Small Business Economics

, Volume 41, Issue 2, pp 379–399

Productivity, wages and intrinsic motivations

Authors

    • Department of Economics Law and InstitutionsUniversity of Rome II, Tor Vergata
  • Stefano Castriota
    • Department of EconomicsUniversity of Perugia
  • Ermanno C. Tortia
    • Department of EconomicsUniversity of Trento
Article

DOI: 10.1007/s11187-012-9431-2

Cite this article as:
Becchetti, L., Castriota, S. & Tortia, E.C. Small Bus Econ (2013) 41: 379. doi:10.1007/s11187-012-9431-2

Abstract

There is a long-standing debate in labour economics on the impact of workers’ intrinsic motivations on equilibrium wages. One direction in economic theory suggests that intrinsically motivated workers are willing to accept lower wages and “donate” work, for example, in terms of unpaid overtime (the donative-labour hypothesis). In the other direction, intrinsic motivations are expected to increase worker productivity and, in turn, wages (the intrinsic motivation-productivity hypothesis). Using a new database of a sample of workers in the cooperative non-profit sector, we find that, consistently with the motivation-productivity hypothesis, more motivated workers earn significantly higher wages, which signals higher productivity. Evidence supporting the donative-labour hypothesis is weaker, even though a generally positive connection between motivations and work-donation is confirmed. We interpret these findings by arguing that the impact of the donative-labour effect is dominated by the intrinsic motivation-productivity effect.

Keywords

Labor-donationIntrinsic motivationsProductivityWagesWage differentialsNon-profit

JEL Classifications

J30J31I30L26

1 Introduction

The influential theory of the donative-labour hypothesis (Hansmann 1980; Preston 1989; Rose-Ackerman 1996; Frank 1996)1 predicts a negative relationship between intrinsic motivations and workers’ pay. The common rationale, consistent with the principle of compensating wage differentials, is that wage-earners will accept lower pay if they find intrinsic (non-monetary) value in their jobs. This implies that intrinsically motivated workers who find that their motivations are satisfied in their occupations and in the missions of their productive organisations, are willing to donate labour to them.

Put differently, these workers exchange monetary incentives for intrinsic, relational, and social incentives which have a positive impact on their well-being, even where they do not affect their income directly (Borzaga and Tortia 2006). The extreme bound of the donative-labour hypothesis is voluntary work, where individuals enjoy non-monetary compensation for their activity to such an extent that they are willing to “work for nothing” (Freeman 1997). Volunteering can also be partial, and in certain organisational environments workers may be ready to donate at least a part of their work effort without monetary compensation. It is important to provide convincing theoretical explanations here, and to assess the modalities and the extent of partial work donations empirically, for example in terms of unpaid overtime and taking holidays in arrears, an issue that has been underestimated by the discipline to date.

Several empirical papers on the profit/nonprofit wage differentials find evidence consistent with the donative-labour hypothesis. Weisbrod (1983) studies the difference between lawyers working in the profit and non-profit sectors, and documents a significant non-profit wage gap. Preston (1989) finds a similar result for different types of white-collar pay (managers, professionals, and clerical and sales workers) compared across the two sectors. Evidence from Europe seems to go in the same direction (see Mosca et al. 2007, for Italy; Narcy 2009, for France).

However, in a thorough empirical analysis on US Census data, Leete (2001) demonstrates that, when finer controls at an industry and occupation level are introduced into the analysis, the non-profit negative wage gap persists only in a few cases. To account for the puzzle, Leete (2001) comments that “the pattern of nonprofit wage differentials across disaggregated occupations and industries is suggestive of a number of forces affecting nonprofit wages simultaneously” (p. 138).

The main point which is implicitly raised here is that, in the absence of explicit data on workers’ intrinsic motivations, the effect of these motivations on wages has been tested by looking at the profit/non-profit differential, the hypothesis being that the factors driving this differential are: (i) higher levels of satisfaction of intrinsically motivated workers in the non-profit industry, assuming homogeneity in intrinsic motivations; and (ii) the difference in workers’ intrinsic motivations in the two sectors, assuming worker heterogeneity in intrinsic motivations (for example, Benz 2005).2 In essence, given the impossibility of testing the intrinsic motivation/job donation/wage nexus directly, affiliation with the non-profit industry has been considered to be a good proxy for higher satisfaction (case (i)), or relatively higher intrinsic motivations (case (ii)).3

This approach is not without its problems. First, intrinsic motivations may vary significantly within industries, and between profit and non-profit sectors. Second, following Leete (2001), the profit/non-profit wage differential may depend on a series of other factors. For instance, in an era of increasing government debt, the differential may be caused, inter alia, by the higher dependence of non-profit organisation revenues on government contracts won in public procurement auctions with significant rebates. On the other hand, Preston (1988) finds that, in the past, this same dependence on the public sector determined a differential in the opposite direction. Her result—of a positive non-profit/profit wage differential where the non-profit industry is highly subsidized—is explained by the argument that non-profit managers have lower cost constraints. Third, the negative wage gap may be determined by quality differences between workers in the two sectors (Preston 1989).

The central argument of our paper is that this area of research may be enriched if we consider additional aspects of the relationship between intrinsic motivations and wages. Our point is that intrinsically motivated individuals may be more productive and, for this reason, if a part of this greater productivity is remunerated by their employers, they may end up having higher, and not lower, wages than their non-intrinsically motivated colleagues. What must be tested empirically, therefore, is whether the effect of the donative-labour hypothesis on wage differentials (the labour supply schedule of intrinsically motivated workers is shifted to the right with respect to the standard labour supply curve, and therefore their equilibrium wage is lower) is stronger or weaker than the effect of the intrinsic motivation-productivity hypothesis (intrinsically motivated workers are more productive, and therefore end up earning more).

Our hypothesis can also be tied in nicely with other strands of inquiry in the literature. In some works on the economics of identity (Akerlof and Kranton 2005; Goette et al. 2006), it has been shown that identity can substitute economic incentives in motivating individuals to pursue organizational goals. This argument seems to suggest a hypothesis similar to the labour donative hypothesis in non-profit organizations, since it would imply that insiders whose identity has been moulded by organizational culture end up earning lower incomes than outsiders and new entrants. This conclusion is very much in contrast with the empirical evidence we presented in the sections that follow. Sorting out a donative-labour hypothesis vis à vis the positive impact of intrinsic motivations on insiders’ income can help solve the puzzle where an organizational culture as a proxy for individual identity is understood to be the relevant motivator. Unlike in our work, however, in Akerlof and Kranton (2005), identity as motivational capital is moulded within the boundaries of the organization, and intrinsic motivations prior to enrolment are not taken into consideration. We, on the other hand, understand intrinsic motivations to have a dynamic impact on productivity, since they relate to a person’s own ability to develop new skills and competencies. From this perspective, the donative-labour hypothesis outlined in the literature would represent a static effect, since it pertains to a one-shot trade-off between monetary and non-monetary remuneration. Our perspective, instead, is more similar to the arguments put forward by Rob and Zemsky (2002), who show that a corporate culture which places value on intrinsic preferences for cooperative effort can increase future economic returns.

We test the effect of intrinsic motivations on wage differentials in the cooperative sector in Italy. Since we have different proxies for intrinsic motivations and are working within a single sector (and not on the comparison between profit and non-profit), we can test the intrinsic motivation-wage nexus directly. The clear advantage of our study is that, unlike in previous works, the risk of the omitted variable bias raised by Leete (2001) is minimized, since (i) we consider only one sector, and (ii) our survey provides an impressive number of control variables. Contrary to common belief, there are also significant differences among workers in both wages and intrinsic motivations in the non-profit sector. The reason for this variability in intrinsic motivations lies in the low level of salaries, which means that applicants range from people with high intrinsic motivations (ready to accept a lower salary in order to achieve a social goal) to those with high extrinsic motivations (because they do not have better opportunities in the profit sector, they end up in the non-profit sector as a second-best alternative).

The obvious limitation of our analysis is that what we find works in the specific case of the cooperative non-profit industry. However, even though the average level of intrinsic motivations in this industry may be higher than it is in the profit industry, we believe that our results are also likely to work there. Therefore, under the highly plausible assumption of a heterogeneous distribution of intrinsic motivations among workers in the profit industry, we expect to find the same static and dynamic wage setting mechanisms.4 We shall leave this last point to future research, however.5

2 Data

The data we use (ICSI 2007) were collected in 2006 by a pool of six universities6 by means of questionnaires submitted to a representative sample of 4,134 employees and 338 managers of 320 Italian cooperatives. The survey instrument comprises a large set of questions, ranging from socio-demographic controls (age, gender, education, etc.) to economic variables (e.g. wage), job characteristics (tasks, working hours, overtime) and job satisfaction with regard to a number of possible domains (relationship with colleagues, wage, type of job, etc.). The result is an extremely rich database, which allows us to study the conditions and motivations of individuals employed in Italian non-profit enterprises.

