Crowding in or crowding out? A laboratory
experiment on intrinsic motivation and extrinsic incentives
Zack Dorner∗1 and Emily Lancsar2 1Department of Economics, Monash University
2Centre for Health Economics, Monash University September 26, 2016
Abstract
This paper uses a laboratory experiment to investigate the extent to which intrinsic motivation can be crowded in or out by adding and then removing monetary or non-monetary incentives. The impact of size and type of incentive on motivation is tested between subjects. Furthermore, we investigate whether this effect is homogeneous or heterogeneous depending on baseline intrinsic mo- tivation to address a gap in the literature. The analysis includes survey data on participants’ pro-environmental and health behaviours, along with physically measured body mass index and waist size. The findings of the project may be useful for informing health and environmental policy.
∗Corresponding author: .
Lata Gangadharan in particular is acknowledged for the extensive supervision given throughout this project. A thank you also for the supervision provided by Anke Leroux and Paul Raschky.
Funding for this project is from the Australian Research Council (ARC), through a Discovery Early Career Researcher Award (DECRA) (Grant ID DE140101260). This project has been approved by the Monash University Human Research Ethics Committee; MUHREC#: CF16/618 - 2016000300.
Preface
Thesis title: Three Essays in Environmental Economics
Supervisors: Professor Lata Gangadharan, Associate Professor Paul Raschky and Dr Anke Leroux.
Within the broader field of environmental economics, the chapters in my thesis look at how preferences - in particular risk preferences, social preferences and intrinsic mo- tivation - drive decisions and behaviour. These preferences are identified using exper- imental methods, both in the field and the lab, with a key element of interest being heterogeneity. A further theme is technological change, in terms preferences over utilis- ing new technologies, and how technological change might influence pro-environmental behaviours through a behavioural rebound effect. Finally, as in the paper presented here, the thesis investigates the impact of incentives - monetary and non-monetary - on behaviour change. Along with an introduction and conclusion, the thesis will have three main chapters. They are as follows:
2. Preferences for Intrinsically Risky Attributes – with Daniel Brent and Anke Ler- oux (Monash Business School Department of Economics Discussion Paper 32/16).
In this chapter, we develop a novel approach to leverage data on risk preferences from a fully incentivized risk elicitation task to model intrinsic riskiness of alter- natives in a choice experiment. In a door-to-door survey, 981 respondents partici- pated in a discrete choice experiment (DCE) to elicit preferences over alternative sources of municipal water, conditional on water price and quality. Participants were not given information about supply or technological risks of the sources to avoid framing effects driving the results. Controlling for water quality and cost, we find that supply risk is an important determinant of participants’ choices, while respondents are not concerned about technology risk.
3. Crowding in or Crowding Out? A laboratory experiment on intrinsic motivation and extrinsic incentives – with Emily Lancsar (presented in this document).
4. Laboratory Experiment Investigating the Behavioural Rebound Effect.
The focus of the third chapter will be investigating the environmental rebound ef- fect in a laboratory setting, in order to better understand some of the behavioural drivers. Evidence around the rebound effect and its size to date is primarily from secondary field data, such as household energy usage and household production technology change, and most studies are not designed to identify behavioural drivers (Gillingham et al., 2016; Sorrell et al., 2009). The laboratory setting allows particular aspects of the rebound effect to be isolated; in this case the be- havioural impact of exogenous versus endogenous technological change and the importance of the baseline environmental impact of consumption.
