 2000 (average)

Norway

Intergenerational Income Data (from tax registers) Register data

Denmark Register data

19581960

Sweden

Register data

1962

Finland

Quinquennial census panel

19581960

a

Sons Earnings Measure 2000 (Age 30)

Measure of Parental Status

Intergenerational partial correlation1 .271a

Parental income 1980 and 1986 (average) Parental income .289 a when son age 10 and age 16 (average) Parental income .171a 1984 and 1988 (average) Parental income .143 a when son aged 16 Father’s earnings 1974

.139 b

Father’s earnings 1980

.143 b

Father’s earnings 1975

.143 b

Father’s earnings 1975

.147 b

Blanden (2005) Table 3.3

b

Bjorklund et al (2005) Table 3.

The partial correlation is equal to the beta coefficient scaled to adjust for changes in inequality across generations. This is important as inequality grew at different rates for the countries in this sample. 1

These results differ slightly from those in Table 5 owing to some adjustments required to ensure that

results are comparable across countries and over time.

The results in Table 2 that compare intergenerational mobility across the eight countries suggest a clear pattern.

America and Britain have the highest

intergenerational persistence (lowest mobility). Germany is around the middle of the estimates, while the Nordic countries and Canada all appear to be rather more mobile. 2

The earlier data uses a two year average of earnings which will include earnings measured at a younger age, the later data has a single year’s measure for earnings as an adult.

6

Among the Nordic countries the levels of mobility are similar: Norway has the greatest mobility and Sweden the least. The estimates shown here are broadly in line with what we would expect from the current literature which takes one country at a time (as reviewed by Corak, 2004) Thus the picture that emerges is that Northern Europe and Canada are particularly mobile and that Britain and the US have the lowest intergenerational mobility across the European and North American countries studied here. The USA is seen by some as a place with particularly high social mobility. In part this is a consequence of using measures of class to estimate mobility (these will be affected by changes in the class structure over time). However, the idea of the US as ‘the land of opportunity’ persists; and clearly seems misplaced. As we go on to show below, low mobility in Britain is partly explained by the strong relationship between parental income and educational attainment. For the US, the picture is slightly different - parental income leads to a less marked advantage in terms of education, but this educational advantage is worth more in the labour market in the US than in the other countries. Another important dimension of the low mobility in the US is related to race, with Hertz (2004) showing that mobility is substantially more restricted for black families than white families, although he does not show precisely how much of the persistence this accounts for.

Changes in Intergenerational Mobility in Britain In the internationally comparable estimates, information has been used for two cohorts of British sons, those from the National Child Development Survey (NCDS) who were born in 1958, and those from the British Cohort Study (BCS) who were born in 1970, which were used in the first section. If care is taken in treating these datasets comparably then they can be used to explore how intergenerational mobility has changed over time. Looking at the transition matrices reported in Tables 3 and 4 it is instantly clear that many more children from the poorest quarter remain in poorest quarter as adults in the more recent cohort. Likewise among the most affluent far more stay among the most affluent as adults than was the case for the earlier cohort.

7

Table 3: Transition Matrix for Sons born in 1958 Sons’ earnings quartile when aged 33 in 1991 Parental income Bottom 2nd 3rd Top quartile when son aged 16 Bottom .31 .28 .23 .17 2nd .30 .28 .23 .19 3rd .22 .25 .25 .28 Top .17 .20 .28 .35 Data drawn from the National Child Development Survey.

Table 4: Transition Matrix for Sons born in 1970 Sons’ earnings quartile when aged 30 in 2000 Parental income Bottom 2nd 3rd Top quartile when son aged 16 Bottom .38 .25 .21 .16 2nd .29 .28 .26 .17 3rd .22 .26 .28 .25 Top .11 .22 .24 .42 Data drawn from the British Cohort Study.

Considering estimates of the intergenerational elasticity confirms that intergenerational mobility has fallen over time in Britain; equality of opportunity declined for those born in 1970 compared with those born in 1958. Table 5 provides the results for males, showing the estimated mobility parameter - the intergenerational elasticity of earnings with respect to family income - and the partial correlation across the generations. It is clear that by either of these measures intergenerational mobility has declined substantially and that these changes are statistically significant. In our underlying papers we explore these results and we are confident that they are robust.