In order to proceed with the survey, an initial sample of 411 social enterprises was extracted from the 2003 ISTAT7 census on social cooperatives (ISTAT 2003, 2007), which recorded 6,168 active cooperatives (with at least one employee) in the country (Carpita 2009, pp. 1–32). Representativeness was guaranteed by stratification on the basis of three parameters: type of cooperative (Type A and Type B),8 geographic representativeness by province (the Italian State is made up of 20 Regions and 107 Provinces), and size by number of employees. After accounting for organisations which were not willing to participate, the final sample consisted of 320 organisations.

Four different questionnaires were distributed to the selected sample. They concerned, respectively, paid workers, volunteer workers, cooperatives, and managers. Only data from the questionnaire distributed to paid workers and the cooperatives will be used in this article. The former represents the main source of data, while the typology of services provided has been compiled from the questionnaire administered to organizations. The rate of individual non-responses for paid workers was extremely low, since 85 % of the workers involved answered an average 90 % of the 87 questions (56 single choice and 31 multiple choice questions).9

As for the ICSI 2007 survey, which reports 2006 data, in Table 1 we describe the variables used, whilst in Table 2, we provide the corresponding descriptive evidence. The average after-tax monthly wage (net of bonuses and premiums) is 867 euros. A large share of the workers are women (74 %) and young (the average age is 37), thus reproducing the salient features of the universe of cooperative workers. On average, they have been working in the cooperatives investigated for 6.2 years (median value 5 years). Although 29 % of the sample has a university degree, the average length of education is close to 13 years, and therefore coincides with a high-school diploma. Seventy-five percent of the respondents are also members of the cooperative for which they work.
Table 1

Description of the socio-economic variables

Variable

Meaning

Wage

Average monthly wage net of taxes and bonuses

Ln of wage

Natural logarithm of average monthly wage net of taxes and bonuses

Hourly wage

Average hourly wage net of taxes and bonuses

Ln of hourly wage

Natural logarithm of average hourly wage net of taxes and bonuses

Bonus

Amount of average yearly monetary premiums

Hours

Average number of weekly working hours

Extra hours

Average number of weekly overtime working hours

Holidays in arrears

Number of days of holidays in arrears

Male

Dummy: gender

Age

Respondent’s age

Education

Years of education

Italian

Dummy: Italian nationality

Member

Dummy: member of the cooperative

Years in coop

Number of years of work in the cooperative

Permanent

Dummy: permanent position

Full-time

Dummy: full-time position

Internship

Dummy: internship position

Medium coop

Dummy: medium size (15–49 paid workers)

Large coop

Dummy: large size (50 and more paid workers)

Type A

Dummy: type a social cooperative (delivers social services)

North-west

Dummy: north-west of the Italian territory

North-east

Dummy: north-east of the Italian territory

Centre

Dummy: central Italy

Table 2

Summary statistics of the socio-economic variables

Variable

Obs.

Mean

SD

Min

Max

Wage

3,744

867.45

298.85

100

6,453

Ln of wage

3,744

6.70

0.37

4.61

8.77

Hourly wage

2,698

6.70

2.29

1.46

50

Ln of hourly wage

2,698

3.25

0.25

1.76

5.30

Bonus

4,134

77.20

285.34

0

6,000

Hours

3,740

31.31

8.66

2

50

Hours extra

3,092

1.79

3.30

0

30

Holidays in arrears

4,073

6.80

12.42

0

189

Male

4,082

0.26

0.44

0

1

Age

3,986

37.38

9.02

17

73

Education

3,759

12.93

3.35

0

21

Italian

4,134

0.95

0.22

0

1

Member

4,134

0.76

0.43

0

1

Years in coop

3,905

6.21

4.89

0

36

Permanent

4,134

0.80

0.40

0

1

Full-time

4,063

0.56

0.50

0

1

Internship

4,157

0.02

0.15

0

1

Medium coop

4,134

0.32

0.46

0

1

Large coop

4,134

0.43

0.50

0

1

Type A

4,134

0.78

0.41

0

1

North-west

4,134

0.40

0.49

0

1

North-east

4,134

0.22

0.41

0

1

Centre

4,134

0.22

0.41

0

1

3 Intrinsic motivations: descriptive findings

Intrinsic motivations are described in the specialised literature as attitudes or drivers of human behaviour extending beyond satisfaction of mere physiological and security needs which require monetary, or in any event material, remuneration. In the Maslow scale (1943, 1954), higher-level needs are linked to non-material, or psychological, aspects of human activity. Even though their satisfaction is not necessary for physiological survival, they have a central role in defining psychological health.10 The well-known original definition of intrinsic motivation (Deci 1975) states that:

One is said to be intrinsically motivated to perform an activity when he receives no apparent reward except the activity itself.

More recently, Deci and Ryan (2000) state that:

Perhaps no single phenomenon reflects the positive potential of human nature as much as intrinsic motivation, the inherent tendency to seek out novelty and challenges, to extend and exercise one’s capacities, to explore, and to learn. … The construct of intrinsic motivation describes this natural inclination toward assimilation, mastery, spontaneous interest, and exploration that is so essential to cognitive and social development and that represents a principal source of enjoyment and vitality throughout life.

What is relevant to our analysis here is that these definitions are linked to the non-material aspects of the job and psychological or intellectual satisfaction. Since higher level needs do not require (though they do not exclude either) monetary remuneration, workers may seek satisfaction irrespectively of the level of pay. Hence, even if the monetary remuneration is kept to a minimum, workers may be satisfied with their jobs where immaterial needs are satisfied. From another perspective, some workers may decide to refuse jobs that satisfy intrinsic motivations poorly, even where the wages are higher. Intrinsic motivations relate to interest in the activity performed, good relationships with other colleagues and superiors, and involvement and autonomy in decision-making at the operational and strategic level. The intrinsic nature is specific to the task and directed to the flow of activity, to goals that are self-defined, and to the obligations of personal and social norms—benevolence, identity, and fairness (Frey 1997; Depedri et al. 2010).

Intrinsic motivations can imply both self-regarding preferences, as in the case of professional growth, and other-regarding preferences, as in the case of attention paid to the needs of the community (Ben-Ner and Putterman 1998). Both of these aspects are closely linked to the perception of the organizational environment, for example, to procedural and distributive fairness, and to involvement in decision-making (Tyler and Blader 2000). Hence, complementarity more than substitution can be expected among the different kinds of preferences linked to intrinsic motivations.

In order to measure intrinsic motivations, we consider one question in the workers’ survey, which asked respondents to report their degree of agreement (on a 1–7 Likert scale) with various definitions of their job in the cooperative (see Table 3).11 The question was:

“Your relationship with your Cooperative is:
  1. (i)

    a mere contractual relationship where labour is exchanged for pay (RSE1: pure contract);

     
  2. (ii)

    a contribution which helps the cooperative to reach its goals (RSE2: contribution to social enterprise);

     
  3. (iii)

    a mix between professional growth and personal development (RSE3: mix-job and personal growth);

     
  4. (iv)

    a set of relationships which goes beyond a mere contractual relation (RSE4: mix-beyond mere job);

     
  5. (v)

    a social engagement common to the respondent and the cooperative (RSE5: social commitment);”

     
where RSE is the acronym for Relationship with the Social Enterprise. We believe that a high percentage of agreement is negatively correlated in the case of the first definition of the current job (RSE1), while it is positively correlated with the other definitions (RSE2 to RSE5), with intrinsic motivations. As is clear from the chosen definitions, our selection includes both self-regarding (RSE3) and other-regarding intrinsic motivations (RSE2, but especially RSE5). More generally, we argue that agreement with RSE2 to RSE5 indicates that the workers have an interest (training, identification with the cooperative’s goals, community and relational elements of the job, and social engagement) in the job, and gain utility from it, which go beyond mere remuneration and other contractual entitlements.
Table 3

Description of the questions concerning intrinsic motivations

Variable

Description

Relationship with the social enterprisea

 RSE1: pure contract

You consider your relation with the social enterprise to be a mere labour contract

 RSE2: contribution to the social enterprise

You consider your relation with the social enterprise to be a contribution towards achieving the social enterprise’s targets

 RSE3: mix (job and personal growth)

You consider your relation with the social enterprise to be a mixture of job and personal growth

 RSE4: mix (beyond mere job)

You consider your relation with the social enterprise to be a set of relationships which go beyond the mere work relationship

 RSE5: social commitment

You consider your relation with the social enterprise to be a joint social commitment from you and your institution

 RSE6: useful job for people

You were looking for a job which could be useful to other people

 RSE7: personal accomplishment in job

You were looking for personal accomplishment in your job as well

Effort at workb

 EFF1: social enterprise needs

What is your effort with respect to the SE’s needs

 EFF2: clients’ needs

What is your effort with respect to the clients’ needs

 EFF3: colleagues’ effort

What is your effort with respect to your colleagues’ effort

 EFF4: wage

What is your effort with respect to your wage

aValues range from 1 to 7, where 1 stands for “I completely disagree” and 7 for “I completely agree”

bValues range from 1 to 7, where 1 stands for “Much less than required” and 7 “Much more than required”

We also include two other dummy variables which measure whether, before finding the current job in the cooperative, the respondent was looking for a job12:
  1. (i)

    which enabled him/her to be helpful to other people (RSE6)

     
  2. (ii)

    which promoted his/her self-fulfilment (RSE7)

     

These items focus more on the other-regarding than the self-regarding component of intrinsic motivations. Here again, we consider that high points given to these two statements are a signal that the individual does not look for monetary compensation alone, but also pursues self-fulfilment and the satisfaction of intrinsic motivations in his/her professional activity.

