Work Practices and Productivity

An extensive literature, both in management science and economics, stresses the role of 'high-involvement' and 'high-performance' workplace practices in business performance. High-performance practices seek to improve the flexibility and the quality of the production process in conjunction with ICT. High-involvement practices such as employee shareholding, profit sharing, or labour-management information sharing seek to enhance employees' motivation, engagement, and loyalty. If the spread of these practices is well documented, for example in France during the 1990s (Coutrot, 2000), their actual impact on productivity is still an unsettled issue (for a review, see Bloom and Van Reenen, 2010). The main concern is the potential reverse causality and unobserved heterogeneity in empirical estimations. The aim of this subsection is not to resolve these caveats but rather to see if, with the same estimation procedure, we can observe breaks in the relations between work practices and labour productivity, before and after the Great Recession, that may have contributed to the slowdown in productivity.

The waves of REPONSE are the only French employer surveys providing information on workplace practices before and after the shock of 2008. Managers were questioned on a large variety of practices. We select here some of the key practices that are retained in numerous studies.[1] In contrast to recent research, we did not aggregate the different practices into a single index.

More specifically, two high-involvement dimensions are used.[2] Employee shareholding is reported by managers interviewed in about one-third of the establishments in our samples. In most firms, managers are the main subscribers to shareholding schemes, but in some firms— even among large multinationals (Societe Generale, Auchan, etc.)—a large proportion of (permanent) workers hold shares. In addition, employees are the main, and even sole, shareholders of certain firms, for example cooperatives. The second dimension is the organized employee-voice groups in the workplace. We built a variable adding the implementation of regular workplace meetings and of employee- voice groups in working conditions and workplace organization. This variable is then normalized to one (thus taking on the values 0.5 or 1).

Three dimensions of high-performance practices are studied. Quality management is captured by adding managers' declarations about quality circles and total quality management (the variable is normalized). Managers are asked about job rotation and the existence of autonomous work teams as well.

All these variables are included in the estimates of the production function (1) for both 2005 and 2011. The results are presented in Table 4.7 for 2005 and in Table 4.8 for 2011. We use various specifications.

Table 4.7. Workplace practices and productivity in 2005

(1)

(2)

(3)

(4)

(5)

Mono-establishment

firms

Panel

05/11

Panel

99/05

Ln(Assets per employee)

  • 0.312***
  • (0.030)
  • 0.333***
  • (0.038)
  • 0.333***
  • (0.038)
  • 0.287***
  • (0.051)
  • 0.440***
  • (0.155)

Organized empl. voice

  • 0.124***
  • (0.032)
  • 0.106**
  • (0.042)
  • 0.123***
  • (0.052)
  • 0.074***
  • (0.023)
  • 0.130**
  • (0.060)

Empl.

shareholding

  • 0.074*
  • (0.040)
  • 0.039
  • (0.033)
  • 0.097***
  • (0.034)
  • 0.057*
  • (0.033)
  • 0.015
  • (0.041)

Quality

management

  • -0.010
  • (0.044)
  • -0.058
  • (0.058)
  • -0.026
  • (0.096)
  • 0.006
  • (0.042)

Autonomous

team

  • -0.030
  • (0.027)
  • 0.035
  • (0.040)
  • 0.003
  • (0.054)
  • -0.016
  • (0.026)

Job rotation

  • 0.002
  • (0.022)
  • -0.008
  • (0.035)
  • 0.007
  • (0.038)
  • 0.019
  • (0.021)

Organized employee voice in 1999

  • 0.009
  • (0.039)

Employee shareholding in 1999

  • 0.005
  • (0.043)

Ln (Productivity per employee in 1999)

  • 0.679***
  • (0.041)
  • 0.848***
  • (0.070)

Ln (Assets per employee in 1999)

  • -0.183***
  • (0.041)
  • -0.358***
  • (0.131)

Two-digit

industry

Yes

Yes

Yes

Yes

Yes

Other controls

Yes

Yes

Yes

Yes

No

N

1,469

531

446

1,203

463

R2

0.60

0.65

0.68

0.72

0.72

Interpretation: Controls are capital intensity, share of employees aged >fifty-five, establishment age, percentage of women, percentage of low-skilled, percentage of high-skilled workers and firm size category. Establishments with twenty or more workers in the private non-agricultural sector. Robust standard errors clustered by two-digit industry code.

