Does Performance Pay Increase Job Satisfaction?
This paper investigates the influence of performance-related pay on several dimensions of job satisfaction. In cross-sectional estimates performance-related pay is associated with increased overall satisfaction, satisfaction with pay, satisfaction with job security and satisfaction with hours. It appears to be negatively associated with satisfaction with the work itself; yet, after accounting for worker fixed effects the positive associations remain and the negative association vanishes. These results appear robust to a variety of alternative specifications and support the notion that performance-related pay allows increased opportunities for worker optimization and does not generally demotivate workers or crowd out intrinsic motivation.
The use of performance pay schemes by employers has been shown to increase workers' productivity, effort and earnings (Lazear 2000; Paarsch and Shearer 2000; Parent 1999). However, it remains unclear a priori what effect performance pay schemes have on worker satisfaction with the job. While increased earnings will increase worker satisfaction, other aspects of performance pay schemes may have less beneficial effects on job satisfaction. Pay schemes based on performance may introduce large variations in periodic earnings, reducing the utility of risk-averse workers. The performance monitoring associated with pay schemes may result in increased effort that workers dislike. While some types of performance pay (such as profit sharing) may increase job security, others will increase earnings dispersion within the firm and may reduce perceptions of fairness or lower morale and motivation. In this way, performance pay schemes may increase worker satisfaction with pay while reducing their satisfaction with other dimensions of the job, such as effort, risk or perceived fairness.
This paper presents evidence on the impact of performance pay schemes on job satisfaction in the United Kingdom. Specifically, we use the British Household Panel Survey (BHPS) to investigate the impact of profit sharing, bonuses and performance pay on several dimensions of job satisfaction, including overall satisfaction and satisfaction with pay, with hours, with job security and with the work itself. The paper provides the first evidence of the influence of performance pay schemes on several of these different aspects of job satisfaction. In traditional cross-section estimates, we demonstrate that performance-related pay schemes are positively related to satisfaction with pay, to satisfaction with job security and to a lesser extent with satisfaction with hours and overall satisfaction. Performance-related pay schemes appear negatively related to satisfaction with the work itself. Next, we utilize the panel nature of the BHPS and demonstrate that controlling for individual fixed effects confirms the positive influences of performance pay. Including the fixed effects, profit sharing and/or performance pay remain associated with higher satisfaction overall, with pay, with job security and with hours. The negative influence on satisfaction with the work itself vanishes in the fixed-effects estimates. A series of robustness checks confirm these general patterns but show that results vary by gender and by union status.
The paper is structured as follows. The next section provides an overview of evidence on the effect of payment schemes on worker outcomes with specific attention to job satisfaction. Section II describes the data source and empirical methodology. Section III presents the central econometric results contrasting the cross-section and fixed-effects estimates. Section IV describes robustness checks and presents results within several subsamples. The final section concludes.
I. PAST THEORY AND EVIDENCE
Payment schemes are seen as more closely aligning the interests of workers and firms. The critical characteristic of linking effort and pay allows a more complete optimization by workers than is possible with time rates (Heywood and Wei 2006). In the extreme, workers paid piece rates or commissions equate the marginal cost of effort with the marginal value added (see Lazear 1995). The infra-marginal effort brought forth by the piece rate generates worker earnings above the cost of effort. While this generates the maximum surplus for workers, workers paid time rates stop short of the optimal effort, failing to produce units that generate value added greater than effort. Thus, many studies confirm that those on piece rates increase effort and earn more than otherwise-equivalent workers on time rates (Ewing 1996; Lazear 2000; Oettinger 2001; Parent 1999).1 All else equal, the ability to optimize more fully should make workers more satisfied.
Yet all else may not be equal. Some factors may further increase the satisfaction associated with performance pay. Brown and Sessions (2003) suggest that workers prefer employment environments seen as rewarding their productivity, and that such environments increase worker optimism about future employment. In addition, some types of performance pay may be part of a bundle of HRM innovations associated with high-performance workplaces. Such workplaces may create greater feelings of belonging, esteem and commitment. Goddard (2001) and Bauer (2004) present evidence on the association between job satisfaction and high-performance workplaces.
There exist a number of reasons, however, why performance-related pay may reduce job satisfaction. First, performance pay may not be successful. Evaluations may be overly subjective, or objective measures of performance may be poorly tied to actual firm profit. In theses cases overall surplus may not be increased as workers try to maximize an objective such as the favour of their supervisor or the number of pieces produced (Baker 1992). In this second case workers may increase productivity in response to piece rates but may not increase profitability—as, for instance, quality falls, maintenance deteriorates or injuries increase (Freeman and Kleiner 2005). Similarly, schemes that rely on a supervisor's evaluation are known to suffer from a variety of biases that may cause them to fail to increase surplus (Prendergast 1999). Some schemes have also been shown to undermine valuable teamwork and co-operation (Drago and Garvey 1998). Schemes with such malfunctioning characteristics might well be anticipated to lower worker satisfaction.
