Low education High education
Constant -4.775 (0.359)** -3.852 (0.351)** Dummy for 1981 0.238 (0.135) 0.243 (0.140) Dummy for 1983 -0.203 (0.141) -0.006 (0.134) Dummy for 1984 -0.085 (0.144) 0.050 (0.140) Dummy for 1985 -0.219 (0.149) -0.126 (0.142) Dummy for 1987 0.097 (0.134) 0.110 (0.132) Dummy for 1988 -0.081 (0.138) 0.022 (0.133) Dummy for 1989 0.287 (0.129)* 0.212 (0.128) Dummy for 1990 -0.085 (0.143) -0.033 (0.135) Age 16 - 24 1.425 (0.159)** 1.322 (0.135)** Age 25 - 34 0.471 (0.180)* 0.688 (0.140)** Age 25 - 34 *D81 -0.151 (0.220) -0.196 (0.195) Age 25 - 34 *D83 0.291 (0.238) -0.071 (0.188) Age 25 - 34 *D84 -0.072 (0.259) -0.123 (0.197) Age 25 - 34 *D85 0.288 (0.252) 0.302 (0.191) Age 25 - 34 *D87 0.029 (0.238) 0.135 (0.182) Age 25 - 34 *D88 0.449 (0.231) 0.126 (0.186) Age 25 - 34 *D89 0.328 (0.218) -0.013 (0.181) Age 25 - 34 *D90 0.291 (0.239) 0.040 (0.189) Age 50 - 64 -0.710 (0.093)** -0.913 (0.100)** Single -0.100 (0.144) -0.062 (0.115) Working wife -0.635 (0.064)** -0.521 (0.054)** Dependant child -0.138 (0.073) -0.215 (0.060)** Manual HOH 0.128 (0.083) 0.105 (0.069) Skill unavailable -0.185 (0.182) 0.219 (0.156) Energy -0.354 (0.353) -0.190 (0.309) Manufacture 0.506 (0.294) 0.081 (0.296) Construction 0.438 (0.298) -0.333 (0.305) Services 0.872 (0.250)* 0.326 (0.266) Industry unavailable 1.272 (0.345)** 0.629 (0.342) Council tenant -2.923 (0.370)** -4.611 (0.524)** Private tenant 1.119 (0.726) 0.180 (0.712) Unemployed 0.793 (0.153)** 0.920 (0.156)**
Log regional relative wage -4.446 (1.200)** -4.065 (0.947)**
Log regional relative wage *D81 -0.728 (1.664) -0.001 (1.524)
Log regional relative wage *D83 2.651 (1.674) 0.336 (1.328)
Log regional relative wage *D84 0.147 (1.793) -0.629 (1.376) Log regional relative wage *D85 4.032 (1.755)* 0.041 (1.303) Log regional relative wage *D87 1.878 (1.615) -1.522 (1.246) Log regional relative wage *D88 3.608 (1.568)* 0.174 (1.201)
Log regional relative wage *D89 2.546 (1.427) 0.714 (1.135)
Log regional relative wage *D90 3.228 (1.546)* 1.337 (1.180)
Unemployment differential 0.113 (0.048)* 0.100 (0.053)
Cost of unemployment -0.425 (0.106)** -0.272 (0.098)*
Relative vacancy rate -0.085 (0.160) -0.076 (0.145)
Sample size Log Likelihood Pseudo (%) 184229 -8648.53 7.53 108655 -10172.97 5.73 Notes: As notes 1-2 in Table 5.10.
allows a test of the hypothesis that all coefficients are unchanged across the two periods. This is rejected; -twice the difference of the log-likelihood’s is equal to 977.75 which is bigger than the critical value (60.19 according to the conventional chi-square statistic and 555.54 for the Schwartz
criterion).
One may expect that wages disaggregated by region and occupation are too broad to be relevant for the wage expectations each individual has^. That is, few wiU expect to receive the mean wage for their skill type. Job experience, racial/ethnic background, age, etc. will influence the individuals potential wage and marketability. Here we investigate how our results change if the level of disaggregation is increased to include age categories. That is wages now differ by, the ten standard regions, 16 OPCS occupational orders and 10 year age bands. Table A5.9 displays smaller wage effects (compared to Table 5.8) which imply some of the previous wage effects were actually age effects. The influence of aU other variables in the migration model do not change much.
5.7 Conclusions
The aim of this chapter has been to analyse the robustness of conventional models of migration in two ways. First by re-assessing the influence of housing tenure on mobility decisions and secondly by examining regional mobility models over a time period of 10 years.
Previous research has found that being in council housing deters individuals from migrating even after personal characteristics have been taken into account. However there is a case for believing council tenants are a self-selected group having characteristics that make them less likely to migrate. This self-selectivity has probably been re-enforced with changes in housing policy (in particular the 1980 Housing Act ( ‘Right to Buy’)) that transformed public housing into housing of last resort. As individuals in housing of last resort are less likely to have the preferences and characteristics often seen in migrants, housing tenure can be seen as endogenous to the migration
Ideally one would like to have panel data with past wage levels as this takes into account the individuals observable and unobservable earnings capacity.
decision. The standard method of dealing with such endogeneity is to instrument the housing tenure variables. Our results illustrate, even after we have use an instmmental variable strategy, we continue to find that hving in council housing significantly reduces the likelihood of changing regions.