***significantatthe 1 percentlevel; **significantatthe5 percentlevel,-*significantatthe10 percentlevel.

In both tables, column 1 is based on the largest sample; controls are similar to those described in Table 4.5. Estimates on mono-establishment observations are given in column 2. Column 3 provides the results of the regression for an alternative subsample: the establishments present in the REPONSE 2011, which by definition are those having survived the first years of the Great Recession and thus may have unobserved

Table 4.8. Workplace practices and productivity in 2011

(1)

(2)

(3)

(4)

(5)

Mono-estab.

Panel 05/11

Relative productivity

Ln (Assets/ employee)

  • 0.302***
  • (0.028)
  • 0.284***
  • (0.028)
  • 0.361***
  • (0.044)
  • 0.340***
  • (0.056)

Organized

employee

voice

  • 0.004
  • (0.026)
  • 0.004
  • (0.040)
  • -0.003
  • (0.025)
  • -0.022
  • (0.025)

(+) ns

Empl.

shareholding

  • -0.033
  • (0.023)
  • 0.038
  • (0.028)
  • -0.040
  • (0.026)
  • -0.009
  • (0.049)

(-) ns

Quality

management

  • -0.020
  • (0.011)
  • 0.000
  • (0.038)
  • 0.006
  • (0.024)

(+) ns

Autonomous

team

  • -0.041**
  • (0.020)
  • -0.077***
  • (0.028)
  • 0.000
  • (0.000)

(+) ns

Job rotation

  • -0.010
  • (0.019)
  • -0.043
  • (0.041)
  • 0.008
  • (0.017)

(-) ns

Organized employee voice in 2005

  • 0.004
  • (0.029)

Employee shareholding in 2005

  • 0.080
  • (0.055)

Ln (Productivity per employee in 2005)

  • 0.550***
  • (0.059)

Ln (Assets per employee in 2005)

  • -0.264***
  • (0.047)

Two-digit

industry

Yes

Yes

Yes

Yes

Yes

Other controls

Yes

Yes

Yes

No

Yes

N

1,857

717

1,426

530

2,569

R2

0.63

0.61

0.71

0.60

0.03

Interpretation: Controlsare capital intensity, shareofemployees >fifty-five, establishmentage, percentage ofwomen, percentage of low-skilled workers, percentage of high-skilled workers, and firm size category (except column 5, establishment size). Establishments with twenty or more workers in the private nonagricultural sector. Column 5, ordered logit on relative productivity indicated by the manager, and pseudo-R2. Robust standard errors clustered by two-digit industry codes.

***significantatthe1 percentlevel; **significantatthe5 percent level;*significantatthe10 percentlevel.

characteristics that led to sustainable performance. In Table 4.7, column 4 presents the estimation on the large subsample of firms that are on average older and for which accounting data in 1999 are also available in our database; we control both by the labour productivity in 1999 and by the capital intensity in 1999, in order to capture a part of the heterogeneity in the information and also to reveal potential reverse causality in the implementation of work practices. Column 5 concerns the panel of firms for the period 1999-2005.

None of the models shows a significant positive correlation between productivity and high-performance practices. We do not report here the similar results of regressions run with a regressor that is an aggregate index of these practices, in application of the idea of bundling practices. Given the methodological limitations stressed above, we do not conclude that these practices are inefficient, but rather that our data and approach do not capture an effect of such practices on productivity.[3]

On the contrary, in estimations run on the three cross-sectional subsamples, an organized employee voice is associated with significantly higher labour productivity. The magnitude of the coefficient is large: one standard deviation implies about a 3 per cent gain in productivity. Results are less robust for employee shareholding, but again the magnitude of the estimated coefficient is significant. The statistical weakness in mono-establishment firms may be linked to the fact that only one- fifth of the managers interviewed reported employee shareholding, while one-third of the managers of multi-establishment firms did.

Including the productivity level and the capital intensity in 1999 among the regressors confirms the qualitative results; however, the estimated coefficients for both employee shareholding and employee voice management are reduced by about one-third. Note that the negative correlation between capital intensity in 1999 and productivity in 2005—with the knowledge of the capital intensity in 2005—is consistent with declining efficiency of ageing capital. Since we use the logarithm of productivity, a coefficient lower than 1 for past productivity is consistent with the beta-convergence of productivity (e.g. on French firms, see Chevalier, Lecat, and Oulton, 2012).

