The dynamics of second job holding in Britain

Transcription

1 The dynamics of second job holding in Britain René Böheim Department of Economics Johannes Kepler University Linz, Austria Mark P. Taylor Institute for Social and Economic Research University of Essex,UK June 10, 2003 Abstract In Britain, about 10 per cent of workers have a second job. Possible motives for holding a second job are hours constraints on the first job or different utilities derived from the first and second job. Alternatively, second job holding may reflect hedging behaviour in an uncertain environment. The holding of a second job may be triggered by changes in family circumstances or negative financial shocks. We describe the dynamics of second job holding in Britain during the 1990s using panel data from the British Household Panel Survey. Our results show that second job holding is surprisingly persistent over time. We find that negative financial shocks trigger second job holding, but second jobs are not a measure to smooth labour supply over time. Hours constraints are of lesser importance in motivating second job holding than heterogenous job characteristics. Keywords: second job, labour supply, BHPS. JEL classification: J2, J6. Thanks to Rudolf Winter-Ebmer for comments on an earlier version.

2 1 Introduction A significant number of workers hold a second job, Kimmel and Conway (2001) state that about 6% of US men had a second job in 1993, and Paxson and Sicherman (1994) state that about 20% of working men in the US had a second job in any one year between 1976 and (See also Averett (2001).) The number of people who have a second job in the USA has increased since the 1980s. In Europe, the evidence of second job holding is more scarce. Bell et al. (1997) state that about 10% of workers hold a second job in Britain and that the number of workers who have a second job is increasing. 1 For Germany, Schwarze and Heineck (2001) report the proportion of second job holders to be 6% of all employed workers. There are three main reasons for second job holding. The first is that a worker would like to work more in her job but cannot supply more labour. She is hours constrained and needs to work in a second job if she wants to optimise her labour supply. The second reason for working in a second job may arise because of complementarities of the two jobs, an example is the professional who also works as a consultant. A third explanation for working in a second job is uncertainty. A worker may work in a second job if she believes that the first job has a high risk of termination. Workers who fear losing their first job may hedge their chances by diversifying their human capital into new jobs. However, Bell et al. (1997) have found little evidence of hedging behaviour of this type. It is important for both policy makers and the (labour) economist to understand who holds a second job, and for what motive. Kimmel and Conway (2001) find for the USA that those who hold a second job typically work full-time on their first job where they earn a lower wage and work fewer hours than those who do not have a second job. If predominantly poor people hold a second job, then we may see this as evidence that they have main jobs which are not flexible and that they may be forced into second jobs. It also raises questions about the working poor, and whether or not a second job may lift them out of poverty. There is recent evidence about hours constraints in Britain. In previous work, we have found that 40% of men and women in paid employment prefer to work a different number of hours at their current wage, and about 8% wish to increase their number of working hours, part-time workers in particular would like to work more hours (Böheim and Taylor, 2002). This constraints may result from changed 1 Bell et al. (1997), who focus on the link between job security and multiple job holding, state that a replication of their [Conway and Kimmel (1994)] study for the United Kingdom would be useful (p8). 1