Table 5: Changes in Intergenerational Mobility in Britain NCDS 1958 .205 (.026)

BCS 19703 .291 (.025)

Change Mobility .085 (.036) Parameter Partial Correlation .166 (.021) .286 (.025) .119 (.033) Sample Size 2163 1976 Source: Blanden (2004) Table 4.2 Standard errors are shown in parentheses. 3

These results differ slightly from those in Table 2 owing to some adjustments required to ensure that results are comparable across countries and over time.

8

As a check that this is not a widespread phenomenon in developed countries experiencing rising inequality, we have also explored the extent to which this experience was shared by the US over the same period. The conclusion from this exercise is that there has been no similar change in intergenerational mobility in the US to match the one that occurred in Britain between the 1958 and 1970 cohorts. This indicates that what happened in Britain is exceptional even when compared with a country experiencing similar changes in inequality4. Decomposing the role of education What lies behind this large reduction in mobility in Britain? Many commentators and political parties link mobility and the education system. In our studies the persistence between incomes across generations is decomposed into that part which is related to education and the part which is not. Further, the education-related aspect of persistence can be split into that part due to the difference in education attainment between people from different income groups and the value of this education in the labour market (i.e. the extent to which those with more education are paid more). Formally: It is clear that education attainment varies according to parental income, such that Edijson = α 0 j +ψ j ln Yijparents + eij , where j refers to the cohort. Education

has

benefits

in

the

labour

market

such

that

ln Yijsons + α1 j + φ j Edijsons + uij where φ j denotes the return to education in cohort j. This means that the overall intergenerational elasticity can be decomposed into the return to education multiplied by the relationship between parental income and education, plus the unexplained persistence in income that is not transmitted through education.

β j = φ jψ j +

Cov(uij , ln Yijparents )

(3.3)

Var (ln Yijparents )

To explore this decomposition the highest qualification levels of the cohort members are translated into the number of years generally taken to obtain them. The results of the decomposition are shown in Table 6.

It is clear that

education has an important role to play, with around 35 to 40 percent of the intergenerational coefficient being accounted for by the measures of education used in the decomposition. 4

In addition, the Table demonstrates that the increase in

See Blanden (2005) Chapter 5 for more information on changes in mobility over time in the US.

9

intergenerational mobility is explained by two factors: an increase in the sensitivity of education to parental income and educational attainment, and an increase in the link between parents’ incomes and sons’ earnings which is not explained by education. An increase in the wage returns to education in the labour market would also lead to an increase in persistence, however the evidence presented here suggests that for these cohorts at least, the returns to education have not changed.

Table 6: Education and Intergenerational Mobility: Decomposition β

1958 Cohort 1970 Cohort

Return to education (φ j )

Relationship between parental income and education (ψ j )

Persistence through Education ( φ jψ j )

Persistence not through education

.205 (.026)

.081 (.004)

.947 (.121)

.077

.132 (.024)

.291 (.025)

.075 (.005)

1.350 (.098)

.101

.191 (.024)

Cov(u son , ln Y parents ) Var (ln Y parents )

Changes in the Impact of Education on Family Income As the link between family income and educational attainment of children has increased between the two cohorts, the next step is to assess how education levels achieved have evolved for young people from different family backgrounds. We consider two stages of educational performance, staying on at school after the compulsory school leaving age of 16 and Higher Education degree attainment. As we are now considering educational attainment rather than adult earnings we can add a third cohort of British children reaching age 16 in the 1990s (from the British Household Panel Survey). This gives a partial picture of how mobility may be changing for a more recent birth cohort. Figure 1 shows the proportions of young people (both males and females) staying on in education beyond age 16 over time. Educational inequality is measured as the difference in the staying on rate of young people with parental income in the richest 20 percent compared with young people with parents in the poorest 20 percent. The first thing to note about these results is that the staying on rate has increased from 1974 to the late 1990s for young people from both income groups. The more interesting result is that the speed of the increase has varied substantially for young people in different periods. It is clear that between 1974 and 1986 (when the cohorts used in the previous section were aged 16) staying on rates for children from the

