Abstract
Reputation systems aim to induce honest behavior in online trade by providing information about past conduct of users. Online reputation, however, is not directly connected to a person, but only to the virtual identity of that person. Users can therefore shed a negative reputation by creating a new account. We study the effects of such identity changes on the efficiency of reputation systems. We compare two markets in which we exogenously vary whether sellers can erase their rating profile and start over as new sellers. Buyer trust and seller trustworthiness decrease significantly when sellers can erase their ratings. With identity changes, trust is particularly low towards new sellers since buyers cannot discriminate between truly new sellers and opportunistic sellers who changed their identity. Nevertheless, we observe positive returns on buyer investment under the reputation system with identity changes, and our evidence suggests that trustworthiness is higher than in the complete absence of a reputation system.
Similar content being viewed by others
Notes
See, for example, the reports prepared by the National Consumer League (www.fraud.org) or the Internet Crime Complaint Center (www.ic3.gov).
Two baseline paradigms have been used to study reputation systems. Keser (2003) and Masclet and Peenard (2012) who also use the standard trust game, while Bolton et al. (2004) use a binary variant (“buyer–seller game”), in which efficiency gains only arise when trust is honored. We use the standard trust game for our main analysis because it allows studying certain types of strategic seller behavior. For example, we can study whether sellers build up their reputation when it is relatively cheap to do so and then “milk” this reputation later on. However, we also test the robustness of our results using the buyer–seller game (see Sect. 3.2.2).
We thus abstract from the public good problem of voluntary feedback provision. Gazzale and Khopkar (2011) study a setup without identity changes, in which buyers can decide whether they want to leave feedback or not. Since there is no scope for opportunistic buyer behavior in our setup, sellers cannot rate buyers in our experiment.
We chose this timing of entry to minimize the number of repeat encounters. In the post-experimental questionnaire, only 2 out of 192 subjects indicated that they thought they had recognized someone whom they had played with before.
A translated version of the instructions can be found in the “Appendix”.
The behavioral reputation literature starting with Camerer and Weigelt (1988) tests the predictions of this type of reputation models by exogenously introducing uncertainty about sellers’ preferences via the experimental design. The results of this literature also suggest that other-regarding preferences or an intrinsic concern for appropriate behavior play an important role in understanding reputation (Grosskopf and Sarin 2010).
In total 98 rounds were played in each session (3 players in each role played 20 rounds, 1 player each played 17, 13, or 8 rounds respectively). For each role there are thus \((60+17+13+8)\times 8=784\) rounds per treatment.
Note that, in principle, lower average trustworthiness in the change treatment could simply be driven by lower buyer investments if sellers reciprocate lower investments with lower returns on investment. Two analyses indicate that the treatment difference in our setup is not just due to the different distribution of investments, but persists conditional on investment. First, we calculate a counterfactual average return on investment by using the investment distribution of the change treatment and the returns on investment observed for each investment level from 1 to 10 in the no-change treatment. The counterfactual return on investment is 56 % and thus substantially higher than the 36 % actually observed. Second, we regress return on investment in the two reputation treatments on a treatment dummy controlling for investment in a random-effects estimation with standard errors adjusted for clustering at session level. While the coefficient on investment is significant and positive, the treatment dummy remains economically and statistically significant in this regression. Detailed results of both analyses are available upon request.
While some previous studies observe a negative or zero return on investment in trust games in a stranger environment (e.g., Keser 2003), the results of our control treatment are in line with those of a recent meta-study of 162 trust games regarding investment and return on investment (Johnson and Mislin 2011), which finds an average investment of 5.00 and a small, but significant positive return on investment of 0.11.
Since buyers did not make a decision for every potential return on investment only intervals can be determined for the \(b_{neg}\) and \(b_{neu}\), which explain the highest percentage of choices. All \(b_{neg}\epsilon [0.58;0.6]\) and all \(b_{neu}\epsilon [0.88;0.88]\) are optimal in the no-change treatment (change: \(b_{neg}\epsilon [0.17;0.2]\); \(b_{neg}\epsilon [0.63;0.66]\)). The values reported represent the mid points of these intervals.
Players without reputation include those players who entered as a new seller in the previous round but did not receive any investment. If we only consider players with the label “new” the difference in investment is even larger (see below).
