Public Goods: A Survey of Experimental Research
Summary (4 min read)
1.1 A Simple Public Good Experiment
- Both economists and sociol ogists recognize that the desired outcome is for all to contribute $5.
- The experimental evidence suggests that voluntary contributions will not produce that desired outcome.
- Economic theory suggests1 that it may be possible to change the institutions by which group choices are made in a way that ca. uses the outcome to be closer to the group op timum.
- To know how to do that, however, requires anticipating how individual choices will change as the institutions change.
2.1.1 Procedures
- The persons in each subgroup were placed into a room with two TV-sets and were, for allegedly "practical reasons," immediately given the fees promised them in four ten-Crown bills, one five-Crown bill and small change to make Kr.50.
- The administrator gave an oral presentation of the test which involved a half-hour program by Hasse Alfredsson and Tage Danielsson,32 not yet shown to the public.
- The subjects were given the impression that there were many groups of the same size simultaneously being asked the same questions in other rooms elsewhere in the broadcasting company.
- The subjects in group VI, who received instructions which differed from the instructions to the first five groups, were simply asked how much they found the program to be worth at a rnaxirnurn.
- In a second round, these people were asked to give their highest bids for a seat to watch the program and were told that the 10 highest bidders out of an alleged group of some 100 persons were to pay the amount they had bid and see the program.
2.1.3 Comments
- Bohm's imaginative study was, for its time, a major advance in the attempt to identify the extent of voluntary behavior in the presence of public goods.
- His experiment was seriously flawed in at least three ways.
- As a result, the data were not convincing and he was forced to conclude correctly that "the test would seem to encourage further work in the field of experimental economics.".
- The question of cooperative vs. selfish behavior remained open.
2.2 Dawes et al .: Soc ial Dilemmas
- While economists were struggling to get their experiments under control, social psychol ogists were independently studying a phenomenon which, I would argue, is a special case of public goods; social dilemmas.
- One of the best and most persistent groups has included Robyn Dawes and John Orbell.
- Let us look at Dawes, McTavish, and Shaklee (1977) for an example of this type of work that avoids many of the flaws of Bohm.
2.2.2 Results
- The last observation concerns the most obvious and least informative result: relevant talking matters a lot, but although four different types of communication were tried the data provide little information as to why.
- Just letting subjects talk in an uncontrolled framework opens up the chances for all sorts of contamination and unintended effects.
- If the authors are to understand the role of communication in encouraging voluntary contributions, they need better control and precision in their experimental designs.
- The state of the art is described in Section 3.3.3.
2.3 Marwell et al .: The Free Rider Problem
- In the process, they tested the distribution of resources, group size, heterogeneity of benefits, provision points, strength of induced preferences, experience of subjects, the divisibility of the public good, and the economics training of the subjects.
- This was a caref u lly thought out research program focused on an important phenomenon.
- In fact it was in response to this study that experimental economists finally began laying the groundwork needed to study free riding.
- No longer would provision of public goods be just a theoretical debate.
2.3.1 Procedures
- In small groups there were a total of 900 tokens and in large groups there were a total of 18,000 tokens.
- In some small groups one individual might have as many tokens as the provision point and everyone knew this.45 But as in the Bohm experiments some of this was a fiction.
- Group size was specified as "small" when there were four members in the group and "large" when there were 80 members.
- All groups contained just four real subjects.
- Because group members never interacted with one another it was 45Theory suggests in this case that one Nash equilibrium involves only that person contributing.
2.3.3 Comments
- A second observation echoes one I made in Section 2.1 on Bohm's research.
- Even though groups were actually of size 4, half of the subjects were told they were in a group of 80.
- It is believed by many undergraduates that psychologists are intentionally deceptive in most experiments.
- If undergraduates believe the same about economists, the authors have lost control.
- Honesty is a methodological public good and deception is equivalent to not contributing.