3 labour supply preferences, for example caused by negative financial shocks. Such shocks may motivate workers to increase their labour supply beyond previous levels. In the absence of hours constraints in the first job, a second job may allow the worker to pursue activities which may otherwise be unavailable to her. For parents, and especially lone parents, having two jobs may allow them more flexibility in combining work and family responsibilities (Plewes and Stinson, 1991). For example, a worker may have one job while her child is at school, and another in the evenings when her partner is available for child care. We are also interested in the dynamics of second job holding. In the medium-term, workers who wish to work more hours can either renegotiate the employment contract or change employers to supply their preferred number of hours. Job changers are more able than job stayers to adjust their work hours in line with their preferences, suggesting that restrictions on work hours in jobs are important (Böheim and Taylor, 2002). There is little evidence on the dynamics of second job holding. Studies of labour supply typically focus on hours in the first job only (Altonji and Paxson, 1988; Steward and Swaffield, 1997). The recent implementation of the the European Working Directive in the UK (October 1998) provides another motivation for looking at the dynamics of second job holding: the maximum number of hours worked in an average week is limited to 48 hours per week. Workers who wish to supply high levels of labour may be forced to taking up a second job. We test these three explanations of second job holding empirically. Using panel data from the British Household Panel Survey we investigate labour supply decisions over the course of a decade. In a first step, we investigate if workers earn more in their first job than in their second job. If wage rates are lower in the second jobs, than this is evidence for an inflexible labour market. If the opposite is true, i.e. if wages are higher in the second jobs, than we need to know why workers do not stop working in their first job altogether. By using the panel nature of the data, we are able to follow workers over time and see whether or not second jobs are a temporary phenomenon (an additional security), a stepping stone towards a job which allows the supply of more hours, or a transition from the first job to the second job. Our estimations show that first and second jobs are not substitutes, but jobs of different quality. A novel finding of our analysis is that second jobs are not a temporary measure to adjust for fluctuations in labour supply. These fluctuations are in some part triggered by financial shocks, as our estimations show, but second job holding is persistent over time. Even if we focus on two-year periods, more than half of those who had a second job in the beginning of the period will have a 2

4 second job at the end of the period. 3

5 2 Theoretical framework Each worker s labour supply is assumed to result from utility maximisation. Since jobs are not equal, the number of hours worked in the first job, h 1, the number of hours worked in the second job, h 2, and the hours of leisure, l, enter the utility function separately: U = U(h 1, h 2, l; C), (1) where C denotes consumption. 2 budget and time constraint: The utility function is maximised subject to a C = w 1 h 1 + w 2 h 2 + Y, and (2) T = h 1 + h 2 + l. The wages in the first (second) job are denoted as w 1 (w 2 ), income from other sources is denoted as Y, and T denotes the time endowment. Combining these three equations results in the following expression: max h 1,h 2 U(w 1 h 1 + w 2 h 2 + Y, h 1, h 2, T h 1 h 2 ). (3) If the worker would like to supply more hours than she is able to, then she is constrained and h 1 is no longer a choice variable. The decision to work in a second job, given that the maximum number of hours h 1 have already been supplied to the first job, will depend on the marginal utility of working in the second job. In the optimum, the relationship between the two jobs are determined by the marginal disutility of working and the wage rate in the second job: U/ h 2 U/ l U/ C = w 2. (4) The numerator U/ h 2 U/ l is the marginal disutility from an extra hour of work in the second job. Rearranging yields the condition for hours supplied in the second job for those who are constrained in their first job: h 2 = h c 2(w 2, (w 1 + w 2 ) h 1 + Y, h 1 ). (5) If leisure is a normal good, than the derivative of h 2 / Y < 0. The sign of h 2 / w 2 is ambiguous because of income and substitution effects. 2 The theoretical model follows Shishko and Rostker (1976) and Conway and Kimmel (1998). 4

6 If it is true that workers hold a second job because the jobs differ in utility or cost, i.e. it is not the constraint in the first job that causes second job holding, then there are two relationships to consider because the number of hours in the first job are not constrained: U/ h 1 U/ l U/ C U/ h 2 U/ l U/ C = w 1, (6) = w 2. These two equations imply that labour is supplied to either job until the disutility of working in that job is equal to the negative wage of that job. (If the jobs do not differ, than the worker will supply all labour to the job that has the higher wage rate.) Solving the model implies that there are two labour supply equations, h 1 = h n 1(w 1, w 2, Y ), (7) h 2 = h n 2(w 1, w 2, Y ). If leisure is a normal good, than both h 1 / Y and h 2 / Y are positive. Because of income and substitution effects, the partial derivatives of the hours supplied with respect to the wages, h 1 / w 1 and h 2 / w 2, are ambiguous in sign. Conway and Kimmel (1998) state that under standard assumptions these partial derivatives are negative. Distinguishing between those who are constrained in their first job and those who are not leaves us with four different groups, as each of the first two may or may not decide to work in a second job (Conway and Kimmel, 1998). Of interest here are the two groups of workers who decide to work in a second job. How do these two groups differ? Those who are constrained in their first job and decide to work in a second job will have on average shorter tenure in the second job than those who are not constrained in their first job. The reason is that eventually those who are constrained in their labour supply decision will move employers to optimise their hours supplied. Further, wages in the second jobs will be less (or equal) to the wages in the first job if they are constrained. If, on the other hand, the reason for holding a second job is differences in utility, then we will not expect to see any particular relationship between wages in the first and the second job. In addition, we expect to see longer tenures in both jobs. 5