Another data problem is encountered at the modelling stage regarding the non-random availability of skill and industry which may be correlated with the dependent variable. Testing the pooled model we arrive at the conclusion that the associated bias is small.
The broad results we arrive at are in harmony with the findings of Pissarides and Wadsworth (1989) and Hughes and McCormick (1994). The most important result is that inter-regional mobility is sensitive to wage differentials but insensitive to unemployment differentials. This implies price signals are more important than quantity signals in generating migration. However, in 1981 a 1% rise in regional wage differentials induce a 0.5% rise in movements across regions whereas in 1990 a 1% rise induces only a 0.2% rise in regional labour flows.
Analysis of migration over the business cycle reveals it is lower when the economy is in a recession but has no additional impact on wages. That is the sensitivity of migration to wages does not change with macroeconomic indicators. Given that the wage effect is not cychcal but has a smaller impact on regional mobility by the decade imphes there is no clear evidence of increased labour market flexibihty.
Chapter 6- Conclusions
Labour mobility is an important channel through which the economy can adjust to structural and cychcal changes. Labour flows to high wage regions and occupations is evidence that labour is responding to market signals in a timely fashion. During the 1980’s, the British economy experienced a large rise in wage dispersion. This thesis has used econometric techniques to ascertain how the significant rise in dispersion translates into changes in relative wages between regions, occupations and other observable characteristics. It also examines the extent to which male workers responded to relative wages in the eighties. The main finding is that relative wages do influence regional, upward and outward mobility^*.
The principle conclusion reached in Chapter 3 was little of the rise in national wage inequality came from increased wage disparity between regions. Most of the rise was due to increased inequality within regions. The rise within each region however, was not uniform. Some regions experienced large changes in inequality and others small changes. Decomposing the national and regional inequality trends, it was found that about 50% of the rise in wage dispersion was due to unobservable wage components. The contribution of changes in the composition of labour towards higher skill and education had a substantial affect on the inequahty trend but the contribution of increased returns to skill and education was negligible.
The econometric model developed in chapter 4 reveals upward and outward mobility depend on
individual characteristics, occupational earnings differentials and local labour market variables. Non-white individuals were found to be more likely to move outward and less likely to move upward whereas single individuals were more likely to move both upward and outwards. Outward mobility was responsive to local vacancy and unemployment rates {quantity signals) and to the wage differential for being self employed {price signals). Upward mobility was
responsive to quantity signals although only the youngest age groups responded to price signals. One way to interpret such responses is in terms of transaction costs or structural impediments to mobility in the labour market. The larger the fixed transaction costs on labour market agents the smaller will be their response to market signals. Hence, the fact that occupational differentials exert a significant impact on the upward mobility of young workers is consistent with the idea that costs of mobility are lower for this group.
Investigating regional mobility in chapter 5, it was clear that price signals were more important than quantity signals in determining regional mobility. Controlling for the endogeneity of public housing led to the result that public housing deterred migration. These findings are in line with existing work on regional mobility. This chapter adds to the literature through its analysis over a series of cross-sections between 1980 to 1990. The principal conclusion arrived at was, regional mobility was less responsive to regional relative wages at the end of the decade. Although migration did shift with aggregate unemployment (therefore it was lower during the large recovery at the end of the decade), the smaller wage responsiveness was unrelated to the economic cycle. Thus, even though many of the observed structural changes in the British economy suggest that the British labour market has become more flexible (i.e. lower transaction costs), in terms of changes in regional mobility, there is no clear evidence of increased flexibility.
There are several interesting implications of the findings of this thesis for future research. The work undertaken here has concentrated on males. In recent years labour market changes have meant that female labour has become increasingly important. In the U.S., Topel (1992) shows, the increased participation of women has reduced the wages of unskilled men and contributed to the rise in overall inequahty. Similarly Blau and Kahn(1991) find in the U.S., the male female wage distributions have converged more at the bottom of the distribution. Such a
finding suggests that in order to fully understand changes in relative wages, it is necessary to examine changes in male and female wage inequality at the same time.
Our analysis of upward and outward mobility follows the traditional method of examining the influence of expected wage differentials. Other factors could be modelled in future research, such as the layoff risk or the variance of occupational and self-employment wages, both of which determine occupational choice. To elaborate, although utility may be higher with higher expected wages, the wages in these jobs may have high variances which some individuals find unattractive. The variance of wages is of particular importance, because during the 1980s there is evidence of rising wage variance within narrowly defined population subgroups.
The results of chapter 3 illustrated that the wage disparity between regions was very small from the early eighties and the late eighties,. According to traditional theories, inter-regional mobility depends on expected wage differentials, which would imply httle change in mobility over this period. Yet inequality growth within regions was large and varied across regions. The influence of these changes on regional mobility will need to take account of factors such as risk aversion and inequality aversion. The need to incorporate the wage variance and higher moments of the wage distribution exists for regional mobility as well.
Our modelling of upward and outward mobility takes into account only the observable part of wages. Unobservable skills, such as inter-personal or presentation skills are powerful in explaining why two otherwise identical individuals end up in different occupations or receive different wages. Furthermore there may be unobserved heterogeneity that affects both wages and mobihty. This problem is also relevant for our analysis of regional mobility in chapter 5 where individuals in a given region and skill are assigned the mean wage for that group. The importance of such self-selection in the UK could be assessed when adequate panel data^ becomes available.
^^The first two waves of the British Household Panel Survey (the only UK panel dataset) are insufficient as it only has positive earnings for 16 regional migrants.