The panel of establishments surveyed in 1999 and 2005 enables us to go one step further by adding the presence in 1999 of employee voice management and employee shareholding as controls for unobserved heterogeneity and potential endogeneity between practices and better performance. Column (5) may be read as a first difference between 1999 and 2005 as well. On this smaller subsample, the potential impact of employee shareholding vanishes, but the impact estimates for employee voice management are even larger.

We also experimented with instrumental variables to correct for endogeneity. We instrumented a high-involvement practice by the weighted mean average of the practice in other establishments of the full REPONSE sample operating in the same two-digit industry;the weights are the same as the ones indicated by the DARES so as to make the survey representative of French establishments according to size and activity. Both instruments are highly correlated with the seminal variables. However, when controls are included in the estimation, standard tests do not reject,[4] and by far, the hypothesis that each of our two high-involvement practices is exogenous. Therefore, we retain OLS estimators, which should be more efficient.

Overall, our findings point to the positive impact of high- involvement practices on labour productivity in 2005. Note that if we follow the literature focusing on the intensity of the use of innovative practices, the aggregated index summing our five practices is strongly correlated with higher productivity in 2005.

Similar exercises are then run on the data from the 2011 REPONSE survey. Table 4.8 reports the results of the estimations of the production function in 2011. As in 2005, the job rotation and quality management variables are not significantly correlated with higher productivity; the autonomous work team variable is negatively correlated with productivity, but this relation vanishes when we control for the past productivity.

Unlike 2005, in 2011, regardless of the specification, high-involvement practices—employee shareholding, organized employee voice—are no longer associated with enhanced productivity. The estimated coefficients are close to zero, and even negative on some samples. On the largest samples (column 1, Tables 4.7 and 4.8), coefficients associated respectively with organized employee voice and employee shareholding are statistically different between 2005 and 2011 at the 5 per cent and 10 per cent levels. They are still different just above the 10 per cent threshold when past productivity is included (column 3).

In addition to accounting data provided by Risk, in the REPONSE 2011 survey, managers were questioned about the relative productivity performance of their establishments. They had to scale their response from much lower than their competitors to much higher (i.e. according to five levels). This qualitative variable is strongly correlated with the productivity measure derived from accounting information, even within two-digit industry categories. This variable is available for most of the establishments surveyed, and thus for a larger and rather representative sample of the French establishments. Estimations using this relative productivity measure as a dependent variable are presented in column 5. An ordered logit confirms no significant correlation between high-performance or high-involvement practices and this productivity scale.

These contrasting results suggest a break in the relationship between high-involvement practices after the 2008 shock. If we consider employee voice management alone, the potential loss of productivity can be up to 10 per cent for establishments implementing both employee-voice groups and regular workplace meetings, and 5 per cent for an average establishment in our sample. These micro-estimates should be translated into macro-figures with caution: the dispersion of estimated coefficients is large; our non-representative sample includes only establishments with twenty or more workers, for which the high- involvement practices may be more volatile;about 40 per cent of the French private workforce belongs to smaller establishments.

Our findings are consistent with a reduction in the engagement of workers. However, the findings may also be interpreted as the result of a labour-hoarding process: firms may be reluctant to fire their own share- holders;they may retain their workforce—especially with specific human capital—and try to preserve the workers' long-term commitment. The study of relationships between work practices and skilled labour hoarding, however, does not support this last interpretation. When high-involvement practices are included in our nested logit (see Appendix 4.B and Subsection 4.5.2 above), they do not seem to boost the hoarding of managers and professionals.[5]

Whatever the interpretation, the loss of associated productivity associated with lower efficiency of some of the work practices including employee shareholding is probably a reversible consequence of the economic downturn and uncertainty about potential recovery.

  • [1] See Posthuma etal. (2013) for a comprehensive taxonomy of high-performance workpractices.
  • [2] Since profit-sharing schemes (participation) are mandatory in firms with over fiftyemployees, we do not consider this practice here.
  • [3] For example, assuming that the spread of innovative practices is mature; then, thechoice to implement a practice or not is optimal, and the econometric model cannot catchan 'effect' of the practices.
  • [4] p > 25 per cent for organized employee voice and p > 50 per cent for employee shareholding, according to Durbin, Wu-Hausman tests under the assumption of independent andidentically distributed errors, Woolbridge's robust score test, and the regression-based testwhen clustering.
  • [5] Employee shareholding is even negatively related with skilled labour hoarding, suggesting that a higher participation of workers would more likely favour a more homogeneousadjustment of the workforce when employment cutbacks are implemented.
 
Source
< Prev   CONTENTS   Source   Next >