7 3 Econometric specification Our empirical analyses focus on three aspects of second job holding. The first relates to the possibility of hours constraints in the first job. We estimate probit models of holding a second job where we use preferences over hours worked as one explanatory variable. 3 We expect the probability of holding a second job to be positively correlated with wanting to work more hours. We estimate two types of probit equations. The first probit equation controls for a potential selectivity bias by applying a Heckman-type two-step method. 4 The second aspect relates to the wages earned in the first and second jobs. Our data contain information on wages earned on the second job, a variable which is generally not available in other surveys, e.g. Conway and Kimmel (1998). We can therefore directly compare wages in first and second jobs. We estimate selectioncorrected wage equations for working in a first and second job. The variables which are thought to influence labour market participation, but not the wage, are the age of the youngest child and whether or not the worker changes address in the year prior to the interview. The additional identifying variables for the wages in the second job are preferences over hours worked in the first job. If supply constraints are an important determinant for working in a second job, then we expect these preferences to yield well determined and large coefficients. Since the BHPS provides variables of these kind, we can estimate the models directly. 6 The third question we ask ourselves relates to the dynamics of second job holding. Over time, workers will sort themselves into jobs which reflect their desired labour supply (assuming that there are firms which offer jobs with flexible hour- 3 We estimate P i [second job X i ] = Φ(X i β). The vector X i contains the explanatory variables, among these the dummy variables indicating preferences over hours supplies. All our estimations are using Stata 8. 4 We estimate P i [second job X i ] = Φ(X β + λ β λ ) and P i [employed Z i ] = Φ(Z γ) simultaneously by maximum likelihood.) Alternatively, we estimate a random-effects panel probit to explicitly exploit the panel nature of the date. 5 We also experimented with fixed-effects panel probit equations but since second job holding is quite persistent over time, see below, we did not succeed in estimating a satisfactory model. 6 Conway and Kimmel (1998) used the SIPP which does not contain information on preferences over hours worked. They employed a disequilibrium model to estimate differences between those who have a second job and those who do not. We believe that our approach, despite the potential bias which may arise by using subjective variables, is more robust than their approach. 6

8 wage packages). We therefore use the panel nature of our data to investigate the probability of starting a second job and the probability of leaving a second job. To analyse this aspect, we estimate the chances of starting (ending) a second job, given last year s second job status. This amounts to estimating a transition matrix or a two-state Markov model with heterogeneity (Boskin and Nold, 1975). We expect the correlation between subjective information such as perceived job instability and starting a second job to be positive. Another way of looking at the dynamics of hours supplied is to investigate the change of numbers in the first job, conditional on having a second job. We would expect, under the null hypothesis of no hours constraints, that workers increase their hours in the first job if they stop working in a second job. The dependent variable in this set-up is whether or not a worker has changed her hours in the main job, having held a second job in the previous year. If second jobs are only a temporary measure to overcome disequilibrium then we expect those who held a second job in the previous year to adjust their hours supplied in the first job over time. 7