2.4.1 Pro cedures
- There were two standard rules regarding the information of participants: first, the subjects were not allowed to communicate with one another during the experiment.
- Secondly, the individuals had no knowledge about the nature of any payoff charts other than their own.
- In a technical sense it was public information that no one had information about other subject preferences.
- Furthermore, it was public information that the final period was known with certainty to no one.
2.4.2 Results
- On average first period contributions yield a public good level of 8.8 which yields a group payoff of 50% of the maximum possible.
- So the first decisions of sub jects are similar in both studies.
- By the fifth period the average number of units provided has dropped to 2.1 for a group payoff which is 9% of the maximum.
- So, after repetition, one can observe significant underprovision and the free-riding phenomenon.
2.5.1 Pro cedures
- Here each participant knew there would be exactly 10 periods and the participants' en dowments and payoffs would remain constant across the repetitions.
- But keeping all other possible effects constant proved more challenging.
- In particular, they discovered that keeping the marginal indi vidual payoff (a measure of selfish gain) constant and simultaneously keeping the marginal group payoff (a measure of altruistic gain) constant was impossible.
- Finally, experience is measured as previous participation in similar experimental ses s10ns.
2.5.2 Results
- So something more than just 40-60% contribution is going on.
- First, increasing M from .3 to .75 increases the rate of contribution in all cases.
- The effect is dramatic and strong and in the direction one should expect when the strength of the private incentive is reduced relative to the public incentive.
- Second, experience matters with inexperienced subjects contributing more.
- This suggests that some form of learning may be occurring.
2.5.3 Comments
- A second comment foreshadows the rest of this chapter.
- The fact that repetition and group size have a noticeable effect when M =.3 but not when M = .
- 75 signals a real difficulty with public goods experiments and their ability as economists to extract useful information from these experiments.
- To see why, let me try to summarize what the authors know to here.
2.5.4 A Summary to this Point
- The authors task would be easier if there were significant comparability across experiments and experimentors.
- There is precious little and, perhaps as a result, a lot of uncertainty still remains about behavior in public good environments.the authors.
3 .1 Thresholds and Provision Points
- In the Palfrey and Rosenthal (1991a) framework, pure strategy Bayesian equilibrium theory predicts a decrease from J( =1 to 2 and from J( =2 to 3 for their parameters.
- A careful look at mixed strategy equilibria for these environments with thresh olds suggests that game theory would predict that changes in the threshold can have an ambiguous effect on changes in contributions.
- The ambiguity is resolved only when specific parameters are known.
- The theory is telling us the authors should not expect a definitive answer to "does an increase in threshold increase contributions" which is independent of other factors.
3.3 Strong Effec ts
- There are really just two factors that fall into this category: one environmental, marginal payoffs, and one institutional, communication.
- I will, however, include a discussion of numbers and rebates since their effects are virtually impossible to disentangle from those of marginal payoffs.
3.3.l Marginal Payoffs and Rebates
- Two of the variables most easily controlled in public goods experiments are the marginal benefit of the public good relative to the private and the number of subjects in a group.
- Brown-Kruse and Hummels (1992) confirm the effect for MPCR=.5 and .3. Saijo and Yamaguchi (1992) confirm the effect for MPCR=.7 and 1.43.78 771 use marginal incentives here to contrast it with what h� been called the strength or salience of payoffs.
- These puzzling data will be discussed in more detail in Section 4 below.
- Indeed one other source of confirming data comes from the analysis of rebates in threshold situations.
- In a "no greed" condition subjects who do not contribute only get $5 more if at least J{ contribute.
2 p
- Where k is a subject pool specific constant.
- 89Since none predicts splitting of tokens, a well-known fact, all are technically defi cient.
- See Chen (1993) for a theory which might explain splitting.
3.3.3 Communication
- For completeness the authors should consider environments with thresholds.
- Here the evidence is mixed although the theory suggests there should be even more group gains from com munication than in the dilemma environment.