10

richest backgrounds were rising faster; this led to an increase in educational inequality. From 1986 to the late 1990s the staying on rate of those from the poorest backgrounds was rising more quickly, leading to a reversal in the extent of educational inequality5. Over the 1990s young people from poorer backgrounds have clearly taken up the opportunity to stay on in post-compulsory education, as never before. This is likely to be in part a consequence of the introduction of the GCSE Given the variety of courses available at further education colleges there might be a wide variation in the courses that young people stay on to do. Therefore the next question is: are the trajectories that they are on leading to higher qualifications? Figure 1: Staying on Rates (Proportions) by Parental Income Group 1

0.9

0.86

0.8

0.7 0.7

0.61 0.6

0.5

Poorest 20 percent at age 16 Richest 20 percent at age 16 Difference (Educational Inequality)

0.45 0.38

0.4

0.32 0.3

0.25

0.24 0.21 0.2

0.1

0 1958 Cohort (NCDS)

1970 Cohort (BCS)

Late 1970s cohort (BHPS)

Proportion of sample who stayed on at school after age 16

Source: Blanden, Gregg and Machin (2005) Table 5.3.

We can find out more about this by considering the completion of higher education by income group in a similar way. Figure 2 presents results similar to those from Figure 1 but this time treating degree attainment by age 23 as the outcome of interest. Once again, educational expansion is evident, with increases in degree attainment for students from all backgrounds. However, in contrast with the staying on results, educational inequality has risen in all periods. Young people from the poorest income groups have increased their graduation rate by just 3 percentage points between 1981 and the late 1990s, compared with a rise in graduation rates of 26 percentage points for those with the richest 20 percent of parents. 5

This pattern is also found when an alternative data source is used.

11

Figure 2: Degree Completion by Age 23 by Parental Income Group 0.5

0.46 0.45

0.4

0.37

0.37

0.35

0.3 0.3 Poorest 20 percent at age 16 Richest 20 percent at age 16 Difference (Educational Inequality)

0.25

0.2 0.2

0.14

0.15

0.09

0.1

0.07

0.06 0.05

0 1958 Cohort (NCDS)

1970 Cohort (BCS)

Late 1970s cohort (BHPS)

Proportion of samples obtaining a degree by age 23

Source: Blanden, Gregg and Machin (2005) Table 5.4.

The clear conclusion here is that the expansion of higher education in the UK has benefited those from richer backgrounds far more than poorer young people. This occurred over a period when means-tested student support declined sharply in the UK. This evidence is a cautionary tale in the recent debate around the introduction of topup fees for Universities in England and Wales. In the past, increasing the numbers of students has failed to increase the participation of the poorest groups, it is crucial that this situation changes for further expansion of higher education and this is going to require commitments to provide more generous grants, bursaries and other measures to widen participation. We have found a strong increase in the relationship between educational outcomes and parental income between the children reaching 16 in the mid-1970s and mid-1980s. For a more recent cohort there is a further strengthening of the relationship between family income and degree performance but a weakening for staying on at 16. What is less clear is the extent to which the relationships between education and income are causal, in other words whether ‘money matters’ or whether it is that richer families produce more educated children because parental education, motivation and other aspects of family culture differ. Separating the effect of income from the impact of other aspects of the family is a difficult identification problem. In

12

our research we have utilized a number of techniques which net out permanent differences in income (which will be related to factors such as parents’ education), to focus on transitory differences in income and their impact on educational outcomes. This can be done in a number of ways focusing on differences across siblings or across time for the same child. Overall, the results of this investigation provide consistent evidence of a significant causal impact of family income on educational attainment. The results suggest that a one third reduction in income from the mean increases the probability of a child getting no A-C GCSEs by around 3 to 4 percentage points, on average, and reduces the chances of achieving a degree by a similar magnitude (Blanden and Gregg, 2004). Unfortunately it is not possible to judge if the causal effect of income on education has changed across cohorts as our most stringent models cannot be applied consistently across all datasets. While it is clear that family income differences between the rich and the poor do have a substantial impact on children’s educational outcome, the estimated impact of income is modest relative to the large differences in attainment between children from richer and poorer families. Consequently, while reducing child poverty can have some beneficial effects, policies to increase intergenerational mobility will need to focus on raising poorer children’s attainment through targeted services and access to the best schools. Conclusion Social mobility and equality of opportunity have once again become issues of political and social concern in the recent past. Our research has highlighted the decline in intergenerational income mobility in Britain over the last few generations of school leavers. The wider focus of our research is to understand better whether the extent of intergenerational mobility seen in Britain is mirrored in other developed countries and to measure the role of education in this process. The research we have described offers a comparable benchmark of Britain’s performance in this dimension. The results show that Britain and the United States have the lowest levels of crossgeneration mobility, lying well below Canada and the Nordic countries. Education has been often seen as a route to greater intergenerational mobility. So it is natural to ask what role education has in the recent decline in mobility in Britain and whether it can help explain why mobility has not fallen but remained constant in other countries like the US. Our research highlights how the relationship

