Local Level Incapacity Benefits Rolls in Britain: Correlates and Convergence

Abstract

McVicar D. Local level Incapacity Benefits rolls in Britain: correlates and convergence, Regional Studies. There is considerable variation across British Local Authority Districts in the proportion of the working-age population claiming incapacity benefits, ranging from 2% in Hart to 16% in Merthyr Tydfil and Easington in 2008. These claimant rates are strongly correlated with local unemployment rates and self-reported disability rates. But spatial differences in claimant rates for incapacity benefits were even greater a decade ago. This paper describes the extent of this claimant rate convergence and discusses possible explanations for it.

McVicar D. 英格兰地方层面工作能力丧失的相关收益：相关性与收敛，区域研究。就适工年龄人口中申请工作能力丧失的比例而言，英格兰不同地方自治区之间存在着显著的差异，如 2008 年 Hart 地区的 2% 到 Merthyr Tydfil 及 Easington 地区的 16%。上述索赔率与地区失业率以及根据个体自助申报进行统计的工作能力丧失率之间存在着紧密的相关性。但是工作能力丧失索赔的空间性差异在 10 年以前更为明显。本文描述了这一索赔收敛率的程度同时探讨了对其可能的解释。

工作能力丧失收益 地方自治区 残疾 隐性失业 收敛值

McVicar D. La liste des demandeurs des prestations d'incapacité en Grande-Bretagne à l'échelle locale: la corrélation et la convergence, Regional Studies. La proportion de la population en âge de travailler qui demande des prestations d'incapacité varie sensiblement suivant les circoncriptions d'action locales, allant de 2% à Hart jusqu'à 16% à Merthyr Tydfil et à Easington en 2008. Ces proportions sont en corrélation étroite avec les taux de chômage locaux et les taux d'incapacité rapportés par les acteurs eux-mêmes. Mais la variation géographique de la proportion de demandeurs était encore plus importante il y a dix ans. Cet article présente l'importance de la convergence de la proportion de demandeurs et discute des explications éventuelles.

Prestations d'incapacité Circonscriptions d'action locales Incapacité Chômage caché Convergence

McVicar D. Invalidenrenten auf Lokalebene in Großbritannien: Korrelate und Konvergenz, Regional Studies. Der Anteil der Empfänger von Invalidenrenten im arbeitsfähigen Alter unterliegt in den verschiedenen britischen Lokalverwaltungsbezirken erheblichen Schwankungen: von 2% in Hart bis 16% in Merthyr Tydfil und Easington im Jahr 2008. Dieser Anteil hängt in hohem Maß mit der Arbeitslosigkeit vor Ort und dem Anteil der selbst gemeldeten Behinderungen zusammen. Allerdings fielen die räumlichen Unterschiede beim Anteil der Empfänger von Invalidenrenten vor einem Jahrzehnt noch höher aus. In diesem Beitrag wird der Umfang dieser Konvergenz bei den Empfängeranteilen beschrieben, und es werden mögliche Erklärungen dafür erörtert.

Invalidenrente Lokalverwaltungsbezirke Behinderung Versteckte Arbeitslosigkeit Konvergenz

McVicar D. Prestaciones por incapacidad en un ámbito local en el Reino Unido: relaciones y convergencia, Regional Studies. El porcentaje de la población en edad de trabajar que solicita prestaciones por incapacidad varía considerablemente en los diferentes distritos de las autoridades locales británicas: desde un 2% en Hart hasta un 16% en Merthyr Tydfil y Easington en 2008. Estos porcentajes están estrechamente relacionados con los índices de desempleo del lugar y las tasas de incapacidad autoinformadas. Sin embargo, las diferencias espaciales en la proporción de candidatos a prestaciones por incapacidad eran incluso superiores hace diez años. En este artículo describimos el volumen de esta convergencia de candidatos y aportamos posibles explicaciones al respecto.

Prestaciones por incapacidad Distritos de las autoridades locales Discapacidad Desempleo oculto Convergencia

Keywords:

• Incapacity Benefits
• Local Authority Districts
• Disability
• Hidden unemployment
• Convergence

JEL classifications:

• I38
• J23

Acknowledgements

Thanks are extended to Michael Anyadike-Danes, Steve Fothergill, Anne Green and David Webster for helpful comments on an earlier draft of this paper; and to David Webster (and originally the Department for Work and Pensions) for the data on Pathways to Work roll out dates. The usual disclaimer applies.

Notes

In some cases local labour markets, for example as defined by travel-to-work areas, may spill over LAD boundaries. This can lead to spatial correlation in LAD unemployment rates and also, by implication, in LAD incapacity benefits claiming rates. The implications of such spatial correlation are considered in the following sections.

The Disability Discrimination Act defines a disabled person as someone who has a physical or mental impairment that has a substantial and long-term adverse effect on his or her ability to carry out normal day-to-day activities.

Although the results presented here are based on disability data from 1999 and 2008, all results are robust to pooling across 1999/2000 and 2007/08/09 given sample size concerns at the LAD level.

The LFS/APS contains data on unemployment using the alternative International Labour Organization (ILO) definition – out of work, wanting to work, actively looking for work in the last four weeks – but there are some missing values at the LAD level because the LFS/APS suppresses data where numbers are so small as to be potentially unreliable. Residence-based claimant count unemployment rates refer to the proportion of the working-age resident population in receipt of unemployment benefits. Note that this differs from the more standard ‘proportion of the labour force’ definition of the claimant count.