9 4 Data Our analyses use data from the British Household Panel Survey (BHPS). Since 1991, this has interviewed annually a representative sample of 5,500 households containing about 10,000 persons. The same individuals are re-interviewed each year, and if they leave their original households to form new households all adult members of these new households are also interviewed. Similarly children in original households are interviewed when they reach the age of 16. The sample therefore remains broadly representative of the population of Britain as it changes through the 1990s. Nine waves of data are currently available, covering the period from 1991 to Panel data are required to enable observation of individuals desired changes in working hours at time t, and subsequent changes in labour market behaviour between t and t + 1. The relevant questions in the BHPS are asked of all those in employment (full-time, part-time, or self-employment) at each date of interview. We restrict our sample to employees of working age (16 to 65 for men and 16 to 60 for women). We exclude the self-employed as they have by definition more flexibility of choosing their number of working hours. However, some 12% of the self-employed report having a second job. Table 1: Incidence of second job holdings, by sex and year. Year Total Men (N) (2427) (2213) (2097) (2124) (2124) (2230) (2303) (2308) (17826) Women (N) (2406) (2229) (2178) (2189) (2167) (2253) (2282) (2290) (17994) Note: BHPS. Data are weighted using cross-sectional weights. Table 1 confirms earlier results: more women than men hold a second job, and there is little variation over time (Bell et al., 1997). On average, 9% of male and 10% of female employees held a second job between 1991 and The data allow us to investigate the relation between hours constraints and the holding of a second job. All working respondents were asked, Thinking about the hours you work, assuming that you would be paid the same amount per hour would 8

10 you prefer to... work fewer, work more, continue same hours. 7 Table 2 tabulates the numbers of second job holders and their stated preference over working hours. The numbers indicate that 13% of men and 17% of women who want to work more hours have a second job. Table 2: Preference over hours and second job holdings, by sex. Second job Men Women Y es Wants to work Less More Same N 1,684 2,084 No Wants to work Less More Same N 16,142 15,910 Note: BHPS. Data are weighted using cross-sectional weights. Those who want to work more hours are a significant minority: 11% of men and 14% of women who hold a second job want to work more hours. The numbers also suggest that hours constraints are not the only source of second job holding: more than 50% of workers who have a second job want to work the same hours they are currently working. The notion that constraints in the first job is an important motivation for taking up a second job is visible in Table 3. Those who have a second job work on average fewer normal hours and fewer overtime hours. The average number of hours worked in the second job is some six hours for both men and women. Second job holding is persistent. Table 4 shows that of all workers who were employed at two consecutive interviews, about 60% who had a second job in one 7 The question is asked directly after questions relating to the first (main) job. However, it is not clear whether or not the respondent relates this question to the hours worked in the first job only, or for both jobs. 9

11 Table 3: Mean hours worked per week, by sex and number of jobs. One job Two jobs Overall Men Normal hours Normal overtime hours Normal paid overtime hours Hours in second job (16142) (1684) (17826) Women Normal hours Normal overtime hours Normal paid overtime hours Hours in second job (15910) (2084) (17994) N ote: BHPS. Data are weighted using cross-sectional weights. Numbers in parentheses are sample sizes. year also had a second job in the next year. Over the course of two years, i.e. looking at three consecutive interviews, the persistence is somewhat lower, but still of a considerable degree: about half will have a second job at the beginning and at the end of the two years (not shown in Table). Table 4 also shows that those who had a second job in both years worked on average 31 hours/week in their first job and some 6 hours/week in their second job. In comparison, those who had one job in both years worked on average 35 hours/week. Those who stopped working in their second job from one year to the next had an increase of one hour/week in their first job; hours in their second job were some 5 per week. Those who took up a second job from one year to the next, on average did not increase their hours/week in their first job, and supplied some 5 hours/week in their second job. If hours restrictions in the first job are the main motivation for working in a second job we expect to see some adjustment in the hours worked over time. Workers are, at least in the medium term, free to change jobs and bargain a wage-hours bundle that corresponds to their labour supply preferences. In Table 4 we do see some evidence for hours constraints in the first job: those who stop working in a second job work more hours/week in their first job, and those who start working in a second job work on average the same number of hours in their first job in both 10