13

between family income and children’s higher education attainment has grown between cohorts completing education in the 1970s and the late 1990s. This implies that the big expansion in university participation has tended to benefit children from affluent families more and thus reinforced immobility across generations. Income inequality has risen at the same time as the gap between the educational attainments of the richest and poorest has grown. Our evidence indicates that income and educational attainment are causally related, through research that tests this hypothesis using a number of approaches to control for observed family differences and to isolate the impact of transitory income shifts. However, they also indicate that equalizing educational attainments by redistribution alone would be unrealistic. To improve this situation we need also to use more direct means such as early years’ education, improved schools for poor communities and financial support to pursue post-compulsory education. Indeed, this is the policy direction that the Government seems to be taking through programmes like Sure Start, Excellence in Cities and the Educational Maintenance Allowance (EMA). It is important, however, that all policies used have solid evaluation strategies, as has been pursued with these examples, so we can improve our knowledge of what really works (see Machin and Vignoles, 2005). However, the low level of intergenerational mobility for children in Britain means that current extent of policy development is currently insufficient for the task at hand.

14

References and Supporting Papers: Björklund, A. B. Bratsberg, T. Eriksson, M. Jäntti, R. Naylor, O. Raaum, K. Roed and E. Österbacka (2005) ‘Intergenerational earnings mobility in the Nordic countries, the United Kingdom and the United States: An overview’. Mimeo Åbo Akademi University, Finland. Blanden, J. (2005) Essays on Intergenerational Mobility and Its Variation over Time, Place and Family Structure, PhD Thesis, University of London. Blanden, J. and P. Gregg (2004) ‘Family Income and Educational Attainment: A Review of Approaches and Evidence for Britain’, Oxford Review of Economic Policy, 2, 245 – 263. Blanden, J. and S. Machin (2004) ‘Educational Inequality and the Expansion of UK Higher Education’, Scottish Journal of Political Economy, 51, 230-249. Blanden, J, P.Gregg and S. Machin (2005) ‘Educational Inequality and Intergenerational Mobility’ in What’s the Good of Education? The Economics of Education in the United Kingdom, eds. S. Machin and A. Vignoles, Princeton University Press. Blanden, J, A. Goodman, P. Gregg and S. Machin (2004) ‘Changes in Intergenerational Mobility in Britain’ in M. Corak ed. Generational Income Mobility in North America and Europe, Cambridge University Press. Corak, M. (2004) ‘Do Poor Children Become Poor Adults? Lessons for Public Policy from a Cross Country Comparison of Generational Earnings Mobility.’ Mimeo UNICEF Innocenti Research Centre. An expanded version of the introduction to M. Corak (ed.) Generational Income Mobility in North America and Europe, Cambridge University Press. Hertiz, T. (2004) ‘Rags, Riches and Race: The Intergenerational Economic Mobility of Black and White Families in the United States’ in S. Bowles, H. Gintis and M. Osborne (eds) Unequal Chances: Family Background and Economic Success New York: Russell Sage and Princeton University Press. Machin, S. and A. Vignoles (eds.) (2005) What’s the Good of Education? The Economics of Education in the United Kingdom, Princeton University Press.

