Should We Clash or Should I Go? The Impact of Low Wage and Bad Working

Conditions on the Exit-Voice Trade-off  

4th Conference on Regulating for Decent Work: Developing and Implementing Policies for a Better Future at Work

July 10, 2015ILO Geneva

Olivier Godechot (Sciences Po, MaxPo and OSC-CNRS)Zinaida Salibekyan (CEE, Aix-Marseille University, LEST)

MotivationMotivation• Examine exit and voice strategies together in relation to working 

conditions and pay.

• Hirschman (1970). Dissatisfaction with price and quality.

• Two options: exit or voice 

• The central mechanism relies on the information exchange

• Pay is the first informational element that will be disclosed during the recruitment. 

• Uncertainty with regard to working conditions as they are not contractible.

Review of the literature Review of the literature

• The voice strategy  - has a positive impact on wage increase (Freeman and Medoff, 1984).

 - decreases job dissatisfaction of employees (Wood, 2008).

 - is positively correlated with the desirable workplace outcomes (Willman et al., 2009). 

• The exit strategy  - is negatively correlated with quitting behaviour (Clark, 1996; Ophem, 1991; Cottini et al, 2011; Garcia-Serrano, 2004).

 - Adverse working conditions have a negative impact on job satisfaction which leads to quitting behaviour (Böckerman and Ilmakunnas, 2009).

Workers coping with poor working conditions (1)Workers coping with poor working conditions (1)

• Voice is more costly than exit.- Information on the quality of alternative options is poor 

and change is risky. - Voice is chosen when exit is difficult, costly and 

unavailable (Hirschman, 1970).- Voice provides rich and detailed information in 

comparison to exit, transmits direct feedback about overlooked poor working conditions.

- H1: Poor working conditions favour collective voice.

Workers coping with poor working conditions (2)Workers coping with poor working conditions (2)

• Employer confronted with collective complaint can  - improve working conditions, - compensate for the poor working conditions with a pay 

increase (Smith, 1776; Rosen, 1986),

 - offer a combination of these two improvements.

• When working conditions are very bad the cost of improving them is lower than the cost of compensating for them. 

• H2: Collective voice improves working conditions.

Workers coping with low pay Workers coping with low pay

• Workers who ceteris paribus earn lower wages are more likely to find rival offers that will improve their wage among other firms.

• Higher wage workers are less likely to find better offers. 

• Workers benefit from information about which industries pay higher wages (Acemoglu, 2001). 

• H3: Low pay favours exit.

• H4: Exit improves salary.

• Collective voice as a traditional working class strategy to increase wages. 

• Collective voice is costly and depends on collective coordination. - The success rate is low 

• Quitting does not require collective coordination and there is an accurate information on pay increase.  - The success rate is higher

Exit and Voice Exit and Voice

Data (1)Data (1)

• British linked employee-employer data: British Workplace Employment Relations Survey (WERS 2005 & 2011). The WERS panel contains  989 workplaces 

• French linked employee-employer data: Enquête Relations Professionnelles et Négociations d’Entreprise (REPONSE 2005 & 2011). The REPONSE panel contains 872 workplaces.

• Aggregated weighted means of demographic and job characteristics have been calculated for each establishment in the panel. The sample size is 770 in the WERS survey and 805 in the REPONSE survey.

Data (2)Data (2)

• Exit rate has been calculated in the WERS. The average exit rate (2011)=8%.• •To obtain the exit rate in the REPONSE survey matching has been performed with the DMMO survey (2006) and the EMMO survey (2006). The sample size reduced to 597 establishments. •The average exit rate (2006) = 4,3%. 

•Les Salaires vus par les Salariés (SalSa 2009):  a random sample of employees from the 2006 Panel DADS. Matching has been performed with the Panel DADS to obtain the employees’ work career since 1976 .

Exit and Voice in the WERS survey Exit and Voice in the WERS survey 

Which, if any, of the forms of industrial action on this card have taken place at this workplace during the last 12 months?

Yes On how many occasions (if Yes)

1. Strikes of less than a day 3 % 1.352. Strikes of a day and more 15 % 1.373. Strikes 18% 1.404. Overtime ban or restriction by employees 2 % 4.535. Work to rule 3 % 3.006. Other industrial action (for example, go slow, sit in) 1 % 1.757. None of these 82 % /

Exit and Voice in the REPONSE survey Exit and Voice in the REPONSE survey Which types of conflicts did it happen in the establishment between 2008 and 2010?