From a descriptive point of view, the level of intrinsic motivation seems to be generally high (Table 4). This is shown by the fact that the average degree of agreement with statement RSE1 (which is inversely correlated with intrinsic motivations) is 2.55; that is, below the central value of 4, and far lower than that measured for statements which we consider to be positively correlated with intrinsic motivations (RSE2 to RSE7) (which was around and over 5). Together with this average score, we observe that the shares of respondents giving the two highest scores (6 and 7) to the items which measure the presence of intrinsic motivations are always above 50 % (with the exception of the RSE2 and RSE4 questions) (Table 4).
Table 4

Summary statistics of the questions concerning the intrinsic motivations

Variable

Obs.

Mean

SD

Min

Max

Percentage of respondents for each declared value

1

2

3

4

5

6

7

RSE1: pure contract

3,457

2.55

1.91

1

7

47.53

14.03

9.55

11.48

6.25

4.8

6.36

RSE2: contribution to the social enterprise

3,564

5.20

1.57

1

7

2.95

3.28

7.01

18.52

20.17

21.91

26.15

RSE3: mix (job and personal growth)

3,554

5.45

1.48

1

7

2.08

2.95

4.25

15.17

19.3

26.17

30.08

RSE4: mix (beyond mere job)

3,494

4.91

1.75

1

7

5.35

6.44

8.44

17.43

18.83

20.78

22.72

RSE5: social commitment

3,556

5.31

1.61

1

7

3.49

3.68

5.65

15.94

16.96

24.97

29.3

RSE6: useful job for other people

3,865

5.31

1.68

1

7

3.36

4.81

6.39

15.83

15.01

22.04

32.55

RSE7: personal accomplishment in job

3,877

5.68

1.43

1

7

1.93

2.09

3.82

11.63

16.02

27.6

36.91

EFF1: social enterprise needs

4,020

5.39

1.13

1

7

0.22

0.22

0.92

26.09

23.03

30.22

19.28

EFF2: clients’ needs

4,031

5.32

1.36

1

7

1.74

2.94

3.41

26.45

21.50

28.50

15.46

EFF3: colleagues’ effort

3,858

4.77

1.07

1

7

0.54

0.52

1.97

49.07

21.49

18.61

7.80

EFF4: wage

3,907

5.45

1.36

1

7

1.13

1.33

2.46

24.37

17.71

23.57

29.43

4 Econometric findings

In this section, we perform our econometric analysis by testing for the existence of a significant nexus between wages and intrinsic motivations. In order to estimate the impact of intrinsic motivations on wages in cooperatives, we adopt the following standard reduced form specification, where the dependent variable (Y) is regressed on a set of independent regressors which include a gender dummy (Male), the respondent’s age (Age) and years of education (Education), a dummy for Italian nationality (Italian), another for being a member of the cooperative (Member), the number of years worked in the cooperative (Year in coop), two dummies for permanent and full-time job status (Permanent and Full-time respectively), two dummies for medium and large cooperative size (Medium and Large), a dummy which takes a value of one if the respondent works in a Type A cooperative (Type A), and three macroarea dummies (North-East, North-West and Centre).

Among our controls, we include not only industry classification for the two different (A and B) types of cooperative, but also additional dummies based on the types of clients served (for Type A) and integrated employee category (for Type B) which, in our opinion, when combined with the above-mentioned industry classification, provide a finer taxonomy of the cooperatives under scrutiny.13As dependent variables we consider alternatively: (i) the level of monthly after-tax wage; (ii) the same variable plus bonuses and premia; and (iii) the hourly wage (without bonuses and premia). The corresponding regressions are estimated with dependent variables in levels and logs. As well known, the log-linear specification with wage as dependent variable corresponds to the Mincerian (1974) equation. Since a number of the workers are employed in the same social enterprise, standard errors are clustered at the cooperative level to control for intra-firm interdependencies.

The econometric results yield clear-cut evidence of the role of intrinsic motivations. In the simpler specification, with the level of monthly wage (net of bonus and premia) as dependent variable, we find that agreement with statement RSE1 (work as a mere contractual relationship where labour is exchanged for pay) has negative and significant effects on wages (Table 5, column 2). From a quantitative point of view (on the restrictive assumption that the regressor is continuous), a unit point rise in agreement with this statement compared with the sample average level of agreement (2.59) reduces the wage by 8.6 euros. This implies that those declaring the highest degree of agreement with this statement on the Likert scale may earn up to 5 % less than the average. The other intrinsic motivation proxies are significant in the expected direction, with the exception of agreement with statements RSE2, RSE4 and RSE7.
Table 5

Wage regressions (net monthly wage, without bonuses)

Variable

(i)

(ii)

(iii)

(iv)

(v)

(vi)

(vii)

(viii)

Male

61.89

(6.17)

64.33

(6.28)

58.38

(5.68)

58.43

(5.53)

59.54

(5.64)

57.89

(5.64)

64.34

(6.40)

64.12

(6.38)

Age

−0.10

(−0.17)

0.09

(0.13)

0.27

(0.42)

0.26

(0.40)

0.24

(0.36)

0.09

(0.14)

0.27

(0.44)

0.06

(0.09)

Education

0.03

(0.02)

−0.14

(−0.11)

0.03

(0.03)

0.00

(0.00)

0.11

(0.09)

0.17

(0.13)

0.27

(0.22)

0.20

(0.16)

Italian

−59.40

(−1.36)

−85.88

(−1.80)

−77.85

(−1.57)

−74.67

(−1.55)

−78.66

(−1.60)

−71.30

(−1.50)

−66.05

(−1.42)

−68.29

(−1.46)

Member

1.92

(0.16)

−4.96

(−0.41)

−0.68

(−0.06)

3.71

(0.31)

−0.66

(−0.05)

0.07

(0.01)

6.56

(0.55)

3.71

(0.32)

Years in coop

9.93

(6.02)

10.86

(7.33)

9.87

(6.55)

10.57

(7.08)

10.44

(6.99)

10.56

(7.32)

9.84

(5.86)

10.28

(6.86)

Permanent

52.39

(3.35)

44.18

(2.67)

42.70

(2.61)

41.26

(2.41)

39.24

(2.38)

41.74

(2.45)

47.19

(2.94)

47.79

(3.01)

Full−time

282.01

(24.83)

285.16

(23.29)

288.00

(23.80)

282.58

(22.83)

285.12

(23.44)

287.44

(23.30)

281.03

(24.38)

281.36

(24.35)

Internship

−31.76

(−1.40)

−37.94

(−1.67)

−38.62

(−1.69)

−39.72

(−1.74)

−37.54

(−1.62)

−34.96

(−1.53)

−27.48

(−1.16)

−36.40

(−1.50)

Medium

−1.39

(−0.07)

−0.22

(−0.01)

−1.69

(−0.09)

2.54

(0.13)

−3.40

(−0.17)

−0.75

(−0.04)

2.87

(0.15)

0.46

(0.02)

Large

−16.60

(−0.72)

−7.85

(−0.33)

−9.15

(−0.39)

−11.29

(−0.47)

−10.05

(−0.42)

−13.22

(−0.55)

−10.98

(−0.47)

−11.95

(−0.52)

Type A

121.95

(3.44)

116.26

(3.17)

117.93

(3.46)

127.08

(3.34)

122.16

(3.44)

127.09

(3.50)

123.97

(3.31)

122.12

(3.43)

North-west

128.54

(5.21)

120.19

(4.67)

125.55

(5.04)

123.69

(4.81)

128.11

(5.03)

125.43

(4.92)

126.38

(5.07)

124.87

(5.09)

North-east

126.01

(5.11)

111.17

(4.31)

121.07

(4.82)

119.23

(4.70)

122.87

(4.83)

118.71

(4.62)

126.32

(5.04)

122.74

(5.03)

Centre

90.81

(3.90)

84.87

(3.54)

89.52

(3.75)

89.62

(3.73)

88.41

(3.69)

89.43

(3.79)

87.10

(3.72)