Second, Gibbons (1987) formalized the traditional union fear that ‘ratcheting’ would lower rewards and incentives after workers responded with additional effort. Certainly Green (2004) has shown that the use of performance pay is associated with work intensification in Britain. Case studies of call centres by Drago (1996) and Fernie and Metcalf (1999) present bleak pictures of low wages and high stress made worse by computerized monitoring and piece rates. Thus, performance pay may be merely a disciplinary tool that does not allow greater optimization but merely increases work effort and lowers satisfaction, especially for the low skilled (for more on the negative consequences for the low skilled see McCausland et al. 2005). Even group schemes and profit sharing designed to change workplace norms may increase not only effort and earnings but also peer pressure. Kandel and Lazear (1992) emphasize that this peer pressure can be so severe as to actually lower worker utility (job satisfaction) even as earnings rise.
Third, workers may suffer reduced satisfaction from the increased earnings risk associated with performance pay schemes that may not be fully built into compensating wage demands (Milgrom and Roberts 1992). Moreover, typical job satisfaction estimates hold constant earnings, suggesting that the direct negative effect of the earnings risk may be reflected in lower satisfaction. Yet the role of sorting by risk can be critical. Sliwka and Grund (2006) have shown that workers who have a greater tolerance for risk are significantly more likely to choose jobs receiving performance pay. Thus, the fact that performance pay is associated with greater earnings risk need not imply that the workers receiving performance pay are less satisfied with their earnings risk. Moreover, a counterbalance to this increased risk occurs when the variation in workers' payments matches that of firm performance. Profit sharing has therefore been recognized as increasing the job security of workers because worker payments fall during times of low profitability, which reduces the chance of layoff (see Weitzman 1984, and Kruse 1993). Thus, profit sharing may simultaneously increase earnings risk but reduce the likelihood of job loss.
Fourth, greater pay dispersion typically results from individual performance pay schemes, and Kennedy (1995) shows that such schemes reduce the morale of the least productive workers and reduce their effort (productivity). Indeed, his model shows that this effect can be sufficient to overwhelm the increase of effort among the most productive, and to actually lower average productivity compared with time rates. In studying the implementation of performance pay, Marsden et al. (2001) provide statistical evidence of just such reductions in productivity among those with low performance pay increments. This, they say, is combined with the ‘demotivating effect arising from difficulties of measuring and evaluating performance fairly’. Personnel management texts routinely stress that avoiding excessively inequitable pay rates is important in creating harmony and productivity (see the studies cited in Akerlof and Yellen 1988). Brown (2001) demonstrates that workers who believe their payment methods are ‘fair’ report higher satisfaction with their pay. Thus, regardless of the influence on total surplus, it is possible that some types of performance pay will be perceived as unfair and will be associated with reduced job satisfaction.
Finally, Frey and Jegen (2001) review the literature and show that extrinsic incentives such as performance pay can crowd out intrinsic motivation to do a good job. Again, they claim this crowding effect can actually dominate the traditional effect, resulting in lower utility and productivity. Even if it does not dominate the traditional effect, it may show up in reduced satisfaction with intrinsic aspects of the job such as satisfaction with the job itself. Thus, the influence of performance pay on job satisfaction cannot be determined by theory. To the extent that it allows better optimization and an increase in surplus, it should increase satisfaction. Yet influences on risk, effort, morale and intrinsic satisfaction could reverse this suggestion. Understanding the net influence of performance pay on job satisfaction remains important, as job satisfaction has been shown to be closely correlated both with worker effort and with the intention to quit (Clark 2001).
Gender potentially confounds any study of performance pay and job satisfaction. Goldin (1986) argues that women routinely sort, and are sorted, into jobs with piece rates. This view contends that women are less motivated by deferred compensation and so require contemporaneous rewards for productivity. Yet Becker (1985) implies that women, and mothers in particular, will avoid performance pay in general because of the relatively high levels of effort they are devoting to home and children. Thus, Geddes and Heywood (2003) present empirical evidence that, while women are disproportionately paid by the piece, they remain less likely to receive performance pay overall as they are significantly less likely to receive commissions, bonuses or merit pay. Therefore, any estimates attempting to isolate the effects on job satisfaction itemized above should be careful not only to control for gender, but also to examine the genders separately and attempt to control for home responsibilities.
Two papers directly estimate the impact of performance pay schemes on over all job satisfaction. McCausland et al. (2005) use waves 8–11 of the BHPS to contrast the pay satisfaction of those receiving performance pay with that of employees not receiving it. They use an econometric framework that endogenizes wages and accounts for self-selection into the method of pay. They show that for more highly paid workers satisfaction with the job and with pay are both higher under performance pay. On the other hand, for lower paid workers satisfaction is lower under performance pay. McCausland and his co-authors suggest that the difference between the two groups of workers may result if lower paid workers are more likely to see performance pay as a form of monitoring or control. While this certainly follows, it may also be the result of the de-motivation of lower-ability workers modelled by Kennedy (1995) and observed by Marsden et al. (2001).
