How important is pro-social behaviour in the delivery of public services?
Paul Gregg, Paul Grout, Anita Ratcliffe, Sarah Smith and Frank Windmeijer
(CMPO, University of Bristol)
This version July 2009
Abstract
A number of papers have suggested that pro-social behaviour in the workplace may be crowded out in an institutional environment where there are high-powered incentives,butthere is little empirical research that attempts to test this directly using data on worker behaviour.This is the aim of this paper. Making the assumption that individuals in the non-profit sector are less likely to face high-powered incentives, we show that individuals in this sector are significantly more likely to donate their labour, measured by unpaid overtime, than those in the for-profit sector. Further analysispoints to selection as the most likely mechanism through which this relationship arises.
Key words: pro-social behaviour; public services; donated labour; motivation
JEL Classification:H11, J32, J45, L31, L32
1. Introduction
The themes of other-regarding preferences, intrinsic motivation, and pro-social behaviour have been ever- present in some form in economics, albeit often under different nomenclature (Smith, 1759, provides an early reference), but they have recently become the focus of mainstream economic research. There is now a large and growing body of theoretical and experimental research on intrinsic motivation and pro-social behaviour and their sensitivity to the institutional environment surrounding intrinsically-motivated individuals (see Meier, 2006, for an overview).
Within this research, a central question is whether pro-social behaviour in the workplace is crowded out in an institutional environment where there are high-powered incentives and /or aprofit motive. There are various theoretical mechanisms through which this crowding out might occur. One line of theoretical research emphasises individual self esteem and signalling motives (e.g., Benabou and Tirole, 2006, and Ellingson and Johannesson, 2008). Another focuses on the role of the profit motive in diverting the benefits of pro-social behaviour toward greater profit rather than better service, dampening employees’ incentive to engage in pro-social behaviour (e.g., Francois, 2000, 2001, 2003, 2007). Another emphasises the greater trust of individuals and organisations in the absence of for-profit incentives (e.g., Arrow, 1975,Hansmann, 1980, Rose-Ackerman, 1996, and Glaeser and Shleifer, 2001). Finally, it has been suggested that ‘mission-oriented’ individuals (those who are pro-socially motivated) will be attracted to organisations with a similar mission (Besley and Ghatak, 2003, 2005). In this context high powered incentives weaken the matching process.[1]
A number of field and laboratory experiments have been used to explore how pro-social behaviour is affected by the institutional environment. Most of the findings are consistent with the idea that high powered incentives and the profit motive reduce pro-social behaviour (e.g., Ariely, Bracha and Meier, 2009, Frey and Oberholzer, 1997, Fehr and List, 2004, Gneezy and Rustichini, 2000, Mellstrom and Johannesson, 2008). However, while these experiments cover a broad variety of situations and have diverse approaches, there has to date been little analysis of the relationship between pro-social behaviour and institutional environment using data on individuals’ behaviour in the workplace.
Using data from the most recent wave of the World Values Survey (WVS) we can show evidence of a systematic relationship across countries between institutional environment and employees’ intrinsic motivation. The WVS is a survey of individuals and lacks detailed information on workplace incentives. However, making the assumption that individuals in the non-profit sector (combining both public and not-for-profit sectors) are less likely to face high-powered incentives, we can compare employee motivations across the for-profit and non-profit sectors.[2]
Figure 1.a shows that people working in the non-profit sector are less likely to be motivated by money than people working in the for-profit sector. This result is true to varying degrees across the selected group of ten countries shown here, and also holds across almost all other countries in the WVS sample, with the exception of India.[3] The proportion citing a safe job as their primary motivation is, perhaps surprisingly, not higher in the non-profit sector in all countries, although is a particularly important factor in China (Figure 1.b). By contrast, the proportion saying that their main motivation when they look for a job is “doing something important that gives you a feeling of accomplishment” is higher in the non-profit sector in all ten countries.
Figure 1: Primary motivation when looking for a job: Differences between non-profit and for-profit workers
Countries: UK = United Kingdom, CHI = China, UK = United States, AUS = Australia, SWE = Sweden, FRA = France, RUS = Russia, JAP = Japan, IND = India.
This is evidence of a systematic difference in motivation between workers in the for-profit and non-profit sectors across a diverse selection of countries. This finding is consistent with the results from other, single-country studies (e.g. Frank and Lewis, 2004). However, with this kind of self-reported motivation data there is always a potential concern that they are capturing a halo effect rather than differences in behaviour; ideally, we would like to see differences in how people behave not just differences in self-reported motivations.