87.71

(3.88)

RSE 1

 

−8.62

(−3.52)

      

RSE 2

  

3.94

(1.29)

     

RSE 3

   

6.57

(1.84)

    

RSE 4

    

5.03

(1.61)

   

RSE 5

     

6.81

(2.28)

  

RSE 6

      

7.21

(2.61)

 

RSE 7

       

4.15

(1.45)

Constant

 

473.19

(7.59)

529.41

(8.29)

473.97

(6.56)

444.70

(5.93)

466.43

(6.39)

442.44

(6.22)

419.11

(5.90)

452.82

(6.69)

N

3,121

2,700

2,758

2,738

2,710

2,756

2,976

2,968

R2

0.45

0.45

0.45

0.44

0.45

0.45

0.45

0.45

All regressions include dummy variables for type of client served and sector of activity of the social enterprise. T stats are in parentheses. Standard errors are clustered at cooperative level. For the legend of intrinsic motivation items see Table 3

Regarding standard controls, we observe a strong, significant gender differential (females earn between 58 and 64 euros less than the average according to different specifications), a positive effect of seniority (around 10 euros more for every additional year of job experience) and macro-regional wage differentials which compensate for existing imbalances in purchasing power parities among Italian macro-areas (north-east and north-west, centre, and south).14 The insignificant effect of education confirms that the types of activities of social cooperatives do not remunerate skill premia, while the positive and significant effect of the Type A cooperative dummy is consistent with the hypothesis that work integration in Type B cooperatives has extra costs in terms of average productivity. In other words, hiring a disadvantaged worker reduces average productivity, which in turn negatively affects the average wage.

We run regressions with firm fixed effects, since the dataset encompassed many workers from each cooperative, and this makes it possible to eliminate unobserved firm-specific heterogeneity which may affect both ages and intrinsic motivation at the same time. However, the use of firm fixed effects makes it impossible to estimate the effects of regressors varying only at firm level (firm size, location and type in our case). Because of this reason we build a more parsimonious specification by estimating the specifications only with variables measured at the employee level (Table 6). In the new specification the relevance of our variables of interest is confirmed. By controlling for the between variability at the level of individual organizations the introduction of fixed effects not only does not weaken, but actually appears to strengthen, our results. This shows that the positive relation between motivations and wage does not disappear when controlling for unobserved organizational features.
Table 6

Wage regressions (net monthly wage, without bonuses), with fixed effects

Variable

(i)

(ii)

(iii)

(iv)

(v)

(vi)

(vii)

(viii)

RSE 1

 

−11.85

(−5.10)

 

 

 

 

 

 

RSE 2

 

 

6.03

(2.21)

 

 

 

 

 

RSE 3

 

 

 

8.21

(2.86)

 

 

 

 

RSE 4

 

 

 

 

7.82

(3.15)

 

 

 

RSE 5

 

 

 

 

 

8.59

(3.22)

 

 

RSE 6

 

 

 

 

 

 

4.56

(1.84)

 

RSE 7

 

 

 

 

 

 

 

7.26

(2.57)

Constant

 

761.63

(5.34)

813.50

(5.54)

718.71

(4.92)

700.92

(4.77)

707.30

(4.82)

730.10

(4.98)

745.81

(5.15)

716.12

(4.95)

N

3,121

2,700

2,758

2,738

2,710

2,756

2,968

2,976

R2

0.57

0.57

0.57

0.57

0.57

0.57

0.57

0.57

Fixed effects included. Excluded collinear dummies: north-east, north-west, centre, type A, large, medium. All regressions include dummy variables for the type of client served and sector of activity of the social enterprise. T stats are in parentheses. Standard errors are clustered by cooperative. For the legend of intrinsic motivation items see Table 3. Full regression estimates are omitted and available in an appendix upon request

Since Type A and Type B cooperatives have quite distinctive characteristics (see Footnote 8) we remove the restriction of identical slopes for the two types, and perform separate estimates with fixed effects. The significance of our intrinsic motivation variables is confirmed. The impact of RSE1 tends to be much greater in magnitude for Type B cooperatives, but the significance of RSE4 and RSE6 is stronger for Type A cooperatives (see Tables 7 and 8).
Table 7

Wage regressions (net monthly wage, without bonuses), type A only, with fixed effects

Variable

(i)

(ii)

(iii)

(iv)

(v)

(vi)

(vii)

(viii)

RSE 1

 

−9.397

(−3.78)

      

RSE 2

  

4.375

(1.26)

     

RSE 3

   

7.202

(2.02)

    

RSE 4

    

6.813

(2.06)

   

RSE 5

     

6.508

(2.00)

  

RSE 6

      

7.169

(2.19)

 

RSE 7

       

2.864

(0.75)

Constant 

500.57

(10.74)

983.47

(17.68)

650.92

(10.58)

534.84

(9.09)

796.22

(13.28)

612.98

(10.53)

616.93

(10.93)

664.38

(10.09)

N

2,422

2,113

2,158

2,147

2,114

2,163

2,312

2324

R2

0.54

0.53

0.54

0.53

0.53

0.53

0.54

0.54

Fixed effects included. Excluded collinear dummies: north-east, north-west, centre, type A, large, medium. All regressions include dummy variables for the type of client served and sector of activity of the social enterprise. T stats are in parentheses. Standard errors are clustered by cooperative. For the legend of intrinsic motivation items see Table 3. Full regression estimates are omitted and available in an appendix upon request

Table 8

Wage regressions (net monthly wage, without bonuses), type B only, with fixed effects

Variable

(i)

(ii)

(iii)

(iv)

(v)

(vi)

(vii)

(viii)

RSE 1

 

−18.011

(−2.72)

      

RSE 2

  

13.316

(1.63)

     

RSE 3

   

9.440

(1.48)

    

RSE 4

    

10.704

(1.53)

   

RSE 5

     

16.579

(2.06)

  

RSE 6

      

7.275

(1.22)

 

RSE 7

       

8.015

(1.34)

Constant

291.32

(4.02)

466.54

(5.59)

224.13

(2.33)

301.49

(3.10)

256.10

(2.48)

244.30

(2.45)

231.00

(3.33)

241.87

(3.44)

N

699

587

600

591

596

593

664

644

R2

0.65

0.68

0.67

0.68

0.67

0.67

0.66

0.65

Fixed effects included. Excluded collinear dummies: north-east, north-west, centre, type A, large, medium. All regressions include dummy variables for the type of client served and sector of activity of the social enterprise. T stats are in parentheses. Standard errors are clustered by cooperative. For the legend of intrinsic motivation items see Table 3. Full regression estimates are omitted and available in an appendix upon request

As a robustness check of our findings we slightly change our dependent variable (Table 9, line 1) by considering the net after-tax monthly wage plus bonus and premia (Table 9, line 2). The rationale for looking at this indicator is that bonuses and premia are not necessarily unexpected. As a consequence, rational workers should already incorporate expectations relating to these extraordinary items in their total expected after-tax remuneration. The results are similar, but with some differences: RSE2 becomes significant, while RSE5 and RSE6 become insignificant. By adding bonuses and premia to the wage in the dependent variable, the effect of variables measuring self-regarding intrinsic motivations (RSE1, RSE2 and RSE3) is enhanced, while that of variables measuring other-regarding intrinsic motivations (RSE6) is weakened. This is consistent with two hypotheses: (i) self-regarding intrinsic motivations lead to higher productivity and wages, and (ii) bonuses and premia stimulate productivity (only or mainly) where individuals have self-regarding intrinsic motivations.
Table 9

Effect of intrinsic motivations on wages and extra hours

Dependent variable

Regressor

RSE 1

RSE 2

RSE 3

RSE 4

RSE 5

RSE 6

RSE 7

Wage

−8.62

(−3.52)

3.94

(1.29)

6.57

(1.84)

5.03

(1.61)

6.81

(2.28)

7.21

(2.61)

4.15

(1.45)

Wage plus bonus

−15.65

(−3.79)

10.22

(2.08)

10.68

(1.81)

6.81

(1.31)

7.69

(1.59)

5.71

(0.91)

0.20

(0.04)

Hourly wage with extra hours

−0.15

(−1.21)

0.13

(1.09)

0.37

(2.67)

0.02

(0.18)

0.26

(1.88)

0.17

(1.25)

0.15

(1.32)

Hourly wage without extra hours

−0.22

(−2.06)

0.13

(1.03)

0.43

(3.23)

0.17

(1.54)

0.20

(1.51)

0.10

(0.76)

0.31

(2.57)

Extra hours

−0.03

(−0.69)

−0.01

(−0.15)

0.05

(0.95)

0.10

(2.42)

0.02

(0.43)

0.12

(2.32)

0.13

(2.97)

Ln of wage

−0.01

(−3.40)

0.00

(0.86)

0.01

(1.75)

0.01

(1.66)

0.01

(2.03)

0.01

(2.49)

0.00

(1.16)

Ln of (wage plus bonus)

−0.02

(−4.16)

0.00

(1.00)

0.01

(1.39)

0.01

(1.19)

0.01

(1.36)

0.01

(2.13)

0.00

(0.29)

Ln of hourly wage

0.00

(−1.34)

0.00

(0.87)

0.01

(1.76)

0.00

(0.42)

0.01

(2.01)

0.00

(0.82)

0.00

(0.83)

T stats are in parentheses. Standard errors are adjusted for clusters at cooperative level. Regressors are the same as in Tables 5, 6, 7, and 8. All regressions include dummy variables for the type of client served and sector of activity of the social enterprise. Column header variables (consensus on the following items expressed on a 0–7 Likert scale)

A third dependent variable considered in our analysis is the hourly wage. This is not easy to define. Our data measure contractual hours, but also—and separately—extra hours (overtime) which may or may not be remunerated.15 Intrinsically motivated workers are expected to provide extra effort (and part of it is probably not going to be paid), which should translate into more extra hours. Hence, by including extra hours as well, we expect a lower (positive) impact of intrinsic motivations on wages.