More than 5 times

From 3 to 5 times

From once to twice

Never

1. Stoppage 11% 6% 14% 68%2. Strikes of less than 2 days 9% 7% 13% 71%3. Strikes of more than 2 days 1% 1% 7% 90%4. Go-slow strike 1% 1% 2% 96%5. Work-to-rule, slowdown 0% 0% 2% 98%6. Overtime ban 3% 2% 4% 91%7. Assembly, demonstration 6% 4% 12% 77%8. Petition 2% 3% 20% 75%

Other voice variables Yes9. At least one strike in the firm (constructed out of 2 and 3) 33%10. Employee making an application to an Employment Tribunal 51%11. Absenteeism is a problem 43%12. Strong tensions between subordinates and supervisors 36%13. Individual participation between 2008 and 2010 in a work stoppage (stoppage or strike) (n=2579)

22%

Voice and Exit strategies in SalSaVoice and Exit strategies in SalSa

Yes No Doesn’t know

1. During the last five years, did you participate in a collective action (strike, demonstration, petition) linked to your work? (n=3117)

23% 76% 1.0%

2. In the last five years, have you ever voluntarily left your job? (n=3117)

18% 81% 0.5%

Exit voice trade-off. Exit voice trade-off. Variation of exit rate depending on the use of voice strategyVariation of exit rate depending on the use of voice strategy

 Workplace (or individual) recently voicing: 

SurveyYes No

ΔT. 

Test

Correlation coefficient

WERS : 2011 workplace exit rate depending on strikes declared in 2011

4.5% 8.7% *** -0.14***

REPONSE : 2006 workplace exit rate depending on strikes declared in 2011 at the establishment level

2.6% 5.4% *** -0.23***

REPONSE : 2006 workplace exit rate depending on voice declared at the individual level

2.1% 4.4% *** -0.19***

SalSa :  Exit at the individual level depending on voice declared at the individual level

11% 19% *** -0.09***

Poor working conditionsPoor working conditions

• Objective and subjective approaches of job quality.

• Subjective approach: the best way to know about employees’ working conditions is to ask them directly how satisfied they are with it (Stride et al., 2007; Clark, 2011).

• The objective approach: to define well-being of individuals in relation to ‘capabilities’ of achieving certain things (Sen, 1999).

• Comprehensive approach (Spencer and Budd, 2014).

• A composite variable offers the opportunity of comparing responses across surveys in a relatively simple way.

• “My job requires that I work very hard” (Strongly agrees or agrees: 75%), 

• “I never seem to have enough time to get my work done” (Strongly agrees or agrees: 41%),

•  “Influence what tasks you do in your job” (Little or none : 26%), 

• “Influence on how you do your job” (Little or none : 15%),

•  “Satisfaction with the work itself (Very dissatisfied or dissatisfied : 9%), 

• “How much of the time the job made you feel tense over the last few weeks” (All or most of the time : 19%), 

• “Relations between managers and employees here” (Very poor and poor: 15%). 

• Poor working conditions index at the establishment level is the firm weighted mean of the index at the individual level. Similar in 2011.

Poor working conditions in the WERS surveyPoor working conditions in the WERS survey

Poor working conditions in the REPONSE

• The REPONSE (2004) provides a global question on working conditions. 

• “Do working conditions limit your involvement at work? : - Yes absolutely (23%), - yes somehow (33%), - not really (26%),-  not at all (17%). 

• We use this four scale ordinal variable as a continuous one (ordered from good to poor conditions).

• In order to measure the impact of voice on quality change, we use in REPONSE (2011) a four scale that measures the degree of satisfaction with working conditions. 

-very satisfied: 14%- rather satisfied: 55%,-rather unsatisfied: 25%,- very unsatisfied: 6%).

Poor working conditions in the SalSa surveyPoor working conditions in the SalSa survey

• In the SalSa survey we built the index BQ2008. – These variables are 

- high-speed work (all the time: 43%), - physically hard work (yes: 37%), - mentally hard work (yes: 64%), - dangerous work (yes: 27%), - convenient working schedule (no: 17%), - and the fact of liking the work (Sometimes and never: 10%). 

Pay

• In the WERS (2004): aggregated weighted mean of the log hourly wage (2004) at the establishment level. 

• In the REPONSE (2004): aggregated weighted mean of the log hourly wage (2004) at the establishment level from DADS.

• SalSa : detailed information on full career of employees. The log hourly wage is calculated both for 2003 and 2008.

Empirical strategy Empirical strategy • We explore the links between poor working conditions 

and pay in t-1 and the strategy declared in t and that occurred sometime between t-1 and t.  - Pay and poor working conditions not to be estimated after the strategy occurred.-Establishment level regressions might be biased by ecological fallacy. A confirmation of establishment results at the individual level is important.

• The impact of strategy Yt either on the improvement of working conditions or on pay increase. 