12 years. Table 4: Dynamics of second job holding, by sex. No second job t Second job t Men Women Total Men Women Total Second job t 1 Mean hours t Mean hours t Mean hours second job t Mean hours second job t N No second job t 1 Mean hours t Mean hours t Mean hours second job t N N ote: BHPS. Data are weighted using cross-sectional weights. Numbers in parentheses are sample sizes. Sample restricted to workers who were employed at two adjacent interviews. Table 5 tabulates the summary statistics for the estimating sample, by second job status. The two groups appear rather similar but for a few differences. Hourly wages in the second job are relatively high, the average is more than twice the average of wages earned in the first job. There are relatively more among those who want to work more hours who have a second job. Unsurprisingly, those who want to work fewer hours are relatively more frequent among those who do not have a second job. Following Bell et al. (1997) we also use variables in our estimations that proxy job security. First, we use the person s job tenure and the job retention rate. The job retention rate is constructed as the percentage of individuals in an occupation with less than one-year tenure (two-digit SIC). The second measure of job security is a subjective evaluation of a respondent s job security. This is taken as the answer to the question: I d like you to tell me from this card which best describes how satisfied or dissatisfied you are with that particular aspect [job satisfaction] of your job. The possible answers range from not satisfied at all to completely satisfied. Our variable is coded 1 if the worker was not satisfied, or not satisfied at all, and 0 otherwise. Another proxy variable of job security is whether or not the first job is permanent. 11

13 Table 5: Summary statistics, by second job status. No second job Second job Mean S.D. Mean S.D. Hourly wage first job a Hourly wage second job b Wants to work less more Female Age Job tenure Residential tenure Permanent contract Education Degree A-level O-level Financial development t 1, t Negative shock Positive shock Married/Cohabiting Spouse s hours/week c Household size Number of children London White Household income ( 1,000/month) Difficulties meeting housing costs N 18,522 2,238 N ote: BHPS. Data are weighted using cross-sectional weights. Numbers in parentheses are sample sizes. Sample restricted to workers who were employed at two adjacent interviews. a Sample sizes 11,446 and 1,542 due to missing values. b Sample size 1,854 due to missing values. c Set to zero if not partnered or partner does not work. 12

14 5 Estimation Results Table 6 presents the marginal effects (and the means of the explanatory variables) from two estimations of the probability of holding a second job. The first results, in Columns 2 and 3, are the results from estimating a selectivity-corrected probit on holding a second job. The selection equation uses the same controls as listed in the table plus two exclusion restrictions, the age of the youngest child and whether or not the person moved in the year prior to the interview. These two variables are thought to influence the probability of being in work, but not the probability of holding a second job. The second results, presented in Columns 4 and 5, are the marginal effects from estimating a random-effects panel probit equation. 8 The estimated coefficients show that there is a correlation between the probability of holding a second job and the wish to work more or fewer hours per week. For those who wish to work less, our estimation yields a negative correlation with holding a second job. The marginal effect in the cross-sectional probit is -2.3% and it is -12.2% in the random-effects panel probit. Those who say that they would like to work more are more likely to have a second job, although the coefficient is badly determined. However, we take these results as evidence of constraints on the labour supply. The estimated coefficient on hours supplied in the first job also supports the hypothesis that constraints on hours supplied are the (main) reason for holding a second job. Those who work more hours in their first job are less likely to hold a second job. We also consider dynamic household effects and control for the financial development over the last year. To do this, we construct a variable which is coded 1 if the respondent s subjective evaluation of the financial development between t and t, measured at t, is better than her expectation at time t 1 concerning the financial development between t 1 and t, measured at t 1. It is coded 0 otherwise and corresponds to a positive financial shock. A variable to indicate a negative financial shock is constructed in a similar fashion. The estimated coefficients on these two indicator variables are both negative, although the coefficient for a negative financial development is not significant at conventional error levels. Those who experienced a positive financial shock are less likely to work in a second job, all things being equal. 8 Due to the persistence of second job holding over time, the estimation of a fixed-effects panel probit equation was not successful as too many cases were lost because of no variation over time. 13

15 6 Conclusion Using data from from the British Household Panel Survey we have estimated models of holding a second job. The conventional hypothesis concerning second job holding rests on labour supply restrictions: those who cannot supply labour according to their preferences in their first job are forced to take up a second job. We have examined several aspects of holding a second job. First, we considered whether or not holding a second job is a response to hours constraints in the first job. Our estimates provide some evidence that this indeed the case. Those who wish to work more hours/week are more likely to hold a second job. Also, the more hours worked in the first job the less likely a worker is to work in a second job. However, second job holding is persistent over time and we consider hours constraints not satisfactory as an explanation of second job holding. Our results support rather the hypothesis that first and second jobs are different. The estimations also confirm that labour supply decisions are not independent of the labour supplied by household members. A novel finding of our analysis is that second job holding is not a temporary measure to adjust for fluctuations in labour supply. These fluctuations are in some part triggered by financial shocks, as our estimations show, but second job holding is persistent over time. Even if we focus on two-year periods, more than half of those who had a second job in the beginning of the period will have a second job at the end of the period. 14