15

Addendum, Response to Comments and Criticisms from Stephen Gorard: Reconciling Cross Country and Over-Time Comparisons of Intergenerational Mobility in the UK This addendum provides further details on the findings described in the report ‘Intergenerational Mobility in Europe and North America’. This is in response to comments and criticisms from Stephen Gorard, and the aim is to offer more clarity about comparing estimates based on different measures of parental economic status. In Table 2 of the report we show comparisons of intergenerational mobility across countries. These do not refer to the same cohorts and time periods across all countries, nor are they necessarily based upon the same statistical models (some relate earnings to father’s earnings; some relate earnings to parental income). An alternative way to present the results is to break down this analysis into two sets of consistently defined comparisons. Neither of these is able to include all the countries, but when taken together they confirm the picture that the US and UK are less mobile than the other countries under consideration. Table A1 shows details of the data sources. The estimated intergenerational parameters are given in Table A2. Results in Panel 1 are for older cohorts of sons compared to those in Panel 2. Panel 1 shows correlations of earnings with father’s earnings; Panel 2 shows correlations of earnings with parental income. The use of father's earnings or parental income as an independent variable clearly matters for interpretation. It is important to ensure that one is comparing like with like.6 Indeed, estimated parameters are different for the two independent variables, usually being higher at a point in time for models using father’s earnings rather than parental income as the independent variable of interest. International Comparison For data reasons, to do with availability of data on father’s earnings and/or parental income, in the first period (shown in Panel 1 of Table A2) the US, UK and West Germany are compared with the Nordic countries, and in the second period (shown in Panel 2 of Table A2) Canada is compared with the US, UK and West Germany. The US and West Germany can be compared in both periods as the available data spans 6

Running a regression (or computing a correlation between) of son’s earnings on father’s earnings is not the same as a regression of son’s earnings on parents’ combined earnings plus non-labour income except under very special circumstances (e.g. if only father’s worked and there was no non-labour income in the household). In the absence of these special circumstances, it is simply not correct, and would be unscientific, to carry out comparisons based on the two different models. The implications of this are considered in Blanden (2005, Chapter 3) and in Blanden, Goodman, Gregg and Machin (2004).

16

both time frames. For the UK, data from two cohort studies (the NCDS 1958 cohort and the BCS 1970 cohort) are available, enabling comparisons to be made between the UK and other countries in each time period. This, however, does NOT mean that this Table can be used to make comparisons of changes over time in the UK since the independent variables differ across the Panels (consistently defined changes over time are discussed below).

Finding 1: Table A2 shows, as stated in the report, the extent of intergenerational mobility for sons is lowest in the UK and US, is at intermediate levels for West Germany and is highest for the Scandinavian countries.

The point being made here is that is important to compare like with like. If non-comparable estimates are compared then the conclusions reached can be misleading. For example, if the NCDS 1958 parental income estimate of 0.166 (given in Table 5 of the report) was used in the international comparisons, then mobility in the UK would be at a similar level to what is observed for the Nordic countries. This comparison would be misplaced however, as the Nordic estimates are derived from specifications use fathers’ earnings as the independent variable. It should be noted that the ranking we find in Table A2 is completely consistent with comprehensive scholarly reviews of the intergenerational literature found in Solon (2002) and Corak (2004). Changes over Time in the UK In the results shown in the first panel of Table A2, an average measure of son’s earnings is regressed on a single measure of father’s earnings. In the second panel, a single measure of son’s earnings is regressed on an averaged measure of parental income. It is not legitimate to compare the NCDS and BCS results in the Table to obtain a picture of changes over time for the UK. The difference between using father’s earnings and parental income is particularly important. Table 5 of the report (reproduced below) shows changes over time in the extent of intergenerational mobility in the UK. We compare estimates from the same datasets used for the UK in panels 1 and 2 of Table A2, but change the estimation approach used to ensure that the comparison across time is legitimate. In Table 5 we report estimates for the two cohorts on a consistent basis, looking at the relationship

17

between son’s earnings in his early 30s and parental income when the son was aged 16.