The results presented in Table 9 (third line) confirm our expectations, and show that only two indicators (RSE3 and RSE5 (but only weakly)) remain positive and significant when we include extra hours in the calculation of the hourly wage. To check whether the conjecture that intrinsically motivated workers have more overtime hours is correct we recalculate hourly wage by excluding extra hours from the denominator. In this case, we have three significant indicators: RSE1, RSE3 and RSE7 (Table 9, line 4).16 The findings documented in Table 9 appear to be robust to the introduction of organizational fixed effects (Table 10).17
Table 10

Effect of intrinsic motivations on wages and extra hours, with fixed effects

Dependent variable

Regressor

RSE 1

RSE 2

RSE 3

RSE 4

RSE 5

RSE 6

RSE 7

Wage

−11.846

(−4.88)

6.028

(1.89)

8.208

(2.59)

7.816

(2.60)

8.589

(2.85)

7.258

(2.52)

4.557

(1.44)

Wage plus bonus

−18.319

(−3.81)

14.626

(2.60)

13.737

(2.30)

13.037

(2.40)

11.574

(2.23)

6.641

(0.92)

5.734

(0.90)

Hourly wage with extra hours

−0.114

(−0.81)

0.215

(1.63)

0.370

(2.22)

0.042

(0.35)

0.299

(1.86)

0.200

(1.25)

0.150

(1.14)

Hourly wage without extra hours

−0.28

(−2.21)

0.10

(1.32)

0.51

(3.02)

0.12

(1.31)

0.15

(1.32)

0.12

(0.54)

0.23

(2.31)

Extra hours

−0.037

(−0.91)

−0.016

(−0.33)

0.057

(0.98)

0.106

(2.39)

0.016

(0.25)

0.079

(1.44)

0.092

(1.97)

Ln of wage

−0.014

(−4.17)

0.005

(1.28)

0.008

(1.80)

0.010

(2.73)

0.008

(2.23)

0.009

(2.22)

0.004

(0.91)

Ln of (wage plus bonus)

−0.018

(−4.39)

0.008

(1.62)

0.008

(1.60)

0.012

(2.64)

0.008

(1.87)

0.010

(1.92)

0.003

(0.69)

Ln of hourly wage

−0.005

(−1.31)

0.005

(1.21)

0.006

(1.24)

0.002

(0.58)

0.010

(2.06)

0.004

(0.88)

0.002

(0.55)

T stats are in parentheses. Standard errors are adjusted for clusters at cooperative level. Regressors are the same as in Tables 5, 6, 7, and 8. All regressions include dummy variables for the type of client served and sector of activity of the social enterprise. Column header variables (consensus on the following items expressed on a 0–7 Likert scale)

4.1 Principal component analysis

Each of the seven proxies considered so far captures only certain limited and specific aspects of self- or other-regarding intrinsic motivations (lack of non-monetary motivations, presence of intrinsic motivations related to personal growth, identification with the cooperative’s goals, community and relational elements in the job, social engagement, and non-monetary motivations measured before entering the cooperative). Each of these items taken individually is an imperfect proxy for the focus of our investigation, as workers can be characterized by more than one kind of intrinsic motivation. This clearly emerges from the pairwise correlation between the items (Table 11).
Table 11

Pairwise correlation coefficients among intrinsic motivation variables

 

RSE1

RSE2

RSE3

RSE4

RSE5

RSE6

RSE7

RSE1

1

−0.209

−0.345

−0.368

−0.344

−0.083

−0.081

RSE2

 

1

0.359

0.316

0.391

0.114

0.111

RSE3 

  

1

0.493

0.417

0.130

0.163

RSE4

   

1

0.447

0.102

0.096

RSE5

    

1

0.223

0.124

RSE6

     

1

0.309

RSE7

      

1

See Table 3 for a description of the variables

Hence, we decide it is appropriate to transform the larger number of correlated variables into a smaller number of uncorrelated ones revealing the internal structure of the data in a way which may capture their variability better and more parsimoniously. This is why we apply principal component analysis to the vector of the seven considered proxies. This solution appears to be superior to (i) the inclusion of single indicators, given the inadequacy of each of them taken individually for giving a complete picture of the phenomenon and (ii) the inclusion of all the indicators in the same specification, given the redundancy of information.

Table 12 documents the relevance of the first component, which captures 39 % of data variability. This component is negatively correlated with the first item and positively correlated with all the others, with correlation coefficients which never fall below 26 % (Table 13).18
Table 12

Principal component analysis (PCA)—loadings

Component

Eigenvalue

Difference

Proportion

Cumulative

S1

2.73

1.41

0.39

0.39

S2

1.32

0.55

0.19

0.58

S3

0.77

0.14

0.11

0.69

S4

0.63

0.04

0.09

0.78

S5

0.59

0.10

0.08

0.86

S6

0.49

0.03

0.07

0.93

S7

0.46

 

0.07

1.00

S stands for principal component

Table 13

Principal components

Variable

Correlations

S1

S2

S3

S4

S5

S6

S7

RSE1

−0.37

0.16

0.66

0.55

0.30

0.00

0.07

RSE2

0.37

−0.20

0.72

−0.35

−0.36

0.21

−0.14

RSE3 

0.45

−0.14

0.00

0.52

−0.32

−0.64

−0.04

RSE4

0.44

−0.22

−0.17

0.49

0.19

0.66

−0.16

RSE5

0.44

−0.14

0.12

−0.24

0.71

−0.24

0.38

RSE6

0.26

0.66

0.02

−0.09

0.24

−0.12

−0.65

RSE7

0.27

0.65

0.00

0.07

−0.27

0.21

0.62

See Table 3 for a description of the variables

The Kaiser–Meyer–Olkin measure of sampling adequacy (.76) (Kaiser and Rice 1974) excludes the selected variables having too little in common to warrant a factor analysis. We therefore repeat estimates for all the different dependent variables considered so far by replacing the individual items proxying intrinsic motivations with the first principal component. Our results now show positive and significant effects in all cases (including that of the hourly wage which includes overtime at the denominator), where fixed effects are not included (Table 14). The inclusion of fixed effects (Table 15) tends to strengthen the results for the wage and the wage plus bonus (columns 1 and 2), while results for the hourly wage appear substantially unchanged. This confirms the existence of an impact of individual intrinsic motivations on wages. The results are weaker in the cases of extra hours and holidays in arrears (columns 6 and 7), on the other hand, signalling the presence of unobserved organizational effects.19
Table 14

Regression results with principal component

Regressor

Dependent variables

(i)

(ii)

(iii)

(iv)

(v)

(vi)

(vii)

Wage + bonus

Wage

Ln of wage

Hourly wage

Ln of h. wage

Hol. in arr.

Extra h.