- Van de Kragt , Orbell, and Dawes (1983) report communication increases efficiency and contribution while Chamberlin (1978) and Palfrey and Rosenthal (1991b) report no discernible effect.
3.4.2 Systemic
- The authors are left with the undefined and unmeasured concepts of discussion induced group solidarity (Orbell, van de Kragt and Dawes (1988) ) and trust (Brown-Kruse and Hummels (1992) ) to explain part of the rate of contribution.
- There may be something here but it has not yet been isolated, measured and controlled.
3.4.3 Institutional
- The idea of revision is also not new since it can be found in one of the oldest market institutions, the English Auction.
- Dorsey (1992) , using the Isaac, Walker, and Thomas (1984) design, with MPCR=.3 and N=4, made one change and allowed subjects to adjust their planned contributions in real time.
- Only the final contribution levels were binding.
- He found 11.5% contribution rates when allowing both increases and decreases (compared to Isaac, Walker, and Thomas who found 26%).
- It is not obvious that revisions are helping in this public goods environment.
3.5 Unknown Effec ts
- It is obvious that subjects bring motivations beliefs and capabilities to the lab that may be vastly different from those assumed in standard game-theoretic models.
- Some experimental situations such as Double Oral Auctions appear to be very robust against such variations.
- That sensitivity can be controlled with high payoffs but little is learned.
- The hard problem is to isolate and measure the effects of the variations.
- This will keep experimentalists busy for a long time.
Did you find this useful? Give us your feedback
Citations
8,783 citations
6,919 citations
Cites background from "Public Goods: A Survey of Experimen..."
...Well known examples are that many people vote, pay their taxes honestly, participate in unions and protest movements, or work hard in teams even when the pecuniary incentives go in the opposite direction.1 This is also shown in laboratory experiments [Dawes and Thaler 1988, Ledyard 1995]....
[...]
...Working Paper No. 4 A Theory of Fairness, Competition and Cooperation Ernst Fehr and Klaus M. Schmidt April 1999 A THEORY OF FAIRNESS,...
[...]
5,391 citations
3,465 citations
3,231 citations
Related Papers (5)
Frequently Asked Questions (8)
Q2. What is the common way to eliminate or control for the two types of experience effects?
Repetition (not replication) has become a common feature72 of much research in experimental economics in an effort to eliminate or control for at least two types of experience effects: learning how to play the particular class of games, such as what keys to press in a computerized continuous auction or how to read a particular payoff schedule, and learning about the specific game one is in, such as what the environment is and what the other subjects are like.
Q3. What does Palfrey and Prisbrey (1993) suggest?
Palfrey and Prisbrey (1993) suggest that experience does not actually have a significant effect on the % of contributions, because, although experienced subjects contribute less , they also make fewer errors .
Q4. What is the marginal incentive to contribute?
In the Isaac and Walker environments ui = Pi (Zi - ci) + l{ (Ej Cj ) · The marginal (selfish) incentive to contribute is pfl = M PC R. The marginal (selfish) incentive not to contribute is p�N .
Q5. What can be done to test and identify the existence of effects which are otherwise unmeasurable?
comparative statics , and statistical procedures can allow us to test and identify, using indirect evidence, the existence of effects which are otherwise unmeasurable and, perhaps, uncontrollable.
Q6. What would have been the way to save money and avoid forcing subjects into losses?
An alternative way to save money and to avoid forcing subjects into losses would have been to add $9.50 to each entry (so all payoffs are non-negative) and then divide all entries by some number to lower the total paid out.
Q7. What does the t.he control of the experimenters suggest?
This suggests to me that, on average, about 10% of laboratory subjects may be simply immune to t.he control that experimenters try to exert by paying them.
Q8. What is the way to find out more about be havior?
Since both the economic/ game-theoretic and socio-psychologic theoretical predictions are wrong, the authors need to discover more about be havior not only in the context of voluntary contributions but also in the presence of many institutional designs.