Finding 2: When considered on a consistent basis (relating sons’ earnings to parental income) the intergenerational association increases over time. The estimate for the first (NCDS) cohort reported in Table 5 of the report is substantially lower than the one shown in Table A2 (.166 in Table 5 of the report compared to .260 in Panel 1 of Table A2). When this estimate of 0.166 is squared up against the comparable estimate for the BCS cohort (.286 in Table 5) it is very clear that intergenerational mobility for sons has fallen over time. Robustness In an ideal world, it would be reassuring if results were proven to be robust to using alternative datasets to consider the same research questions. However, research into transmissions of incomes across generations is challenging precisely because good data on parents’ incomes and children’s outcomes is rare. It is however, possible to compare results on changes in the relationship between educational outcomes and family incomes with those from another dataset. In Figure 1 of the report we show that the impact of parental income on staying on beyond age 16 increases between the 1958 and 1970 cohorts. In Blanden, Gregg and Machin (2005) we repeat this analysis using data from the Family Expenditure Survey for the relevant years when the cohort members were leaving school. We find exactly the same pattern. This is further reassuring evidence that the patterns between family income and children’s outcomes found in the cohort studies are genuine.

18

Table A1: Samples Available for Internationally Comparable Estimates of Intergenerational Mobility Country Dataset Data for older cohorts: UK National Child Development Study Norway Register data Denmark Register data Sweden Register data Finland Quinquennial census panel Data which spans cohorts: US Panel Study of Income Dynamics

West Germany

Socio-Economic Panel

Sons Born

Sons Observed

Measure of Parental Status

1958

1991 and 2000

Fathers’ Earnings 1974

1958 1958-1960 1962 1958-1960

1992 and 1999 1998 and 2000 1996 and 1999 1958 and 1960

Father’s Earnings 1974 Father’s Earnings 1980 Father’s Earnings 1975 Father’s Earnings 1975

1954-1970

1995 and 1999 (or age 30 for the BCS comparable results) 1995 and 1999 (or 2000 for the BCS comparable results)

Father’s earnings 1978 Parental Income son age 10 and age 16

1960-1973

Data for younger cohorts: UK British Cohort Study

1970

2000

Canada

1967-1970

1998

Intergenerational Income Data (from tax registers)

Father’s earnings 1986 Parental Income 1984 and 1988

Parental Income 1980 and 1986 Parental Income son age 1418

Table A2: International Estimates of Intergenerational Mobility Panel 1 Earlier Cohorts - Fathers’ Single-Year Earnings as Measure of Status – Two year average of son’s earnings. Country Sons Born Partial Correlation US 1954-1970 .3483 UK 1958 .2601 W. Germany 1960-73 .1803 1958-1960 .1471 Finland 1962 .1431 Sweden 1960-1973 .1431 Denmark Norway 1958 .1391 Panel 2 Later Cohorts - Parental Income Average as Measure of Status – Single Year Measure of son’s earnings Country Sons Born Partial Correlation US 1954-1970 .2892 UK 1970 .2712 W. Germany 1960-1973 .1712 Canada 1967-1970 .1432 Notes: 1Bjorklund et al Table 3 2 Blanden Table 3.3

3

Additional estimates.

19

Table 5: Changes in Intergenerational Mobility in Britain NCDS 1958 .205 (.026)

BCS 1970 .291 (.025)

Change .085 (.036)

Mobility Parameter Partial Correlation .166 (.021) .286 (.025) .119 (.033) Sample Size 2163 1976 Source: Blanden (2004) Table 4.2 Standard errors are shown in parentheses. This table is reproduced from the main report.

Addendum References Blanden, J. (2005) Essays on Intergenerational Mobility and Its Variation over Time, Place and Family Structure, PhD Thesis, University of London. Blanden, J, P. Gregg and S. Machin (2005) ‘Educational Inequality and Intergenerational Mobility’ in What’s the Good of Education? The Economics of Education in the United Kingdom, eds. S. Machin and A. Vignoles, Princeton University Press. Blanden, J, A. Goodman, P. Gregg and S. Machin (2004) ‘Changes in Intergenerational Mobility in Britain’ in M. Corak ed. Generational Income Mobility in North America and Europe, Cambridge University Press. Solon (2002) ‘Cross-Country Differences in Intergenerational Earnings Mobility’, Journal of Economic Perspectives, Vol. 16, pp. 59-66. Corak, M. (2004) ‘Do Poor Children Become Poor Adults? Lessons for Public Policy from a Cross Country Comparison of Generational Earnings Mobility.’ Mimeo UNICEF Innocenti Research Centre. An expanded version of the introduction to M. Corak (ed.) Generational Income Mobility in North America and Europe, Cambridge University Press.

Jo Blanden, Paul Gregg, Stephen Machin June 15 2005

20