Two additional points deserve notice. First, the definition of performance pay adopted by McCausland et al. (2005) does not include profit sharing. Indeed, because profit sharing is asked as a separate question on payment method in the BHPS, estimating the determinants of performance-related pay versus all other methods would include profit sharing among the base group of other methods. If some of the influences of profit sharing and other performance-related pay are similar, they could be obscured as a consequence. Second, while the BHPS examined fewer dimensions of job satisfaction by wave 8, it continued to examine several beyond overall job satisfaction and satisfaction with pay.2 While it was not the focus of the study by McCausland et al., using these dimensions might help shed light on some of the conflicting theories described above. For instance, satisfaction with job security might be useful in unravelling the role of performance pay on risk and so on satisfaction.
Heywood and Wei (2006) use the National Longitudinal Study of Youth to examine the influence of both profit sharing and individual performance pay in the United States. They show that both profit sharing and individual performance pay are associated with greater job satisfaction and greater satisfaction with pay. Profit sharing is also associated with greater satisfaction with the worker's supervisor. Interestingly, within the measure of individual performance pay there is a suggestion that piece rates may reduce overall satisfaction even as it increases satisfaction with pay. Neither the broad measure of individual performance pay nor profit sharing were correlated with satisfaction with co-workers, although studies of German workers (Heywood et al. 2005a, b) show that profit sharing tends to positively influence satisfaction with both supervisors and with co-workers. While these studies control for earnings, they do not try to correct for selection.
In what follows, we return to the BHPS to examine the determinants of job satisfaction. We examine those determinants for all of the dimensions of job satisfaction available in the data and simultaneously examine the influence of both performance pay and profit sharing.
II. DATA AND METHODOLOGY
The data used in this paper are drawn from the British Household Panel Survey (BHPS). The BHPS is a nationally representative sample that each year interviews approximately 10,000 individuals from roughly 5500 households (McCausland et al. 2005). We use the waves of the BHPS corresponding to 1998–2004, as earlier waves do not contain information on pay schemes. We restrict our sample to those individuals aged 20–65 and not self-employed. This yields an unbalanced panel of 11,849 individuals.
The information on payment schemes in the BHPS is available for 1998 onwards, and the questions asked are:
In the last 12 months have you received any bonuses such as a Christmas or quarterly bonus, profit-related pay or profit sharing bonus, or an occasional commission? [this excludes overtime payments]; and Does your pay include performance-related pay? (Taylor et al. 2006)
From these two questions, we create three mutually exclusive variables: (i) the individual received profit shares/bonuses only, (ii) the individual received performance-related pay only and (iii) the individual received both profit shares/bonuses and performance pay. Our data do not identify the proportion of earnings attributable to different pay schemes, a limitation shared with most individual and establishment data sets (see Heywood et al. 1998).
All job satisfaction questions in the BHPS are reported on a seven-value Likert scale, 1 being the least satisfied, 7 the most satisfied. At different times a variety of job satisfaction questions have been included in the BHPS, but for the period in which pay scheme information is available five job satisfaction questions were used: (i) overall job satisfaction, (ii) satisfaction with pay, (iii) satisfaction with hours worked, (iv) satisfaction with job security and (v) satisfaction with the work itself. Table 1 provides mean job satisfaction levels for these five categories, disaggregated by the type of pay scheme. For brevity, and to avoid double-counting individuals, we report these for the 1998 sample only.3
SUMMARY STATISTICS, PAY SCHEME AND JOB SATISFACTION: EMPLOYEES AGED 20–65, 1998
|Profit sharing/bonuses only (996)||5.31||5.02||5.08||5.48||5.38|
|Performance pay only (332)||5.21||4.89||5.26||5.19||5.37|
|Profit sharing/bonuses and performance pay (502)||5.20||5.03||5.00||5.35||5.23|
|No. of individuals|| || ||4399|| || |
Overall job satisfaction is highest for those who do not receive any form of performance pay or profit sharing/bonuses with those who receive profit sharing or bonuses only coming second. Satisfaction with pay is the highest and nearly identical for those both receiving profit sharing/bonuses only and for those receiving these in conjunction with performance pay. Performance pay appears to be associated with lower levels of satisfaction with job security. In general, the pattern of unconditional means reported in Table 1 does not appear to support a strong link between job satisfaction and performance pay schemes.