For the analysis of identity change behavior we exclude the first round of each seller since identity could not be changed in this round. 736 rounds are thus included in the analysis.
The conditional probability of changing identity is also highest for an aggregate reputation of −1. 86 % of the sellers with an aggregate reputation of −1 change their identity. Note also that the substantial positive impact of a positive last rating on investment found above could make a strategy profitable where sellers split equally in one period to receive a positive rating, behave opportunistically in the next, and change their identity after having received a negative rating. This could explain why 25 % of sellers change their identity at an aggregate reputation of zero.
A qualitatively similar picture emerges if we consider aggregate reputation at the end of each period instead of return on investment.
The correlation is also significant if we take each session as an independent observation (Spearman rank order correlation \(\rho =-0.90\); p < 0.01; N = 8).
Given that all players were new players at the start of the experiment and the label could thus not carry any information we exclude the first period from the analyses in this paragraph. The qualitative results are robust to including these observations.
One reason why buyers might not treat new sellers in the worst possible way could be that they are guided by “preferences for appropriate choices” (Grosskopf and Sarin 2010).
This result also holds for the control treatment where buyers earn 11.07 and sellers earn 19.25.
There are 20 (23) bad and 27 (24) good sellers in the no-change (change) treatment. For our analysis, we ignore the remaining seller in each treatment who ends the session with an equal number of good and bad ratings. Including them either with the good or the bad sellers does not change the results.
Note that in the change treatment a classification into “bad” and “good” based on identity changes is possible. Our result also holds if we classify someone who changes identity at least once as “bad”, or if we classify someone who returns nothing and changes identity afterwards as “bad”.
In view of this result one might ask why not more subjects choose to behave opportunistically and change their identity afterwards. Analyzing behavior in the last round suggests that a substantial fraction of sellers does not act purely out of strategic concerns. 11 out of the 34 sellers who receive an investment in the last period provide a return on investment for the buyer of 50 % or more, and thus seem to have an intrinsic preference for trustworthy behavior.
Friedman and Resnick (2001) draw the same conclusion from their model and provide a discussion of different measures to achieve this.
References
Bajari, P., & Hortaçsu, A. (2004). Economic insights from internet auctions. Journal of Economic Literature, 42(2), 457–486.
Berg, J. E., Dickhaut, J., & McCabe, K. (1995). Trust, reciprocity, and social history. Games and Economic Behavior, 10, 122–142.
Bohnet, I., Harmgart, H., Huck, S., & Tyran, J. R. (2005). Learning trust. Journal of the European Economic Association, 3(2–3), 322–329.
Bolton, G. E., Katok, E., & Ockenfels, A. (2004). How effective are online reputation mechanisms? An experimental investigation. Management Science, 50(11), 1587–1602.
Bolton, G. E., Loebbecke, C., & Ockenfels, A. (2008). Does competition promote trust and trustworthiness in online trading? An experimental study. Journal of Management Information Systems, 25(2), 145–170.
Bolton, G. E., Greiner, B., & Ockenfels, A. (2013). Engineering trust: Reciprocity in the production of reputation information. Management Science, 59(2), 265–285.
Brown, M., Falk, A., & Fehr, E. (2004). Relational contracts and the nature of market interactions. Econometrica, 72(3), 747–780.
Camerer, C., & Weigelt, K. (1988). Experimental tests of a sequential equilibrium reputation model. Econometrica, 56(1), 1–36.
Cassar, A., & Rigdon, M. (2011). Trust and trustworthiness in networked exchange. Games and Economic Behavior, 71(2), 282–303.
Dellarocas, C., & Wood, C. A. (2003). The digitization of word of mouth: Promise and challenges of online feedback mechanisms. Management Science, 49(10), 1407–1424.
Dellarocas, C., & Wood, C. A. (2008). The sound of silence in online feedback: Estimating trading risks in the presence of reporting bias. Management Science, 54(3), 460–476.
European Commission. (2004). Issues relating to business and consumer e-commerce. Special Eurobarometer 60.0/Wave 201, European Opinion Research Group, Brussels.