2 Introduction
Evidence on the potential for extrinsic incentives to reduce intrinsic motivation moti- vation has been gathering since Deci’s seminal article in the field of psychology (Deci, 1971). The counter-intuitive notion that external incentives, especially monetary in- centives, could lead to the crowding out of effort is now widely accepted in economics too (Frey, 1997; Frey and Jegen, 2001; Gneezy et al., 2011). With crowding out, one or both of the following occurs: effort is reduced after the application of incentives or effort is reduced below pre-incentive levels after the removal of temporary incentives. Intrinsic motivation to do an activity can take many forms, such as the enjoyment of a particular activity, the desire to engage in productive and meaningful work, the benefits to one’s self image from undertaking an activity and prosocial motivation (Promberger and Marteau, 2013). Take exercise as an example. Generally individuals must find some way of motivating themselve to exercise, rather than rely on exernal incentives. One person may enjoy exercise, someone else might motivate themselves by striving to be a fit, healthy and attractive person. Another person may motivate themselves by wanting to stay healthy so that they can be around for their family as long as possible. Crowing out theory notes the possibility that paying someone to exercise could lead to this external incentive dominating the motivation for exercise and leading to an overall reduction in the level of exercise undertaken after payments start. If the incentive is large enough, the level of exercise may increase. However, when the incentive is removed, if crowding out has occurred then the level of exercise may drop below is original pre-incentive level. The possibility of crowding in also exists – after payment ceases, exercise may remain at a higher level than before.
As noted in this example, there are a range of types of intrinsic motivations someone might use to exercise. Naturally, there is great heterogeneity in the level of exercise undertaken by heterogeneous individuals. This fact leads to several questions. What are some of the factors that determine intrinsic motivation? And, how might external incentives affect a person with low intrinsic motivation compared with someone with a high level of intrinsic motivation?
In this paper we aim to address these questions. In a laboratory setting we give a real effort task to a heterogeneous, non-undergraduate student participant pool. We give the task to the participants for one round, with no incentives or mention of future incentives. In a second round, the participants are given the task again, with all but the control group given some sort of incentive (which vary by size and type). In the third round, the incentive is removed. A fourth round, again without incentives, is also given to participants, after a break in which another task is completed. Thus, we have a measure of baseline intrinsic motivation (round 1), test the effects of a range of incentives (round 2) and test the ongoing effect of removing those incentives (rounds 3
and 4). We also measure health and environmental variables related to the participants, which provide useful applications for the study results. We find the low power incentive is the only incentive effective at increasing effort in round 2, whereas the high power incentives only serve to crowd out effort in the subsequent rounds, particularly amongst those with the highest levels of original intrinsic motivation.
The crowding out effect can be identified when raising incentives decreases rather than increases supply of effort. Under the motivation crowding model, a small monetary incentive may lead to a decrease in supply, due to the crowding out effect dominating the positive effect of the incentive. An increase in supply can be garnered, provided there is a large enough external incentive provided. If this crowding out effect has operated through changed preferences, changed information (on the side of the agent about the principle’s motivations), lowering self agency or lowering the enjoyment of the task, the crowding out effect will persist after the incentive is removed (Frey and Jegen, 2001; Gneezy and Rustichini, 2000a,b). A large portion of the economic literature has focussed particularly on prosocial motivation and potential crowding out effects in this area (Promberger and Marteau, 2013). More recent theoretical models emphasise the roles of image motivation (both self- and public image) and core beliefs as assets to explain prosocial behaviours and cases of crowding out (B´enabou and Tirole, 2006, 2011). Conversely, the emphasis of intrinsic motivation literature in psychology has been motivation to do enjoyable activities – activities which are done for their own sake (Deci et al., 1999; Promberger and Marteau, 2013). This focus is not without contention. Cameron et al. (2001) argue that intrinsic motivation should be approached with a more broach definition. Given this broad definition, the literature shows that crowding out is not pervasive, but applicable only to high interest tasks with tangible rewards that are at least loosely performance-based.1
The laboratory setting has provided important insights into intrinsic motivation and crowding out. Indeed, of the three experiments conducted by (Deci, 1971), two of them are laboratory experiments. Particularly for the within-subject component of our experimental design, we build on the basic structure of Deci’s laboratory experi- ments. He measures baseline intrinsic motivation in the first round of a puzzle task, gives a monetary or non-monetary (verbal affirmation) incentive in a second round and then removes the incentive in the third round. There is also a control group with no incentive. The monetary incentive is found to crowd out intrinsic motivation (Deci’s experiment 1) and the verbal affirmation crowds in intrinsic motivation (Deci’s experi- ment 3). Ma et al. (2014) present a recent laboratory study with the same three round design showing that crowding out effect of a monetary incentive operates at a neuro- logical level. To the best of our knowledge, there is a gap in the literature in terms of testing a range of incentive types, which is covered in this paper by the between sub-
1For the response to Cameron et al. (2001), see Deci et al. (2001).