Male

73.99

(3.38)

62.25

(5.93)

0.07

(5.61)

1.18

(2.05)

0.03

(2.13)

0.03

(0.05)

0.61

(2.37)

Age

0.92

(0.80)

0.39

(0.54)

0.00

(0.89)

0.05

(1.55)

0.00

(1.69)

−0.02

(−0.51)

−0.02

(−1.87)

Education

1.67

(0.83)

0.18

(0.14)

0.00

(0.52)

−0.03

(−0.59)

0.00

(−0.82)

0.03

(0.35)

0.00

(0.24)

Italian

−89.42

(−1.33)

−88.92

(−1.73)

−0.10

(−2.07)

−0.04

(−0.06)

0.00

(−0.11)

−1.52

(−0.71)

−0.27

(−0.46)

Member

24.06

(1.23)

−0.07

(−0.01)

0.01

(0.32)

−1.47

(−2.63)

−0.05

(−2.87)

0.76

(0.67)

0.43

(1.94)

Years in coop

18.43

(6.16)

10.69

(6.87)

0.01

(7.21)

0.27

(5.26)

0.01

(6.22)

0.20

(2.03)

0.00

(0.10)

Permanent

71.37

(3.36)

36.84

(2.14)

0.06

(2.58)

−1.06

(−1.22)

−0.01

(−0.28)

4.68

(5.67)

0.17

(0.92)

Full-time

306.39

(14.78)

285.88

(21.84)

0.39

(24.46)

−2.35

(−4.18)

−0.06

(−4.50)

1.61

(2.32)

0.43

(2.65)

Internship

−70.46

(−2.29)

−30.83

(−1.32)

−0.05

(−1.26)

4.75

(1.08)

0.01

(0.09)

−1.53

(−1.19)

0.73

(1.12)

Medium

19.97

(0.66)

−0.71

(−0.03)

0.02

(0.66)

0.01

(0.02)

−0.01

(−0.44)

0.31

(0.29)

0.16

(0.48)

Large

32.88

(0.96)

−5.78

(−0.23)

0.00

(0.04)

0.35

(0.41)

−0.01

(−0.29)

0.89

(0.74)

−0.15

(−0.41)

Type A

163.32

(2.79)

122.18

(3.31)

0.18

(3.43)

4.94

(4.01)

0.19

(4.57)

−2.02

(−1.03)

−0.37

(−0.46)

North−west

187.91

(4.87)

128.50

(4.96)

0.17

(5.45)

1.89

(2.19)

0.08

(2.60)

4.41

(2.79)

0.46

(1.45)

North−east

132.19

(3.24)

121.07

(4.65)

0.18

(5.46)

1.63

(1.81)

0.06

(1.95)

2.51

(1.96)

0.30

(1.04)

Centre

100.55

(2.49)

91.36

(3.79)

0.13

(4.16)

1.91

(1.66)

0.05

(1.74)

2.90

(2.03)

0.58

(1.60)

S1

12.39

(2.03)

8.84

(2.60)

0.01

(2.37)

0.28

(2.37)

0.01

(2.00)

0.44

(1.84)

0.08

(1.81)

Constant

 

320.80

(2.77)

491.78

(7.03)

6.13

(69.96)

21.61

(11.78)

3.02

(46.77)

−0.18

(−0.05)

1.80

(1.74)

N

2,516

2,516

2,516

1,959

1,959

2,672

2,166

R2

0.33

0.44

0.47

0.06

0.11

0.11

0.09

T stats are in parentheses. Standard errors are clustered at cooperative level. All regressions include dummy variables for the type of client served and sector of activity of the social enterprise. S1 = first principal component (see Tables 12, 13)

Table 15

Regression results with principal component, with fixed effects

Regressor

 

Dependent variable

(i)

(ii)

(iii)

(iv)

(v)

(vi)

(vii)

Wage + bonus

Wage

Ln of wage

Hourly wage

Ln of h. wage

Hol. in arr.

Extra h.

Male

83.48

(3.32)

51.05

(4.67)

0.06

(4.54)

1.23

(1.58)

0.03

(1.69)

0.13

(0.21)

0.40

(1.88)

Age

1.04

(0.88)

0.47

(0.61)

0.00

(0.50)

0.05

(1.46)

0.00

(1.27)

−0.04

(−1.35)

−0.01

(−0.80)

Education

2.06

(1.00)

0.48

(0.39)

0.00

(0.53)

0.03

(0.63)

0.00

(0.34)

−0.04

(−0.53)

0.02

(1.03)

Italian

3.60

(0.07)

−18.32

(−0.64)

−0.05

(−1.43)

0.51

(0.58)

0.01

(0.42)

−2.51

(−1.16)

0.22

(0.56)

Member

44.16

(1.84)

23.93

(1.77)

0.03

(1.86)

−0.79

(−1.15)

−0.03

(−1.27)

1.77

(3.19)

0.57

(2.01)

Years in coop

18.61

(5.45)

10.25

(6.89)

0.01

(7.07)

0.27

(3.89)

0.01

(4.98)

0.29

(3.76)

0.00

(−0.03)

Permanent

103.10

(4.39)

54.35

(3.12)

0.07

(3.14)

−1.24

(−1.07)

−0.01

(−0.32)

3.85

(5.33)

0.15

(0.58)

Full-time

253.74

(10.27)

261.88

(15.91)

0.35

(19.58)

−2.96

(−3.81)

−0.08

(−4.64)

1.25

(1.88)

0.42

(2.33)

Internship

−77.06

(−2.16)

−43.51

(−1.67)

−0.03

(−0.79)

6.06

(1.05)

0.04

(0.46)

−1.27

(−0.78)

0.26

(0.31)

Medium

144.29

(21.74)

−182.36

(−47.28)

−0.26

(−48.12)

4.00

(5.48)

0.16

(8.10)

−7.44

(−10.97)

−1.92

(−7.39)

Large

116.61

(4.12)

6.65

(0.35)

0.02

(0.99)

0.06

(0.10)

−0.02

(−1.26)

−3.36

(−5.49)

0.63

(2.82)

Type A

−169.62

(−6.09)

111.52

(7.39)

0.33

(16.59)

−0.85

(−0.33)

0.01

(0.10)

3.71

(3.41)

−0.88

(−2.02)

North-west

273.89

(8.51)

40.55

(2.67)

0.06

(2.59)

5.05

(4.01)

0.28

(12.04)

14.99

(21.10)

−2.29

(−22.65)

North-east

71.19

(1.57)

142.72

(5.47)

0.26

(6.79)

6.05

(6.45)

0.25

(13.66)

16.41

(19.65)

−2.59

(−9.97)

Centre

−120.37

(−4.17)

−47.99

(−2.85)

−0.09

(−3.44)

6.67

(10.49)

0.27

(16.89)

17.31

(28.50)

−1.81

(−8.21)

S1

20.34

(2.81)

12.38

(3.70)

0.01

(2.82)

0.31

(2.34)

0.01

(1.90)

0.23

(1.22)

0.06

(1.14)

Constant

 

426.62

(4.64)

478.47

(9.57)

6.08

(106.46)

20.31

(6.23)

2.98

(36.17)

−7.80

(−2.67)

1.33

(1.54)

N

2,516

2,516

2,516

1,959

1,959

2,672

2,166

R2

0.46

0.57

0.59

0.21

0.29

0.35

0.28

T stats are in parentheses. Standard errors are clustered at cooperative level. All regressions include dummy variables for the type of client served and sector of activity of the social enterprise. S1 = first principal component (see Tables 12, 13)

4.2 Intrinsic motivations, productivity and job donation

The previous sections identified a significant positive correlation between intrinsic motivations and wages, which seems to contradict the standard donative-labour hypothesis. We aim to show by additional empirical analysis that our findings are consistent with the mutual validity of the donative-labour and intrinsic motivation-productivity hypotheses. To demonstrate this, we decompose the donative-labour hypothesis into two testable hypotheses: (i) intrinsically motivated workers donate more work hours, and (ii) the wage of intrinsically motivated workers is significantly lower than that of non-intrinsically motivated workers.

The first part of the hypothesis (workers with stronger intrinsic motivations donate more work) is only weakly confirmed by our results on the effect of some of our proxies and the principal component on holiday in arrears and extra hours (Tables 14, 15, columns 6 and 7), since the inclusion of fixed effects may signal the presence of concomitant organizational effects interacting with motivations. This is likely to be due to unobserved idiosyncratic organizational dimensions, which may relate, for example, to the “organizational climate” in terms of interactional and procedural fairness (i.e. a perceived reduction of management fairness may typically crowd out intrinsic motivations at cooperative levels, leading to a reduction of extra hours worked) (Tyler and Blader 2000).20

Do we have a measure of the link between productivity and intrinsic motivations which is independent from wages? We already obtain signals of this higher productivity when we find that intrinsic motivations lead to higher overtime and arrears of holidays. We find additional evidence by looking at the answers to a question in which workers are asked to judge their effort relative to: (i) what is required by the cooperative, (ii) the needs of the beneficiaries of the cooperatives’ services, (iii) the effort of other workers in the cooperative, and (iv) their pay.