Table 2 presents the correlations between the five measures of job satisfaction. While the correlations between overall job satisfaction and the other dimensions are reasonably large, the correlations among those other dimensions are sufficiently small to indicate that they measure different aspects of satisfaction. Thus, it makes sense from a statistical as well as a theoretical point of view to estimate the determinants of the individual dimensions separately.4
CORRELATIONS BETWEEN DIFFERENT DIMENSIONS OF JOB SATISFACTION, 1998–2004
|Overall|| || || || |
|Hours||0.511|| || || |
|Pay||0.483||0.345|| || |
|Work itself||0.701||0.430||0.340|| |
A number of variables are available in the BHPS that allow us to control for other sources of variation in work conditions, many of which we would expect to influence job satisfaction. Thus, we observe the workers' pay rate (log pay), the number of hours worked and whether they worked overtime. When estimating the impact of performance pay schemes on job satisfaction it is important to control for these factors, as, from the discussion in the previous section, we would expect these to vary with pay scheme. Other included variables relating to job characteristics are whether the individual had a work pension; whether the employer provided health insurance; whether the employer provided training; whether the individual was promoted in the last year; whether the individual was on a temporary contract; whether the individual had a management or supervisory role; and size of the firm.
Table 3 presents sample means and standard deviations of the independent variables. The variables included are largely standard in the literature on job satisfaction. Again, for brevity sample means for the first wave used (1998) are presented in this table, with full pooled (i.e. multiple individual–time observations) sample means presented in the Appendix (Table A2). The variables include the controls for work conditions listed above, gender, whether the respondent is in a union and whether the respondent has a disability. Also included, but not reported in the sample means, are 10 controls for industry, 9 controls for occupation and 11 controls for region.
SUMMARY STATISTICS: EMPLOYEES AGED 20–65, 1998
|Profit sharing/bonuses only*||0.226|| |
|Performance pay only*||0.075|| |
|Profit sharing/bonuses and performance pay*||0.114|| |
|Highest level of education|
| < A level||0.541|| |
| A level||0.216|| |
|Higher degree||0.029|| |
|Normal hours worked||35.907||9.327|
|Union member||0.319|| |
|Public-sector worker||0.303|| |
|Temporary job||0.050|| |
|Promoted in last year||0.075|| |
|Employer provided health insurance||0.104|| |
|Employer provided training||0.210|| |
|Firm size||0.320|| |
| 1–24 workers|| || |
| 25–99 workers||0.264|| |
| 100–499 workers||0.247|| |
|No. of observations||4399|| |
Following past research, the values of job satisfaction are fitted to the cumulative normal distribution through ordered probit estimates (see e.g. Clark and Oswald 1996 and Clark et al. 1997). The ordered probit estimation follows appropriately when the dependent variable has a natural ordering, such as least to most satisfied (see McKelvey and Zavonia 1975).
III. CENTRAL RESULTS
Table 4 presents estimates of job satisfaction using data pooled for 1998–2004, with standard errors clustered at the individual level. These provide initial results upon which the more satisfactory fixed-effects estimates will build. The first column provides estimates for overall job satisfaction. Estimates for controls are largely as demonstrated in previous research. Job satisfaction declines with education level and firm size. Union members are less satisfied, while public-sector workers are more satisfied with their jobs. Females appear to have markedly higher job satisfaction than males, confirming past findings of the ‘paradox of the contented female worker’ (Bender et al. 2005).
DIMENSIONS OF JOB SATISFACTION: COVARIATE ESTIMATES, 1998–2004
|Profit sharing/bonuses only||0.053*||0.030***||0.095*||−0.011||0.115*|
|Performance pay only||−0.0003||0.021||0.071*||−0.043***||0.042***|
|Profit sharing/bonuses and||−0.0002||−0.017||0.098*||−0.024||0.042***|
|Normal hours worked||−0.005**||−0.026*||−0.003||−0.002||0.003|
|Promoted in last year||0.167*||0.105*||0.102*||0.148*||0.131*|
|Employer provided health||0.047***||0.016||0.170*||0.001||−0.057**|
|Employer provided training||0.070*||0.046*||−0.015||0.083*||0.028***|
|No. of observations|| || ||43,215|| || |
In terms of performance pay measures, only those receiving profit sharing or bonuses have a significantly higher level of overall job satisfaction. Columns (2) to (5) display covariate estimates for the various dimensions of job satisfaction reported in the BHPS for 1998–2004. As might be expected, performance pay measures are associated with increases in workers' satisfaction with their pay. As we control for the level of pay, this result goes beyond the recognized influence that performance pay has in increasing earnings and is consistent with an improved ability to more nearly optimize the trade-off of effort and earnings. There is also an indication that profit sharing or bonuses are associated with higher levels of satisfaction with work hours. Again, work hours are accounted for and, as anticipated, workers have lower satisfaction with their hours of work when those hours are greater. The point is that, holding the hours of work constant, those who receive profit shares or bonuses report greater satisfaction with their hours. Taken together, these two results suggest that profit sharing or bonus schemes allow workers to choose effort and pay combinations more in line with their preferences and so generate greater job satisfaction.