The aim of this paper is to provide evidence on the relationship between pro-social behaviour and institutional structure based on actual behaviour in the workplace using unpaid overtime -donated labour - as a measure of pro-social behaviour. We use data from the British Household Panel Survey (BHPS) which, as discussed further in section 3, has the required information on sector of employment and on unpaid overtime. We also exploit the panel to look at what happens when individuals change sector.
The next section briefly outlines our empirical strategy, while section 3 contains further details on the data and definitions of key variables. In section 4 we show that individuals in the non-profit sector are indeed significantly more likely to donate labour, controlling for a wide range of individual- and job-specific characteristics. In section 5 we estimate a simple fixed effects panel data model which shows no evidence that individuals change their donated labour when they switch sector. We can therefore reject that the observed relationship in the cross-section arises simply as a result of implicit contracts. It also rules out any institutional effects operating on individuals’ behaviour, implying that the association between pro-social behaviour and sector of employment must arise as a result of selection. In section 6 we present evidence consistent with this explanation. Section 7 concludes.
2. Empirical Approach
Our primary aim is to explore whether pro-social behaviour is more prevalent in the non-profit sector. We use unpaid overtime as our measure of pro-social behaviour and estimate the probability that an individual does any unpaid overtime using a linear probability model. We show below that the greatest variation is in this extensive margin. We find evidence that individuals in the non-profit sector are significantly more likely to do unpaid overtime. We include four binary indicators representing the non-profit and for-profit “caring” sectors (defined as health, education and social care) and the non-profit and for-profit “non-caring” sectors (all other industries). The existing literature suggests an association between pro-social behaviour and caring services, such as health, education and social care (see Francois, 2003). Since these services are more likely to be delivered by the non-profit sector, it is important to control for service-type in comparing across sectors.
A related literature has tried to capture donated labour indirectly by testing for a wage gap between the private sector and the (narrowly-defined) not-for-profit sector in the US (see inter alia Preston, 1989, Leete, 2001, Ruhm and Borkowski, 2003, Mokan and Tekin, 2004). The findings of these studies are mixed. However, ex ante, individuals in the not-for-profit sector may be paid more than in the private sector because of what Feldstein (1971) termed “philanthropic wage-setting” – the absence of pressure on managers to minimise costs, as well as differences in the tax and regulatory burdens across sectors. Also, in practice, few studies are able to control for all other compensating differentials between the sectors. Looking at the public-private wage gap in the UK, Postel-Vinay and Turon (2007) argue that differences in remuneration between sectors may not be fully captured by current pay because of differential risk of job loss.
An obvious concern with using unpaid overtime is that it may not be donated labour. For example, many individuals do unpaid overtime because it will improve their promotion prospects and result in higher remuneration in the future. We therefore include a number of controls for such career concerns together with a wide range of controls for individual and job characteristics. Another potential concern is if unpaid overtime is part of an implicit contract over hours. For example, there may be a social norm governing how much nominally unpaid overtime is in fact expected of everyone in the job, possibly compensating for shorter basic hours. Another possibility is that unpaid overtime may be a gift exchange in return for other benefits.
To rule out the possibility that unpaid overtime hours are part of an implicit contract, we look at what happens when individuals switch sectors. If the non-profit premium simply reflected differing social norms across sectors, we would expect to see individuals changing behaviour when they switched sector. We therefore estimate a fixed effects regression where the usual error term is decomposed into a constant individual specific effect and a pure random error term: . In the fixed effects specification, the sector effects are identified only from individuals who change sector. As shown in section 5, we find no evidence that individuals change their behaviour when they switch sector. This finding also rules out any institutional effects operating on individuals’ behaviour. This is a strong, and perhaps surprising, result; we show that it is very unlikely to be attributable to measurement error.
This suggests that the estimated non-profit premium reflects the selection of individuals into different sectors on the basis of their pro-social motivation. Put simply, “caring” individuals appear to select themselves into the non-profit sector and “non-caring” individuals into the for-profit sector. Formally, the selection story is that . In section 6, we present additional evidence that supports this selection story. We show that individuals who switch from the non-profit caring sector to the for-profit caring sector are less likely to do unpaid overtime (when they are in the non-profit sector) than those who stay in the non-profit caring sector. We also find that individuals who switch from the for-profit caring sector to the non-profit caring sector are more likely to do unpaid overtime when they are in the for-profit sector than those who stay in the for-profit sector.