The impact of poor working conditions(t-1) and pay (t-1) on strategy (t)The impact of poor working conditions(t-1) and pay (t-1) on strategy (t)

Collective action (2011) Exit (2011)WERS - Establishment level Model 1 Model 2  2004 Poor working conditions 0.40*** [1.45] 0.02 [1.03] (establishment weighted mean) (0.15)   (0.04)  2004 Log hourly wage 0.11 [1.12] -0.11** [1/1.17] (establishment weighted mean of log pay) (0.23)   (0.05)  Pseudo R2 (logistic) or R2 (OLS) 0.40   0.25  Number of observations 764   706  Model Logit   OLS   Collective action (2011) Exit (2006)REPONSE - Establishment level Model 3 Model 4 2004 Poor working conditions 0.20** [1.22] 0.06 [1.08](establishment weighted mean) (0.10)   (0.04)  2004 Log hourly wage 0.52*** [1.68] -0.21*** [1/1.3] (establishment weighted mean) (0.15)   (0.06)  Pseudo R2 (logistic) or R2 (OLS) 0.23   0.33  Number of observations 804   596  Model Logit   OLS  REPONSE - Individual level Model 5      2004 Poor working conditions 0.17** [1.19]    (establishment weighted mean) (0.06)      2004 Log hourly wage 0.40*** [1.49]     (establishment weighted mean) (0.06)      Pseudo R2 0.13      Number of observations 2415      Model Logit      REPONSE - Individual level IV regression Model 6      2011 Poor working conditions (instrumented with 0.86* [4.9]    2004 firm poor working conditions and 2004 firm wage) (0.34)      

2011 Log hourly wage (instrumented with 2004 firm 0.88*** [4.96]    

poor working conditions and 2004 firm wage) (0.18)      R2 0.14      Number of observations 2415      Model 2SLS      

The impact of poor working conditions(t-1) and pay (t-1) on strategy (t)The impact of poor working conditions(t-1) and pay (t-1) on strategy (t)

Collective action (2008)

Exit (2008)

SalSa BQ2008 - Individual level Model 7 Model 8 2008 Poor working conditions 0.37*** [1.45] 0.04 [1.04] (0.05)   (0.06)  2003 Log hourly wage 0.20* [1.22] -0.18** [1/1.2] (0.08)   (0.08)  Pseudo R2 0.11   0.14  Number of observations 2466   2476  Model Logit   Logit  SalSa BQ2003 - Individual level Model 9 Model 10 2003 Imputed poor working conditions 0.58** [1.8] -0.25 [1/1.28] (0.26)   (0.29)  2003 Log hourly wage 0.15* [1.16] -0.16* [1/1.17] (0.08)   (0.09)  Pseudo R2 0.09   0.14  Number of observations 2440   2451  Model Logit   Logit  

Results (1)Results (1)

• Poor working conditions and strategies (H1).- 1 standard deviation decrease in the quality of working 

conditions multiplies the probability of participating in collective action by a factor of 1.2 (SalSa, REPONSE) and to 1.5 (WERS; SalSa) and 1.8 (SalSa with imputed variable).

- Poor working conditions do not appear to be significant in the exit  models.

• Low pay and strategies (H3).-In  all  the  exit  models  1  standard  deviation  decrease  in  log hourly pay increases the probability of quitting by a factor of 1.2 to 1.3. 

The impact of strategy (t) on working conditions’ improvement or pay increaseThe impact of strategy (t) on working conditions’ improvement or pay increase

Working conditions’ improvement

Pay increase

WERS - Establishment level Model 11 Model 12Collective voice: at least one strike reported in 2011 0.022 0.073 (0.115) (0.114)Exit: voluntary quit rate in 2004 0.006 -0.095** (0.043) (0.043)R2 0.004 0.01Number of observations 560 560Model OLS OLSREPONSE - Establishment level Model 13 Model 14Collective voice: at least one strike reported in 2011 0.017 0.012 (0.043) (0.043)Exit: 2006 voluntary quit rate 0.046 -0.125 (0.088) (0.089)R2 0.02 0.01Number of observations 578 578Model OLS OLSSalSa - Individual level   Model 15Collective voice   -0.002   (0.015)Voluntary exit   0.069***   (0.018)R2   0.01Number of observations   1719Model   OLS

Results (2)Results (2)

• Voice improves working conditions (H2).- Voice has a small positive but non significant impact on working 

conditions.- Employees, who suffer the worst working conditions, are the most 

likely to report non-monetary improvements. - Difficult to confirm the H2.

• Exit improves pay (H4)- Voluntary exit significantly increases wages by 7%.- REPONSE and WERS models do not provide a proper test for H4.- Thanks to SalSa we can thereby partially confirm H4.

ConclusionConclusion• The first detailed application of Hirschman’s exit-voice 

framework to the labour market stressing the importance both on working conditions and pay. 

• Poor working conditions have a positive and significant impact on the probability of participating in collective action.

• Low pay tends to favour exit strategy. 

• An increase in log hourly wage decreases the probability of quitting. 

LimitationLimitation  • The lack of information of working conditions prior 

to the measured strategies. Individual panel surveys could provide a way of confirming our results. 

• Different theoretical framework can be used. Collective voice is based on bargaining power and on the degree of competition (Budd, 2005).

DiscussionDiscussion• Future studies can compare different periods and labour 

market regimes. 

• To understand behavioural patterns of individuals in different unionized seetings. 

• To revise our views on collective action. 

• Market view considers voice inefficient and exit as a pure market strategy. 

• Marxist view exit is viewed as an individualistic petit-bourgeois strategy. 

• Our Hirschmanian approach is situated between the two.