We suggested conflicting theoretical influences upon satisfaction with job security. Profit sharing has been recognized to create greater job security by reducing marginal costs of employment when profits are low. This seems born out by the highly significant and large influence of profit sharing on greater job satisfaction with security. The influence of performance pay is less clear. An emphasis on output-based measures might imply a greater chance of being dismissed on the basis of under-performance. On the other hand, the ability to avoid being dismissed may more nearly be in the hands of workers who can influence their output. Importantly, we also know that workers with less aversion to risk sort into jobs with performance-related pay. Column (6) shows that performance-related pay is associated with greater satisfaction with job security, although these are only statistically significant at the 10% level.
Finally, there appears to be a weakly significant negative effect of performance pay on satisfaction with the job itself. As this estimate also holds pay constant, it may reflect the tendency of performance pay schemes to crowd out intrinsic satisfaction. Alternatively, it may reflect the tendency for jobs that are easily monitored and paid by performance to be simple and repetitive (MacLeod and Parent 1999). Nonetheless, this one result does not change the overall pattern in which performance pay and profit sharing tend to have positive influences across a variety of dimensions of job satisfaction.
Not only do the performance pay variables generate a series of statistically significant coefficients, but the magnitudes of the effects are also important. In Table 5 the marginal effects on the probability of reporting the highest level of satisfaction are reported for each of the dimensions of job satisfaction.5 Thus, the first entry indicates that receiving only profit sharing/bonuses increases the likelihood of being in the most satisfied category by one percentage point. This is not a small effect, as the mean level is around 11%. Thus, a one percentage point increase represents an increase of (1/11)th relative to the mean, or a finding that those receiving profit sharing/bonuses are 9% more likely to report being in the most satisfied category. The magnitudes for the other payment schemes and dimensions vary but are largest for pay and for job security.
MARGINAL EFFECTS ON THE PROBABILITY OF BEING IN THE ‘MOST SATISFIED’ CATEGORY, 1998–2004
|Profit sharing/bonuses only||0.010*||0.006***||0.015*||−0.003||0.035*|
|Performance pay only||−0.00006||0.004||0.011*||−0.010***||0.013***|
|Profit sharing/bonuses and||−0.00004||−0.004||0.015*||–0.006||0.013***|
|Proportion of sample in most satisfied category||0.114||0.140||0.085||0.159||0.231|
The estimates presented so far have assumed that individuals are randomly assigned into different pay schemes. There is evidence, however, that this is not the case (McCausland et al. 2005). Lazear (2000) suggests that 56% of the increase in productivity associated with one form of performance-related pay, i.e. piece rates, is associated with the sorting of inherently more productive workers into firms operating these types of pay scheme. In the context of the present paper, this is important in so far as any estimated effect of performance-related pay on job satisfaction may merely reflect the sorting of workers with preferences for performance-related pay to these types of job. In particular, if more productive workers are more satisfied under any pay setting but are attracted to performance pay settings, as Lazear suggests, then the apparent influence of performance pay on satisfaction will be misleading.
The suggestion that fixed effects can be critical in estimating job satisfaction has been confirmed by Heywood et al. (2002), who showed that the dissatisfaction of union workers with their pay in a cross-section is eliminated in fixed-effects estimates that concentrate on workers changing union status. Such fixed effects may also influence the cross-sectional estimates for the facets reported in Table 4. Thus, fixed effects will largely hold risk aversion constant, allowing the actual role of performance pay on satisfaction with security to emerge. Similarly, the unmeasured skills of workers limited to routine and easily monitored jobs can be held constant to focus on the actual role of performance pay on satisfaction with the job itself. In essence, the panel estimates remove the role of sorting by unmeasured worker fixed effects.
To control for unobserved worker-specific effects on job satisfaction, we estimate fixed-effects ordered probits using the within-worker variation across the seven waves.6 The specification used in Table 4 is again used, but time-invariant worker characteristics such as gender will necessarily be omitted from the estimation.
Table 6 provides the fixed-effects estimates of job satisfaction. For brevity, only the coefficients for the pay scheme type are reported. They generally support the earlier estimates. Column (2) confirms the earlier reported significant positive effect of profit sharing/bonuses on overall job satisfaction and now shows a positive coefficient on performance pay that just misses traditional significant tests (tstat=1.59). Moreover, all three of the indicators of payment schemes remain positive and significant indicators of satisfaction with pay and of satisfaction with job security, as they did in the cross-section. Thus, risk sorting does not appear to be driving the job security results in the cross-section. In the cross-sectional estimates both performance pay and profit sharing/bonuses positively influenced satisfaction with hours. Controlling for individual fixed effects, they remain positive, but now it is the coefficient on performance pay schemes that reaches statistical significance. The previously negative influence of performance pay on satisfaction with the work itself is not apparent in the fixed-effects estimate. Such a pattern is consistent with the conjecture that workers who are subject to certain types of performance pay (such as piece rates) may be limited to jobs that tend to be more repetitive and simple, and that once these fixed effects are held constant the nature of the payment schemes does not play an independent role.