3. Data
The data we use are taken from the British Household Panel Survey (BHPS). Since 1991 this survey has annually interviewed members of a representative sample of around 5,500 households, covering more than 10,000 individuals. On-going representativeness of the non-immigrant population is maintained by using a “following rule” – i.e. by following original sample members (adult and children members of households interviewed in the first wave) if they move out of the household or if their original household breaks up.[4]
A key advantage of using the BHPS is that as a panel it allows us to observe the same people working in both the for-profit and non-profit sectors. It also collects a wide range of detailed demographic and employment information. A potentially limiting factor is that the sample sizes in each wave of the BHPS are not sufficiently large to allow us to estimate standard deviations of wages by occupation with any precision. We use these to control for career concerns as discussed further below. We therefore supplement our analysis with data from the Labour Force Survey, a quarterly sample of 60,000 individuals. This limits our analysis to the period 1993 – 2000 for which we have common information across both datasets.
We select a sub-sample of individuals aged 16 – 60 who work between 30 hours and 90 hours per week. We exclude the self-employed and individuals in industries with non-standard working practices such as the armed forces, forestry and agriculture. We drop observations with missing information in key variables and also trim the top and bottom 0.5 per cent of the distributions of key variables such as hours of overtime (paid and unpaid), usual job hours and hourly pay.[5] Our final sample contains 6,061 individuals (24,135 person observations).
The survey does not directly ask individuals how many hours unpaid overtime they work. Instead, they are asked the following three questions about their hours of work:
Thinking about your (main) job, how many hours excluding overtime and meal breaks are you expected to work in a normal week?
And how many hours overtime do you usually work in a normal week?
How much of that overtime (usually worked) is usually paid overtime?
The answer to the first question is assumed to reflect an individual’s basic, contracted hours. The second two questions are used to derive the number of hours of unpaid overtime. Although calculated as a residual, estimates of unpaid overtime using the BHPS compare well to those obtained using the LFS where individuals are asked directly how much unpaid overtime they do.[6]
The main focus of our analysis is a comparison of unpaid overtime worked by individuals in different sectors (for-profit and non-profit). We define individuals’ sector on the basis of the following question:
Which of the types of organisations on this card do you work for (in your main job)?
Individuals are prompted with a list of options. Those who respond “private firm/ company” are allocated to the for-profit sector. All other responses are allocated to the non-profit sector. These include “civil servant/central government”, “local government/town hall”, “NHS or higher education”, “nationalised industry”, “non-profit organisation”.[7]Our non-profit sector therefore includes individuals working in the public sector, as well as in (traditionally defined) non-governmental not-for-profit organisations.[8]
A potential problem with this self-reported measure is that it may be subject to non-random measurement error. Estimates of the public sector workforce based on a self-reported measure in the LFS have been shown to overestimate the size of the public sector workforce. However, this bias has been shown to be mainly attributable to (self-employed) general medical practitioners wrongly classifying themselves as public sector and to staff in higher education classifying themselves as public sector, as opposed to the not-for-profit sector.[9] Since we drop the self-employed from our sample and since we are interested in the distinction between the for-profit and (widely-defined) non-profit rather than between the public and not-for-profit sectors, we would argue that these measurement error issues do not pose a problem for our analysis.
Our analysis of donated labour focuses on individuals working in caring industries since this is where we would expect individuals’ motivation to be manifested in extra donated labour. There is no formal definition of caring industries. To avoid imposing our own, possibly arbitrary, definition we follow Francois (2003) in identifying caring industries as those with “…a public good component. Examples of such services are childcare, medical care, education, and care for the aged”. We therefore define individuals working in health, education and social care industries as being in caring industries using the 1980 Standard Industrial Classification (SIC) two digit codes. Individuals working in these industries comprise 17 per cent of our total sample.
It could be argued that an industry-wide definition of caring is too broad; for example a hospital cleaner may not donate their labour because they work in a hospital rather than in an office, whereas hospital doctors may posses a greater level of attachment to the service they provide. For this reason, we also used a more restricted definition that cross-classifies industry with job occupation and defines caring occupations within caring industries, to include managers, natural scientists, health and teaching professionals and childcare workers. This definition restricts individuals working in caring to 14 per cent of our sample. A third possible definition of caring includes research and development, the arts and culture, corresponding to a broader set of industries where not-for-profit organisations are concentrated according to Rose-Ackerman (1996). This broadens the group of caring individuals to 20 per cent of our sample. We have assessed that our main conclusions are not sensitive to the definition of caring that we use and in the rest of the analysis presented below we focus on the first definition.