In all these cases, most of the proxies of intrinsic motivation (and their principal component) are strongly significant and positive, net of the impact of other standard controls with and without the inclusion of fixed effects at the cooperative level (Tables 16, 17).21 This implies that, according to respondents’ self-evaluation, intrinsically motivated workers attribute a higher level of effort to themselves with respect to (i) their co-workers, (ii) what is required by the cooperative, and (iii) what is needed by the beneficiaries, (iv) their pay. Under the reasonable assumption that effort is positively related to productivity, more intrinsically motivated workers are more productive, if we are to believe their answers. Our findings are consistent with other empirical results showing the importance of intrinsic motivations, commitment, and teamwork for guaranteeing improved organizational performance (Akehurst et al. 2009). One criticism might be that intrinsically motivated workers have reasons for providing biased answers to these questions in order to provide a consistent picture of their “diligent worker” image. Since, however, here we are comparing their answers with those of non-intrinsically motivated workers, there is no particular reason why non-intrinsically motivated workers should not try to overstate their effort as well.
Table 16

Effect of intrinsic motivations on effort levels

Dependent variable

Regressor

RSE 1

RSE 2

RSE 3

RSE 4

RSE 5

RSE 6

RSE 7

S1

Eff. 1

−0.03

(−2.19)

0.06

(4.70)

0.06

(3.98)

0.02

(1.40)

0.04

(2.79)

0.08

(5.08)

0.09

(6.45)

0.08

(4.82)

Eff. 2

0.00

(−0.22)

0.04

(3.04)

0.05

(3.56)

0.01

(1.16)

0.03

(2.07)

0.10

(7.10)

0.11

(7.47)

0.06

(4.42)

Eff. 3

−0.01

(−0.76)

0.05

(3.25)

0.05

(3.31)

0.01

(1.16)

0.02

(1.47)

0.07

(4.46)

0.03

(2.09)

0.05

(4.08)

Eff. 4

0.01

(0.46)

0.02

(1.62)

0.02

(1.76)

0.00

(−0.23)

−0.02

(−1.19)

0.08

(5.63)

0.06

(4.29)

0.03

(2.01)

Results come from ordered probit regressions. T stats are in parentheses. Standard errors are clustered at cooperative level. Regressors are the same as in Tables 5, 6, 7, and 8.

All regressions include dummy variables for the type of client served and sector of activity of the social enterprise. For the legend of intrinsic motivation items see Table 3

Table 17

Effect of intrinsic motivations on effort levels, with fixed effects

Dependent variable

Regressor

RSE 1

RSE 2

RSE 3

RSE 4

RSE 5

RSE 6

RSE 7

S1

Eff. 1

−0.01

(−2.24)

0.03

(3.21)

0.03

(2.93)

0.04

(1.42)

0.02

(2.14)

0.03

(3.42)

0.03

(4.32)

0.03

(3.12)

Eff. 2

0.01

(−0.55)

0.03

(2.32)

0.02

(2.94)

0.03

(1.02)

0.02

(2.32)

0.13

(4.24)

0.09

(5.32)

0.03

(3.21)

Eff. 3

−0.3

(−0.46)

0.03

(3.12)

0.02

(3.12)

0.04

(1.03)

0.01

(1.32)

0.04

(3.43)

0.02

(2.32)

0.04

(2.54)

Eff. 4

0.03

(0.56)

0.03

(1.64)

0.04

(1.65)

0.001

(−0.32)

−0.05

(−1.04)

0.06

(3.94)

0.04

(4.12)

0.05

(1.93)

Results come from ordered probit regressions. T stats are in parentheses. Standard errors are clustered at cooperative level. Regressors are the same as in Tables 5, 6, 7, and 8. All regressions include dummy variables for the type of client served and sector of activity of the social enterprise. For the legend of intrinsic motivation items see Table 3

To conclude, if we define productivity as the ratio between the contribution to a cooperative’s output and the wage, we can say that intrinsically motivated workers are more productive because the ratio between effort at the nominator, and pay, the needs of the cooperative, or the efforts of other workers at the denominator, is positively related to intrinsic motivations. As we have seen, motivated workers also appear to postpone their holidays and work longer hours. This result does not receive strong econometric support from fixed effects estimates. However, from a descriptive point of view, it is remarkable to see that, in spite of the larger amount of extra hours (some of which are unpaid), which may have negative effects on hourly wages, intrinsically motivated workers receive higher pay even when measured in terms of hourly wage. One plausible explanation for this is that intrinsic motivations create extra productivity, which compensates for the amount of work donated and leads to higher pay.22

So far, we have measured the effects of our regressors on the sample means of our dependent variables. As a robustness check, we can verify whether the documented effects also work in other relevant moments of the dependent variable distribution. To this end, we estimate quantile regressions (Koenker and Bassett 1978) in a robustness check adopted in several wage differential studies (see, for example, the set of studies collected in Fitzenberger et al. 2002, and the application to the Italian labour market by Naticchioni et al. 2007). More specifically, we estimate the impact of our controls on the 25th, median and 75th percentiles of the dependent cumulative distribution. Our principal component has a significant impact on all of the three levels of the distribution considered, which confirms that the effect of intrinsic motivation is not limited to the mean wage (Table 18).
Table 18

Coefficient of the first principal component (S1) from quantile regressions for selected variables

Dependent variable

q25

q50

q75

Wage

7.90

(3.14)

9.47

(3.78)

6.53

(1.96)

Hourly wage

0.08

(0.84)

0.15

(1.74)

0.21

(2.52)

Ln of wage

0.01

(4.25)

0.01

(3.79)

0.01

(2.80)

Ln of hourly wage

0.00

(1.07)

0.01

(1.89)

0.01

(2.67)

Results come from quantile regressions. T stats are in parentheses. Standard errors are adjusted for clusters at the cooperative level. Regressors are the same as in Tables 5, 6, 7, and 8. All regressions include dummy variables for the type of client served and sector of activity of the social enterprise

5 Conclusions

The results presented in this paper shed light on an important, and previously unexplored, aspect of the relationship between intrinsic motivations, wages and productivity. Our starting point is the debate on compensating differentials. The findings from several empirical studies in the literature on profit/non-profit differentials have been interpreted as evidence which documents the relationship between intrinsic motivations and wages. In many of these studies, non-profit workers have been shown to earn significantly less than their counterparts in the for-profit industry. This finding has been interpreted as being consistent with the donative-labour hypothesis: intrinsically motivated workers find superior non-pecuniary compensation when working in the non-profit industry, and therefore agree to sacrifice a part of their wages in return. The different kinds of satisfaction provided by intrinsic motivations therefore act as a compensating differential of the profit/non-profit wage difference.

However, there are some important limitations to what the profit/non-profit differential can tell us about the relationship between wages and intrinsic motivation. It is true that, in the absence of frictions between the two sectors, workers from the non-profit sector may move to the for-profit sector, and therefore intrinsic motivations must be relevant. The story may be less convincing in the presence of frictions, however. In addition, the literature has found many other factors which affect the for-profit/non-profit wage gap (softer budget constraints in the non-profit industry where it is highly subsidized, and differences in quality between workers in the two sectors due to self-selection, etc.) which prevent us from interpreting the differentials as the effect of intrinsic motivations alone. It has not been possible to overcome these limitations in the literature to date, owing to the impossibility of measuring directly intrinsic motivations.

In this paper we have endeavoured to shed light on the issue by looking at the relationship between intrinsic motivations and wages within the cooperative (non-profit) sector, and exploiting the benefit of measuring directly intrinsic motivations by means of subjective data on the self-evaluations of 4,134 paid workers employed by 320 Italian social enterprises. Our findings relate to one organizational type (the non-profit social cooperative) and one sector (social services) only. Hence, they do not allow us to test for the differential importance of intrinsic motivations in different organizational forms and sectors. They do, however, permit testing of the impact of intrinsic motivations on labour supply and demand within a homogeneous organizational context. Our findings do not confirm—although they are not incompatible with—the core of the donative-labour hypothesis as traditionally stated in the literature. We find only weak evidence (when not controlling for fixed effects at the level of the cooperative) that more intrinsically motivated workers may have more holidays in arrears and hours of unpaid overtime. Hence, we are not in a position to fully confirm that intrinsic motivations lead workers to accept a lower pay ex-post where they are driven by an ex-ante non-pecuniary interest in the activity they perform. Instead, we document the existence of a positive correlation between intrinsic motivations and wages, and explain it by the intrinsic motivation-productivity hypothesis, which states that intrinsically motivated workers are more productive, and therefore end up earning more.

The total effect of motivations on wages is to be interpreted as the net effect of the movements of both the supply and the demand curves of labour for intrinsically motivated workers. While rightward shifts in the labour supply curve would correspond to the donative-labour hypothesis, as they tend to lower the equilibrium level of wages, rightward shifts in the labour demand curve are due to increased per-hour productivity, and correspond to the intrinsic motivation-productivity hypothesis. They lead to a higher equilibrium wage. Although we are not able to measure the donative and productivity effects separately, our results show unambiguously that the productivity effect overcomes any donative effect, and engenders a net upward dynamic effect in productivity. The positive correlation between motivational variables and wages is limited. It amounts to about 1 % of total wages, that is, about 10 Euros out of a total of wages plus bonuses of 944 Euros a month. However, it signals a forgotten role of motivations in making production processes more stable and efficient, mainly in the long term.