PERFORMANCE-RELATED PAY AND JOB SATISFACTION: FIXED-EFFECTS ESTIMATES, 1998–2004
|Profit sharing/bonuses only||0.047**||0.017||0.097*||–0.027||0.106*|
|Performance pay only||0.051||0.073**||0.161*||0.045||0.099*|
|Profit sharing/bonuses and||0.005||–0.031||0.123*||–0.030||0.134*|
|No. of observations||43,215|| || || || |
In total, these results suggest two specific points. First, performance-related pay schemes are significant determinants of several different aspects of job satisfaction. Second, estimates of the influence of pay schemes on job satisfaction that fail to control for individual specific effects may misrepresent the influence on a number of these aspects.
More generally, the results provide no evidence that performance pay schemes de-motivate workers, lower morale or crowd out intrinsic motivation. While these things may happen in specific cases, they are certainly not sufficiently general to lower any of the measured dimensions of job satisfaction. Instead, the bulk of the evidence is consistent with performance pay improving opportunities to optimize effort and reward trade-offs, improving the sense of fairness in rewards and improving job security. In this sense, the results provide important corroboration of those of Heywood and Wei (2006) and of McCausland et al. (2005), who found important positive effects of performance pay on satisfaction, but for more general measures of satisfaction (fewer dimensions) and for more limited samples (only the more highly paid).
IV. ROBUSTNESS CHECKS AND SUBSAMPLE RESULTS
Before moving on to specific checks and subsamples, we emphasize a general result. All of the estimates of the influence of performance-related pay schemes on all dimensions of job satisfaction are robust to the exclusion of controls for hours worked (and overtime) and pay. While hours itself often matters for satisfaction, it does not interact with performance-related pay in a fashion that alters the results reported in the previous section. This invariance goes in part to the issue of whether longer hours are associated with payment schemes and so might lower satisfaction. We found no evidence of such a suggestion. Performance-related pay is known to increase earnings, and one might argue that this influence should not be held constant in examining the influence of performance-related pay on satisfaction. Indeed, the size and significance of the coefficients generally increase when earnings are excluded as a determinant of satisfaction.
The decision to adopt performance-related pay may be relevant only for certain production technologies, and hence only for portions of the workforce. Equally importantly, the exact nature of performance pay (merit schemes, piece rates, etc.) may be specific to particular production technologies. While our data limit our ability to investigate this, we estimate fixed-effects models for all dimensions of job satisfaction separately by industrial sector: primary sector, manufacturing/construction sector and service sector. These estimates (available on request) demonstrate that workers in manufacturing/construction receive large and significant increases in overall job satisfaction from performance pay. Estimates for the primary sector are positive but insignificant, which may be due to a relatively small sample size. For service sector workers we cannot confirm an influence of performance pay on overall job satisfaction. These results hint that choices over performance pay and types of scheme may be related to production technologies, and that these differences influence satisfaction. This being said, performance-related pay does significantly increase satisfaction with pay for both manufacturing/construction and service sector workers.
McCausland et al. (2005) claim that for lower paid workers performance pay schemes merely act as a discipline and monitoring device. This, they suggest, explains their empirical result that performance pay reduces job satisfaction for the lower paid. We examine this contention by splitting the sample according to wage. Specifically, we stratify the sample by whether the worker was above or below the median wage in the sample for each year. We then estimate fixed-effects models of all dimensions of job satisfaction for the lower-income panel. These estimates (available upon request) indicate that pay schemes have no significant impact on overall job satisfaction for these workers. Despite the reduced sample size, all three categories of pay schemes remain associated with increased job satisfaction with pay (even holding pay constant in the estimates). Moreover, profit sharing remains associated with significantly increased satisfaction with job security. Interestingly, profit sharing is associated with significantly reduced satisfaction with the job itself.7 In total, these results are not suggestive of overall negative effects on morale or motivation of performance-related pay schemes for low paid workers.
A further factor to consider when examining the role of performance-related pay schemes on satisfaction with hours worked is that this may vary markedly according to a worker's skill level. Specifically, the consequences of ratchet effects and work intensification may be concentrated among the lower skilled. We examine such claims by estimating the model for satisfaction with hours worked separately for low-skill workers. There are two main ways one might think of identifying low-skill workers in our data. First, we could focus on workers with low education levels. Estimates based on samples of workers without A levels (n=22,047), or workers without post-school qualifications (n=31,451) reveal no systematic effect of performance pay schemes on satisfaction with hours worked. Second, we look at workers in low-skill occupations by identifying workers in the three lowest skilled occupational groups (sales and customer service occupations; process, plant and machine operatives; and elementary occupations). Again, estimates based on this sample (n=8864) show no statistically significant relationship between satisfaction with hours and performance-related pay schemes.
In light of the large variations in job satisfaction by both gender and union status displayed in Table 4, it is worth considering whether there are variations in the impact of performance-related pay on job satisfaction across both these groups. Table 7 provides estimates of the impact of performance-related pay on job satisfaction split by gender and union membership, respectively. In both cases, all controls are as in Table 4, and in addition controls for individual fixed effects are included.