ject aspect of our experimental design. There is a further gap in terms of investigating heterogeneous responses to such a 3 round laboratory design. Finally, our study adds a fourth round to test persistence.
Our study has a generic laboratory design, using a real effort task without any further framing for participants. Given the importance of context within the intrin- sic motivation literature, we further collect data on two contexts for which intrinsic motivation is important. The health context is of interest given the private nature of benefits to exercise – that is, greater health, plus enjoyment from exercise for some individuals. Health has the added advantage of having a variable that is observable in a laboratory setting - namely weight. The environmental context is a useful example of a context where intrinsic motivation is of a pro-social nature. The advantage of the environmental context for this study is that pro-environmental preferences are perhaps more heterogeneous than other contexts; mitigating climate change is a much more controversial pursuit than feeding people who lack sufficient access to food.
This paper is organised as follows. The second section below outlines the methodol- ogy, including research questions, a description of the experiment and how the sample was recruited. The results are presented in Section 3, while Section 4 provides a dis- cussion and conclusion.
3 Method
Given the general questions and strands of literature identified in the introduction, this paper specifically seeks to address the following main research questions: First, what factors help explain initial level of intrinsic motivation? Second, how does level and type of incentive impact effort? Third, to what extent does level and type of extrinsic incentive crowd in or crowd out intrinsic motivation? Fourth, how does heterogeneity in level of intrinsic motivation impact the efficacy of incentives and the level of crowding in or out of incentives? Fifth, is level of intrinsic motivation and crowding in or out observed in a general lab context consistent with self reported attitudes and behaviours in the field; in particular health and environmental attitudes and healthy and pro- environmental behaviours? Our method is to utilise a generic laboratory setting, which is described in this section.
3.1 Experimental design
The experiment was run over 12 sessions from 6 April to 3 June, 2016, at the Monash Laboratory for Experimental Economics (MonLEE) at Monash University in Mel- bourne, Australia. The overall timeline of each experimental session is shown in Table 1.
Table 1: Overall timeline of each experimental session.
Initialisation / Activities / Surveys / Measurementand payment
Participants ran- / Activities for the / Surveys on health, / Participants in-
domly assigned to / experiment com- / the experimental / structed to proceed
computers, con- / pleted – multiple / activities and the / to a neighbour-
sent forms signed, / rounds of effort / environment given / ing room to be
overview instruc- / task and a time / to participants. / measured and
tions provided in / preferences task / paid in private by
hard copy and / (see Table 2 for / assistants.
read aloud by / more detail).
experimenter.
At the start of each session, each participant took a random number from a bucket, which corresponded to one of the 26 computers in the room. They were seated, signed consent forms and then overview instructions were provided in hard copy and read aloud. The instructions outlined the overall session structure, without giving detail about the activities themselves. At this stage it was explained to participants that they would be paid at the end of the activity by an administrative assistant in a neigh- bouring room. Next, the activities were undertaken, followed by surveys on health, the experimental activities and the environment. These tasks were all undertaken on the computers. When the participant was finished these activities, they were asked to line up outside the neighbouring room where they would be weighed, have their height and waist measured, and be paid.
The activities section of the experiment proceeded as shown in Table 2. We em- ployed multiple rounds of a real effort task in order to address the research questions. We used the word encoding real effort task developed by Erkal et al. (2011), pro- grammed using zTree (Fischbacher, 2007). The number pad on the right-hand side of the keyboard, along with the tab keys were disabled for all participants in all sessions to remove the advantage a particularly experienced computer user could have in the task.