We believe that our approach is superior to that employed previously in the literature. In the absence of data which directly measure intrinsic motivations, the existing tests of the donative-labour hypothesis have looked at the profit/non-profit wage differential under the constant assumption that greater levels of satisfaction from intrinsic motivations are to be found in the non-profit industry. This encroaches on many potential confounding factors, given the heterogeneous features of the profit and non-profit industries. Our results on extra-hours worked and holidays in arrears show that these factors can also be found at the level of organizations, since for-profit and non-profit industries may be characterised by different governance and management styles.

Our paper therefore represents a step forward because: (i) we directly measure intrinsic motivations and relate them to wages and effort, and (ii) we reduce industry heterogeneity by working within the same sector. The clear advantage of our study is that, unlike in previous works, the risk of omitted variable bias raised by Leete (2001) is limited, since (i) we consider only one sector, and (ii) our survey provides an impressive number of control variables. Direct measurement of intrinsic motivations also allows us to exploit within sector motivational variability. Indeed, contrary to common belief, the non-profit sector shows significant differences among workers in both wages and motivations. One reason for the variability in intrinsic motivations is that the low level of salaries means that applicants range from individuals with high intrinsic motivations (who are prepared to accept a lower salary in order to achieve a social goal) to people with high extrinsic motivations (who lack better opportunities in the profit sector, and end up in the non-profit sector as a second-best solution).

Our results can have relevant consequences in terms of policies and industrial relations. They reveal that intrinsic motivations are important (and still only partially explored) drivers of productivity. Even though we do not have valid instruments available to verify whether the relationship includes a causality nexus from intrinsic motivations to productivity, we record that more intrinsically motivated workers report that they make a relatively higher effort. They suggest that a reduced distance between corporate goals and employees’ intrinsic motivations may be associated with significant productivity gains. Managerial policies specifically designed to stimulate intrinsic motivations, especially in the long run and in connection with workers’ commitment and loyalty to the organization, are likely to deliver important results, for example, when it comes to reinforcing stability and resilience to worker turnover. Furthermore, the impact of intrinsic motivations may help to explain a part of the variability of corporate performances which is not accounted for after controlling for the contribution of traditional productive inputs. Without concealing certain weaknesses in our results, we believe that future research on economy-wide samples will confirm many of the hypotheses that have inspired this work.

Footnotes
1

Following Preston (1989), the acceptance of less pay by intrinsically motivated workers is equivalent to a monetary donation to an organization which produces social benefits. In this sense, these workers show a strong willingness to pay for public goods. Frank (1996) emphasizes that intrinsic motivations are a component of job amenities, and therefore interprets the difference in wages as a compensating differential. Rose-Ackerman (1996) argues that it is the alignment between workers’ ideals and corporate goals which leads workers to accept lower pay. Finally, Hansmann (1980) interprets the differential as a sorting mechanism, by which workers who attach a relatively lower weight to pecuniary compensation and a relatively higher weight to contributing to public goods are hired in the non-profit industry. What all these rationales have in common is the capacity to explain why the differential persists, and why it is not bridged by a migration of workers from the non-profit to the profit sector.

 
2

This conclusion hinges on the assumption that below a given threshold of intrinsic motivations, workers would not accept lower pay in the non-profit industry.

 
3

As will be shown below, since we can measure the degree of intrinsic motivations of individual respondents, we are definitely within a framework of heterogeneity of intrinsic motivation (case (ii)).

 
4

In our data, the distribution of workers’ intrinsic motivations does not show particular irregularities (e.g. bimodal distribution), although all distributions are skewed in a pronounced manner either to the right or to the left.

 
5

The International Social Survey Programme on work orientations reports that more than 25 % of workers consider the fact that their job “allows them to help other people” and “is useful to society” as its most important characteristic. This share is equal in size to that of workers who value “high income” as the most important characteristic (see Clark 2005). This evidence on the relevance of non-pecuniary motivations is noteworthy. It implies that when deciding on job offers, a significant proportion of workers is driven by moral and other-regarding considerations, rather than by personal interest.

 
6

The Universities of Trento, Bergamo, Brescia, Milano Bicocca, Napoli Federico II and Reggio Calabria.

 
7

The Italian National Agency for Statistics.

 
8

Social cooperatives (cooperative sociali), like all types of cooperatives, are controlled by non-investor stakeholders, and deliver socially-oriented, public benefit services. According to Italian law (Special Law no. 381/1991), the goals of social cooperatives are the social integration of disadvantaged citizens, the general well-being of the community and human development. Social cooperatives are of two types: Type A social cooperatives manage health and education services, while Type B social cooperatives (also called work integration social enterprises) operate in industry, agriculture, trade or the service sector with the goal of including within the labour market “disadvantaged” workers (for example, the disabled, ex-prisoners, or ex-drug addicts), who must amount at least to 30 % of the workforce.

 
9

The extremely high response rate is due to the careful set-up of the survey. All the organizations involved were contacted in advance, and received the questionnaires. Involvement in and consensus on the survey was also achieved thanks to the support of representative meso-level bodies (for example, ConSolida, the Consortium of Social Cooperatives in the Trentino Province). Trained data collectors were sent on site, usually more than once. All the workers of an organization were involved where the organizations had up to 15 employees. For larger organizations, a random sample of workers was extracted on the basis of the three above-mentioned stratification criteria. Workers either compiled the questionnaires in groups with the support of the data collectors, or took the questionnaires home (in both cases, the questionnaires were always handed in to the collectors anonymously in envelopes), while late spare questionnaires were posted directly to the surveyors by the organizations.

 
10

The pyramid of needs designed by Maslow (1954) identifies five categories: physiological needs (or prime needs), needs of security (including stability and protection), needs of identification and involvement (both in a society and in groups), needs of esteem (such as self-esteem and other rewards), and needs of self-fulfilment (such as realization of personal and professional abilities).

 
11

These items are included in question d8 in the original questionnaire, which is omitted for reasons of space, and is available from the authors upon request.

 
12

These items are included in question d52 in the original questionnaire.

 
13

Distinguishing between Type A and Type B social cooperatives, we control for six Type A and 11 Type B service and product typologies. We further control for eight Type A and seven Type B user typologies.

 
14

Cannari and Iuzzolino (2009) calculate the cost of living by integrating the information provided by the Italian National Institute of Statistics (ISTAT) with data on other services (e.g. energy, public services, and transportation) from other official sources, and on real estate prices and rents. On this basis, they show that PPP correction raises the value of real wages in the south with respect to the centre and (even more so) to the north-east and north-west.

 
15

Unfortunately, the division between remunerated and unremunerated overtime hours is not available in the database.

 
16

A more direct check is obtained by estimating a regression where the number of extra hours is directly considered as a dependent variable. In this case, we have three significant and positive indicators (Table 4, line 4), which highlight the importance of pro-social, other regarding motivations in supporting the donative-labour hypothesis.

 
17

The results for type A and type B cooperatives only provide similar findings. They are omitted for reasons of space and available upon request.

 
18

Note that the second principal component has a much lower explanatory power (19 % of variance) but one which is still higher than one eigenvalue. When we include it in the estimate, however, its effect is not significant, and the impact of all other regressors is unchanged. Hence these results are omitted for reasons of space.

 
19

The significance of the first principal component is confirmed when the estimate is repeated for type A and type B cooperatives separately. In this case, the principal components are obviously recalculated on the restricted sample. The results are omitted for reasons of space and available upon request.

 
20

The results for type A and type B cooperatives only provide similar findings. They are omitted for reasons of space, and available upon request.

 
21

The impact of individual proxies of intrinsic motivations is positive and significant for all of the four relative effort measures for two indicators (RSE6 and RSE7), and is significant for all proxies of intrinsic motivations where the level of effort with respect to the needs of the social enterprise (EFF1) is considered.

 
22

Besides the positive effects on productivity and effort, intrinsic motivations are likely to be still more relevant to fostering worker well-being, in both its material and non-material components, as has already stated in the literature (Borzaga and Tortia 2006).

 

Acknowledgements

We are gratefully indebted to Kaushik Basu, Avner Ben-Ner, Carlo Borzaga, Luigino Bruni, Benedetto Gui, Alois Stutzer, Robert Sugden, an anonymous referee and participants to seminars held in Trento and Rome for their comments and suggestions. The usual disclaimer applies. We are also grateful to EuRICSE (European Research Institute on Cooperative and Social Enterprises, Trento, Italy) and to its staff for data access and support. The survey on which the empirical analysis is based was carried out between March 2004 and February 2008 by five research units at the Universities of Brescia, Milan, Naples, Reggio Calabria and Trento. The survey was granted financial support by: (1) the Italian Ministry for University and Scientific Research (MIUR) awarded to the national research project (PRIN) titled: “The Economic Role of Nonprofit Organizations: New Theoretical Developments and Empirical Tests”, and (2) the Foundation of the Saving Bank of the Lombard Provinces (Ca.Ri.P.Lo. Foundation).

Copyright information

© Springer Science+Business Media, LLC. 2012