PERFORMANCE-RELATED PAY, JOB SATISFACTION BY GENDER AND UNION STATUS: FIXED-EFFECTS ESTIMATES, 1998–2004
|Profit sharing/bonuses only||0.050***||0.050||0.031||0.016||0.116*||0.071**||–0.020||–0.041||0.119*||0.112*|
|Performance pay only||0.061||0.048||0.139*||–0.003||0.190*||0.129*||0.022||0.068||0.149*||0.056|
|Profit sharing/bonuses and||0.018||–0.008||–0.039||0.017||0.124*||0.133*||–0.048||–0.001||0.166*||0.123*|
|No. of observations||21,286||21,928|| || || || || || || || |
|Profit sharing/bonuses only||0.128*||0.030||0.022||0.025||0.118*||0.087*||0.050||–0.033||0.021||0.141*|
|Performance pay only||0.050||0.117*||0.158*||0.052||0.283*||0.128*||0.078||0.069||0.149*||0.093**|
|Profit sharing/bonuses and||0.120**||–0.012||0.073||–0.053||0.295**||0.049||0.045||–0.022||0.218*||0.145*|
|No. of observations||14,478||25,737|| || || || || || || || |
The results stratified by gender reveal a few differences. First, profit sharing increases job satisfaction for men, but not women. Performance pay has a positive effect on job satisfaction with hours, pay and job security for men, but only the pay effect is apparent for women. Profit sharing increases satisfaction with pay and job security for men, but only job security for women. Together, these results suggest that performance-related pay schemes are more desirable for men. To the extent that such schemes create increased competition and rivalry within the workplace, this might be anticipated on the basis of several experiments showing that competition improves the performance of males but not of females (Gneezy and Rustichini 2004). It also remains possible that women have more responsibilities at home, which limit their ability or desire to participate in performance pay schemes. Indeed, as a final note on the role of gender, we re-estimated the fixed-effects equations separately for women with children at home and women without children at home. The significant positive results in determining satisfaction with pay and with security are unique to the subsample of women without children at home. While small sample sizes hurt efficiency, no significant coefficients emerge in the subsample of women with children at home.
The results in the second panel demonstrate that there are marked differences between union and non-union workers in terms of the effect of performance-related pay schemes on job satisfaction. For instance, profit sharing increases the latent measure of overall job satisfaction by almost 0.13 for union workers, while receiving both profit shares and performance pay increases the latent measure of job satisfaction by 0.12. For non-union workers no such effects are noticeable, while only performance pay increases their overall job satisfaction. Performance-related pay schemes increase satisfaction with pay for both groups, but the magnitude of the effect is much larger for union workers who receive only performance pay or in conjunction with profit sharing. Performance pay increases satisfaction with hours for unionized workers, while profit shares increase satisfaction with security for non-union workers.
The remarkably strong role for performance pay among unionized workers may at first blush seem surprising. Unions are presumed to oppose individual performance pay because it hurts the solidarity wage and increases earnings dispersion among workers doing the same job (Issac 2001). At the same time, unions are presumed to oppose explicit profit sharing as it replaces the implicit profit sharing that takes place during bargaining (Drago and Heywood 1995). Yet the international evidence on the association between unions and performance pay is best described as inconclusive. Moreover, US and Canadian evidence suggests that, when unions are involved in the creation and implementation of performance pay schemes, the schemes are associated with better firm performance and greater longevity (Kim and Voos 1997; Kim 1999). German evidence makes clear that active involvement by organized labor (in the form of works councils) actually increases the chance of a firm adopting performance pay (Heywood et al. 1998). Thus, the finding that performance pay is associated with increased job satisfaction for union members may not be paradoxical if the schemes in the union sector are more likely to be both accepted by workers and successful because of the workers' role in its creation.
This work expands on previous research by investigating the role of performance pay in determining many separate dimensions of job satisfaction. Moreover, it uniquely controls for individual fixed effects when examining the dimensions of job satisfaction. Indeed, the only previous study using longitudinal data and fixed-effects estimates examined just overall satisfaction (Heywood and Wei 2006), even as the individual dimensions prove critical for examining competing hypotheses. This study is also unique in the extent to which subsamples are separately examined by gender, union status, presence of children and occupation. It is also novel in allowing interactions between performance-related pay and profit sharing.
We have provided evidence that profit sharing/bonuses tend to increase overall job satisfaction. Moreover, performance-related pay increases satisfaction with both pay and job security. The latter finding is not necessarily intuitive. Performance-related pay may decrease job security in so far as it is indicative of a culture of monitoring work effort. Conversely, linking pay to productivity may increase job security as wages fluctuate positively with the output of the firm (Weitzman 1984; Kruse 1993), reducing the need for firms to lay off workers in periods of weak product demand. It may also attract workers who are willing to tolerate risk and so are more likely to be satisfied with their degree of security. Our findings suggest that the latter two effects dominate.