European Commission (2012) Cyber Security. Special Eurobarometer 390/Wave EB 77.2, TNS Opinion & Social, Brussels
Fischbacher, U. (2007). z-tree: Zurich toolbox for ready-made economic experiments. Experimental Economics, 10, 171–178.
Friedman, E. J., & Resnick, P. (2001). The social cost of cheap pseudonyms. Journal of Economics & Management Strategy, 10(2), 173–199.
Gazzale, R. S., & Khopkar, T. (2011). Remain silent and ye shall suffer: Seller exploitation of reticient buyers in an experimental reputation system. Experimental Economics, 14, 273–285.
Greif, A. (1989). Reputation and coalitions in medieval trade: Evidence on the maghribi traders. Journal of Economic History, 49(4), 857–882.
Grosskopf, B., & Sarin, R. (2010). Is reputation good or bad? An experiment. American Economic Review, 100, 2187–2204.
Jin, G. Z., & Kato, A. (2006). Price, quality, and reputation: Evidence from an online field experiment. The RAND Journal of Economics, 37(4), 983–1005.
Johnson, N. D., & Mislin, A. A. (2011). Trust games: A meta-analysis. Journal of Economic Psychology, 32, 865–889.
Kandori, M. (1992). Social norms and community enforcement. Review of Economic Studies, 59(1), 63–80.
Keck, S., & Karelaia, N. (2011). Does competition foster trust? The role of tournament incentives. Experimental Economics, 14, 273–285.
Keser, C. (2003). Experimental games for the design of reputation management systems. IBM Systems Journal, 43, 498–506.
Klein, T. J., Lambertz, C., Spagnolo, G., & Stahl, K. O. (2009). The actual structure of ebay’s feedback mechanism and early evidence on the effects of recent changes. International Journal of Electronic Business, 7(3), 301–320.
Kreps, D., Milgrom, P., Roberts, J., & Wilson, R. (1982). Rational cooperation in the finitely repeated prisoners’ dilemma. Journal of Economic Theory, 27(2), 245–52.
Liu, Q. (2011). Information acquisition and reputation dynamics. Review of Economic Studies, 78, 1400–1425.
Masclet, D., & Peenard, T. (2012). Do reputation feedback systems really improve trust among anonymous traders? An experimental study. Applied Economics, 44, 4553–4573.
Milgrom, P. R., North, D. C., & Weingast, B. R. (1990). The role of institutions in the revival of trade: The law merchant, private judges, and the champagne fairs. Economics & Politics, 2(1), 1–23.
Ockenfels, A. (2003). Reputationsmechanismen auf internet-marktplattformen: Theorie und empirie. Zeitschrift für Betriebswirtschaft, 73(3), 295–315.
Okuno-Fujiwara, M., & Postlewaite, A. (1995). Social norms and random matching games. Games and Economic Behavior, 9(1), 79–109.
Reichling, F. (2004). Effects of reputation mechanisms on fraud prevention in ebay auctions. Stanford University Working Paper.
Resnick, P., & Zeckhauser, R. (2002). Trust among strangers in internet transactions: Empirical analysis of ebay’s reputation system. The Economics of the Internet and E-Commerce, 11, 127–157.
Resnick, P., Zeckhauser, R., Swanson, J., & Lockwood, K. (2006). The value of reputation on eBay: A controlled experiment. Experimental Economics, 9(2), 79–101.
Acknowledgments
The author would like to thank Johannes Abeler, Steffen Altmann, Christine Harbring and Armin Falk for many insightful discussions and Markus Antony, Holger Gerhardt, Alexander Koch, Sebastian Kube, Rosemarie Nagel, Axel Ockenfels, Gert Pönitzsch, Mirko Seithe, Dirk Sliwka, Florian Zimmermann and seminar and conference participants at Alicante, Amsterdam, Bonn, Cologne, Graz, Milan and Zurich for helpful comments. Financial support from the German Research Foundation (Grant KR 2077/2-1 and SFB/TR 15) is gratefully acknowledged.
Author information
Authors and Affiliations
Corresponding author
Electronic supplementary material
Below is the link to the electronic supplementary material.
Rights and permissions
About this article
Cite this article
Wibral, M. Identity changes and the efficiency of reputation systems. Exp Econ 18, 408–431 (2015). https://doi.org/10.1007/s10683-014-9410-3
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10683-014-9410-3