16 Table 6: Probability of holding a second job: marginal effects. Cross-section Panel dy/dx (SE) x dy/dx x Wants to work less (0.006) more (0.007) Female (0.014) Age (0.005) Age 2 / (0.007) Job tenure (days) (0.001) Residential tenure (years) (0.001) First job Hours (0.000) Permanent (=1) (0.005) Education Degree (0.012) A-levels (0.012) O-levels (0.012) Financial development t 1, t Bad shock (0.005) Good shock (0.006) Married/Cohabiting (0.014) Partner s hours (0.001) Partner s hours 2 / (0.001) Household size (0.006) One child (0.024) Two children (0.025) Three or more children (0.036) London (0.013) White (0.013) Household income ( 1,000/m) (0.004) Housing costs difficult (0.008) N N ote: BHPS. The cross-sectional estimation is a selectivity-corrected probit equation, where the selection equation estimates the probability of being in work. Exclusion restrictions are the age of the youngest child and whether or not the person moved in the year before the interview. The estimated sample includes persons not in work. The panel estimation is a random-effects panel probit. 15

17 Table 7: Selection corrected estimation of wages in first jobs. ln wage 1 Work Coefficient (SE) Coefficient (SE) Female (0.017) (0.027) Age (0.006) (0.012) Age 2 / (0.009) (0.017) Job tenure (0.001) Hours/week (0.001) Permanent contract (0.020) Education: Degree (0.021) (0.036) A-levels (0.021) (0.043) O-levels (0.017) (0.036) Married (0.026) (0.046) Spouse s working hours (0.001) (0.002) sphour (0.002) (0.003) Household size (0.009) (0.015) 1 child (0.038) (0.059) 2 children (0.038) (0.060) 3+ children (0.042) (0.066) London (0.024) (0.048) White (0.024) (0.047) Household income ( 1,000/m) (0.011) (0.007) Housing costs (0.014) (0.025) Age youngest child (0.003) Mover (0.026) Constant (0.112) (0.195) N ,567 N ote: BHPS. The cross-sectional estimation is a selectivity-corrected probit equation, where the selection equation estimates the probability of being in work. Exclusion restrictions are the age of the youngest child and whether or not the person moved in the year before the interview. The estimated sample includes persons not in work. 16

18 Table 8: Selection corrected estimation of wages in second jobs. ln wage 2 Second job Coefficient (SE) Coefficient (SE) Female (0.089) (0.066) Age (0.034) (0.022) Age (0.046) (0.031) Job tenure (0.009) (0.005) Hours/week (0.005) (0.002) Education: Degree (0.117) (0.082) A-level (0.131) (0.093) O-level (0.111) (0.077) Married/Cohabiting (0.143) (0.100) Spouse s working hours/week (0.005) (0.004) Spouse s hours (0.008) (0.005) Household size (0.062) (0.034) 1 child (0.250) (0.120) 2 children (0.244) (0.120) 3+ children (0.235) (0.126) London (0.151) (0.092) White (0.123) (0.081) Household income ( 1,000/m) (0.028) (0.018) Housing costs (0.088) (0.055) ln wage (0.043) Wants to work: less (0.133) more (0.048) Permanent contract (0.049) Short tenure (0.003) Low job security (0.043) Short tenure * work less (0.005) Bad financial shock (0.034) Good financial shock (0.037) Age of youngest child (0.005) Moved in last year (0.054) Constant (0.669) (0.413) N ote: BHPS. The estimation is a selectivity-corrected probit equation, where the selection equation estimates the probability of having a second job. 17