Thanks to David Webster for sending the author these data, which were originally sourced from the DWP.

The higher claiming rate for men partly reflects the gender difference in pension age, with a cohort of men aged sixty to sixty-four on IB for which there is no comparable group of women. When restricted to the sixteen to fifty-nine age group, the female claiming rate is closer to that for males (for example, Anyadike-Danes and McVicar, 2008).

Female claimant rates are more highly correlated with male unemployment rates than with female unemployment rates. For further discussion of the interrelated nature of male and female local labour markets in this context, see Beatty et al. (2009).

This has fallen from 35.6% in 1999, with most of the working-age self-reported disabled who do not claim incapacity benefits being in employment. (This latter group is referred to as the ‘hidden sick’ by Beatty et al., 2000.) Inasmuch as there is consensus in the literature, it is that changes over time in the incapacity benefits claiming rate reflect changes in the proportion of the disabled that are out of work more than changes in the proportion of the working-age population that report a disability for both the UK and USA (McVicar, 2008).

For a discussion of the pros and cons of proxying for disability with mortality rates or other ‘objective’ health measures, see Bound (1991).

Given the difference in sample means the elasticities (magnitudes expressed in percentage change terms) for men and women are both around 0.5. These estimates are not particularly robust to replacing the claimant count unemployment rate with the ILO unemployment rate. The least missing values for this alternative unemployment rate measure are for the data on males and females together, for which the corresponding elasticities for self-reported disability prevalence and unemployment are closer to 1.0 and 0.25, respectively. The estimates are, however, reasonably robust to the inclusion of regional dummies intended to capture common omitted factors across proximate LADs. When these are included the coefficients on self-reported disability prevalence fall by around one-third but the coefficients on the unemployment rate are very similar. Perhaps unsurprisingly the largest regional dummy coefficients are for Wales, Scotland, the Northeast and Northwest. The conclusions are also robust to exploring possible non-linearity by including squared terms in unemployment rates and disability prevalence. Finally, testing for spatial independence in incapacity benefits claimant rates (using Moran's I, Geary's c and Getis and Ord's G tests) suggests one cannot rule out spatial dependence in the error structure. Estimates presented in Table 3, however, are highly robust to spatial regression extensions to allow for either spatial lags (for example, if stigma from claiming is higher in LADs bordered by low-claiming neighbours) or spatial autocorrelation in the error term (for example, if there are common cultural influences across LAD boundaries).

Elasticities and R 2s are broadly similar when estimated on 1999 cross-section data. Further, if self-reported disability prevalence is replaced by 1997–1999 male life expectancy, one obtains similarly plausible results, with negative and significant coefficients on life expectancy in each case and positive and significant coefficients on unemployment rates. The unemployment rate coefficients are smaller, however, by an order of magnitude for females and for all, but not for males.

This is one possible explanation for the smaller elasticity on ILO unemployment rates compared with claimant count unemployment rates, given that the LFS/APS is based on small samples at the LAD level.

Hausman tests for exogeneity of self-reported disability prevalence returned the following chi-square statistics (p-values): 6.54 (0.038), 29.8 (0.000) and 13.6 (0.001) for male, females and all, respectively. Equivalent tests for exogeneity of unemployment rates return the following statistics (p-values): 35.4 (0.000), 19.5 (0.000) and 32.3 (0.000), respectively. These tests suggest both variables should be treated as endogenous in all three cases.

These results are robust to the inclusion or otherwise of regional dummies in terms of sign and statistical significance, although the coefficients on unemployment rates are lower for women and for all by close to an order of magnitude when these dummies are included.

For a detailed discussion of this point in the context of the economic growth literature, see Quah (1993).

The growth literature parallel here is the difference between beta convergence (the negative correlation of growth rates with country initial income levels) and sigma convergence (falling cross-sectional dispersion in country incomes over time).

An alternative (equilibrium labour markets type) explanation of claimant rate convergence is that incapacity benefits claimants have been migrating from the ‘hotspots’ to the Greater Southeast. One would expect to see convergence in median earnings across LADs, however, if this were the dominant factor.

The growth literature parallel is beta convergence, and Quah's (1993) Galton fallacy critique applies.

These estimates are similarly robust to spatial regression extensions to allow for either spatial lags or spatial autocorrelation in the error term and to inclusion of a squared term in the 1999 incapacity benefits claimant rate.

The cohort hypothesis discussed below – that spatial variation in claimant rates is driven by a particular cohort of ex-miners and ex-industrial workers that have recently reached retirement age – is one possible example.

Hausman tests for exogeneity of the 1999 claimant level returned the following chi-square statistics (p-values): 50.1 (0.000), 56.8 (0.000) and 69.2 (0.000) for male, females and all, respectively. Equivalent tests for exogeneity of the change in self-reported disability prevalence return the following respective statistics (p-values): 4.83 (0.305), 3.40 (0.494) and 1.09 (0.897); for change in unemployment rates there are 18.0 (0.001), 9.96 (0.019) and 18.5 (0.001); and for PtW there are 1.32 (0.859), 2.87 (0.579) and 0.15 (0.997).