A concern with performance-related pay is that it can lead to work intensification (Green 2004), and this in turn may lead to dissatisfaction with hours worked. In this study we found no evidence of performance-related pay adversely affecting satisfaction with hours worked, even for low skilled workers whom it has been suggested are adversely affected by performance pay schemes. Indeed, in the fixed-effects estimates we found evidence of greater satisfaction with hours among those receiving performance pay.
A related concern is that the explicit incentives of performance-related pay may crowd out intrinsic motivations. In the cross-section estimates there was, indeed, a suggestion that performance-related pay was associated with reduced satisfaction with the job itself. Yet the fixed-effects estimates revealed that this was the result of sorting, as the association did not persist. Thus, we remain unable to confirm any negative influences of performance pay on job satisfaction, and unable to dislodge a series of positive influences.
Several caveats remain. First, these general tendencies do not mean that the job satisfaction of all workers will increase should their firms adopt performance pay. By its nature, performance pay is suited for some types of production technologies and not for others. Thus, we emphasized in our early discussion that performance pay can, in some circumstances, be counter-productive and can decrease surplus (Freeman and Kleiner 2005). It makes sense that workers in such cases may not enjoy increased satisfaction. Indeed, we presented evidence hinting that the influence of performance pay on satisfaction may be less evident in service industries. Second, other dimensions of job satisfaction may still present negative correlations. Satisfaction with management, co-workers or stress may all be lowered by performance pay. We simply do not have access to those dimensions in our data. Finally, we recognize that our measures of performance pay may aggregate individual practices that have offsetting influences. Thus, piece rates may lower satisfaction even as earnings based on a broader formal appraisal increase satisfaction. We cannot identify whether or not such differences exist. Despite these caveats, the many suggestions that worker welfare will be reduced by performance pay received no support in our inquiry.
Tables A1 and A2 present summary statistics for the full pooled sample, 1998–2004.
SUMMARY STATISTICS, PAY SCHEME AND JOB SATISFACTION: EMPLOYEES AGED 20–65, 1998–2004
|Profit sharing/bonuses only (9318)||5.35||5.02||5.16||5.54||5.39|
|Performance pay only (2865)||5.30||4.98||5.15||5.53||5.34|
|Profit sharing/bonuses and performance pay (3539)||5.25||5.08||5.04||5.43||5.32|
|Samples size|| || ||43,215|| || |
SAMPLE STATISTICS: EMPLOYEES AGED 20–65, 1998–2004
|Profit sharing/bonuses only||0.216|| |
|Performance pay only||0.066|| |
|Profit sharing/bonuses and performance pay||0.089|| |
|Highest level of education|| || |
| < A level||0.510|| |
| A level||0.218|| |
|Higher degree||0.037|| |
|Normal hours worked||35.221||9.561|
|Union member||0.335|| |
|Public-sector worker||0.341|| |
|Temporary Job||0.042|| |
|Promoted in last year||0.065|| |
|Employer provided health insurance||0.090|| |
|Employer provided training||0.198|| |
|Firm size|| || |
| 1–24 workers||0.327|| |
| 25–99 workers||0.267|| |
| 100–499 workers||0.230|| |
| 500 workers plus||0.176|| |
|No. of observations||43,215|| |
We thank the Lancaster University Management School for the support of its Visiting Professor Programme that brought Heywood to Lancaster in the spring of 2006 and allowed us to work together on this project. We also express gratitude to two anonymous referees for helpful comments on an earlier draft.
1. This is not to say that in all circumstances piece rates automatically have this influence. Inability to measure performance or incompatible managerial policies can, for instance, make piece rates unprofitable (Freeman and Kleiner 2005).
2. Earlier waves also included questions on worker satisfaction with promotion prospects, relations with the boss and their use of initiative.
3. Means for the full pooled sample covering 1998–2003 are reported in Table A1 in the Appendix.
4. Indeed, one can use the Cronbach's alpha to show that the individual dimensions should not be aggregated to a single measure (Heywood et al. 2002).
5. As there are seven satisfaction categories, three payment scheme variables and five dimensions of job satisfaction, we have not presented a full set of marginal effects, but they are available from the authors upon request.
6. Specifically, we estimate ordered probits with fixed effects in Limdep 8.0.
7. This could result if the type of jobs and industries in which profit sharing is offered to the lower paid workers differs from those in which it is offered to higher paid workers. To examine whether profit sharing for the lower paid is concentrated in particular jobs, we investigated the incidence of profit sharing/bonuses for the our lower paid sample across 1-digit UK Standard Industrial Classification (SIC) codes. We found only limited variation in the incidence of profit sharing/bonuses at this level of aggregation and it did not differ greatly from the general sample. Among the lower paid, the highest incidence is 30.8% for those in the transport and communication and the lowest is 12.9% for those in ‘other’ services. We recognize that greater detail might show more variation.