19 Table 9: Probabilities of starting and stopping a second job between t and t + 1. Start Stop Coefficient (SE) Coefficient (SE) Wants to work: less (0.443) (0.433) more (0.135) (0.150) Female (0.137) (0.136) Age (0.055) (0.055) Age (0.075) (0.074) Job tenure (0.013) (0.013) Housing tenure (0.010) (0.010) Hours/week first job (0.005) (0.005) Permanent contract (0.152) (0.156) Short tenure (0.008) (0.008) Low job security (0.130) (0.126) work less * short tenure (0.016) (0.016) Degree (0.153) (0.159) A-levels (0.183) (0.188) O-levels (0.153) (0.159) Bad financial shock (0.106) (0.107) Good financial shock (0.121) (0.129) Married/Cohabiting (0.213) (0.215) Spouse s working hours/week (0.007) (0.007) Spouse s hours (0.012) (0.011) Household size (0.077) (0.075) 1 child (0.338) (0.328) 2 children (0.332) (0.324) 3+ children (0.338) (0.333) London (0.205) (0.183) White (0.196) (0.212) Household income ( 1,000/m) (0.035) (0.037) Housing costs (0.141) (0.138) Constant (0.976) (0.964) Note: BHPS. The estimation are random-effects panel regressions. 18

20 References Altonji, Joseph G and Christina H Paxson (1988), Labor supply preferences, hours constraints, and hours-wage tradeoffs, Journal of Labor Economics 6(2), Averett, Susan L (2001), Moonlighting: multiple motives and gender differences, Applied Economics 33, Bell, David N F, Robert A Hart and Robert E Wright (1997), Multiple job-holding as a hedge against unemployment, CEPR discussion paper series No Centre for Economic Policy Research, London, UK. Böheim, René and Mark P Taylor (2002), Option or obligation? The determinants of labour supply preferences in Britain, The Manchester School. forthcoming. Boskin, M J and F C Nold (1975), A Markov model of turnover in Aid for Families with dependent children, Journal of Human Resources 10, Conway, Karen Smith and Jean Kimmel (1994), Male labor supply estimates and the decision to moonlight, unpublished manuscript. W.E. Upjohn Institute for Employment Research, Kalamazoo, MI. Conway, Karen Smith and Jean Kimmel (1998), Male labor supply estimates and the decision to moonlight, Labour Economics 5, Kimmel, Jean and Karen Smith Conway (2001), Who moonlights and why? Evidence from the SIPP, Industrial Relations 40(1), Paxson, Christina H and Nachum Sicherman (1994), The dynamics of dual-job holding and job mobility, NBER working paper series No National Bureau of Economic Research, Cambridge (MA). Plewes, T J and F Stinson, Jr (1991), The measurement and significance of multiple jobholding in the United States, Statistical Journal 8, Schwarze, Johannes and Guido Heineck (2001), Auswirkungen der Einführung der Sozialversicherungspflicht für geringfügige Beschäftigung Eine Evaluation des 630-dm-jobs -reformgesetzes, DIW discussion papers 257. Deutsches Institut für Wirtschaftsforschung, Berlin. Shishko, Robert and Bernard Rostker (1976), The economics of multiple job holding, American Economic Review 66(3),

21 Steward, Mark B and Joane Swaffield (1997), Constraints on the desired hours of work of British men, Economic Journal 107,

22 A Appendix Data (Note: for referees only, not intended for publication) Main variables used in analysis Item Hours first job Hours overtime Paid overtime Hours preference Has a second job? Standard Occupational Code Type of employment in second job? Hours in second job? Income from second job(s) Question asked Thinking about your (main) job, how many hours, excluding overtime and meal breaks, are you expected to work in a normal week? And how many hours overtime do you usually work in a normal week? How much of that overtime (usually worked) is usually paid overtime? Thinking about the hours you work, assuming that you would be paid the same amount per hour would you prefer to...(work fewer hours, work more hours, continue with same hours) Do you earn any money from (a second job) odd jobs or from work that you might do from time to time (apart from your main job)? (inc baby sitting, mail order agent, pools agent etc What is it that you do (and what does the firm or person you work for make or do)? (in your second job Are you an employee or self employed? (in your second job) How many hours do you usually work a month in your second/odd job(s), excluding meal breaks but including any overtime you might do? Before tax and other deductions how much did you earn from your second and all other occasional jobs in the last calendar month? 21