Discussion Paper
No. 2007-21 | May 11, 2007
Annamaria Fiore and Andrea Morone
A Simple Note on Informational Cascades

Abstract

Seminal models of herd behaviour and informational cascades point out existence of negative information externalities, and propose to ‘destroy’ information in order to achieve social improvements. Although in the last years many features of herd behaviour and informational cascades have been studied, this particular aspect has never been extensively analysed. In this article we try to fill this gap, investigating both theoretically and experimentally whether and to which extent destroying information can improve welfare. Our empirical results show that this decisional mechanism actually leads to a behaviour more consistent with the theory that in turn produces the predicted efficiency gain.

JEL Classification:

C91, D62

Links

Cite As

[Please cite the corresponding journal article] Annamaria Fiore and Andrea Morone (2007). A Simple Note on Informational Cascades. Economics Discussion Papers, No 2007-21, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2007-21


Comments and Questions



anonymous - Referee Report
June 12, 2007 - 09:27
see attached file

Andrea Morone - Response to referee
June 25, 2007 - 18:13 | Author's Homepage
The referee has read the paper carefully, and we have found his or her comments extremely useful while working on a new draft of the paper with some substantial changes (addresing Problem 1 and 2). In all cases we have adopted the referee's recommendations. Problem 1: We have reformulated equations 2a, 2b, 3a, and 3b (see page 15 of the new version of the paper). The previous formulation was not wrong, but unacceptably unclear (we would like to thank the referee to have raised the point). Problem 2: We have re-written all the appendix in a clearer way (see pages 14-15 of the new version of the paper). Minor Point 1: typos have been fixed Minor Point 2: ·Sentence at page 4 (page 3 in the new version) has been changed in: Each individual has to decide whether or not to adopt a specific behaviour, for example, whether or not to adopt a new technology.·Sentence at page 6 (page 5 in the new version) has been changed in: (k, number of players observing only their own signal; p, probability of signal correctness; n, number of players having already taken their decision). ·Table 7 has been changed according the referee's suggestion. ·Sentence at page 17 has been changed in: Negative informational externality produced by phenomenon of informational cascade has drawn quite a lot of attention in economic literature. Minor Point 3: reference to Pastine and Pastine has been added.

Anonymous - Referee Report
July 24, 2007 - 10:46
see attached file

Andrea Morone - Response to referee
July 29, 2007 - 16:11 | Author's Homepage
see attached file

Anonymous - Associate editor's response
August 17, 2007 - 13:15
After a careful review of the reports, the responses and my own read of the paper, I do not feel comfortable accepting the paper as it now stands. However, in their response to the second referee, the authors suggest that they are willing to further revise their manuscript in light of of the comments received. I am inclined to recommend publication of a revised version that accounts for the comments as they address. However, in addition to completing these I have some additional concerns: 1 Despite the responsiveness towards referee 1, the Appendix can and should still be improved upon. Indeed, the analysis that it contains should be elaborated on and the specifics for the parameters chosen in the experiment should also be included as an illustrations of the calculations. 2 In Table 3 and Figure 1 the authors include a breakdown of the first four moves in treatment T2. However, these are theoretically all the same and there should not be a distinction. Why are they making one? Can this be elaborated on; or should these observations be grouped together? 3 As evidenced by the fact that there are large differences in payoffs dependent on the order of the first four moves in treatment 2 (see comment 2 above); the overall data set is too small to draw reliable conclusions at this stage. This is indeed what referee 2 points to in his/her first comment – something that the authors seem to have misunderstood (that is, 10 subjects is sufficient; but one needs more than 10 treatments!). 4 It is advisable that the instructions given in the experiment are made available to the referees and the public evaluators – after all space is not a concern here.

Andrea Morone - Responses to Associate editor’s comments (MS 55)
September 07, 2007 - 15:42 | Author's Homepage
We thank the editor for his/her useful suggestions. In the following you find our responses to each of the points raised, along with more details regarding our responses to second referee’s comments.Second referee’s commentsAs stated in the report to the second referee, we carefully followed his/her remarks in order to make our results clearer. In particular:· we modified accordingly table 2 and further clarified the meaning of figures (point 2);· we added two footnotes (7 and 8) in order to explain in more details subjects’ behaviour (point 3);· we substituted “ex –ante” with the term “theoretical” and we provided a closer explanation for it (point 4);· referee rightly notes that chance should play an important role on average earnings. We considered his/her point standardizing our results as showed in Appendix D (point 5);· in the new version, it is clearly acknowledged that winning percentages under two treatments become very similar (point 6);· given that treatment dummy T2 should have a different impact, depending on which stage are considered, we ran a Chow test. Result in new Appendix E. Editor’s comments1. In addition to changes made to Appendix, we further elaborated on it, providing as an example the way to get probability of NO-cascade, of ending up in a CORRECT-cascade or in a WRONG-cascade for the particular values of parameters chosen for the experiment. 2. We completely agree on the fact that the first four moves under T2 are theoretically indistinct (as explained in section 2). As a consequence, we amended table 3 and figure 1 averaging figures over the first four positions. 3. We realized that observations derived from 10 subjects could be not sufficient to obtain reliable conclusion. Consequently, we repeatedly subject behaviour over 20 periods (in addition to 2 trial periods). In Appendix C we provide raw data we based on for our further analysis. 4. Instructions are now available in new Appendix B.

Andrea Morone - Revision of Paper
October 02, 2007 - 11:39
see attached file

Thomas D Jeitschko - Associate editor's response
October 02, 2007 - 11:43
The authors present an original and interesting insight into the value of information in the formation of cascades. The analysis is thoughtfully and competently executed and the results are note-worthy. However, there is some concern as to the strength of these insights due to a small sample size. As one referee notes: “…the authors improved their paper a lot based on the comments they received. But, honestly, as long as the authors just have one independent observation per treatment, I do not really trust their results. In Appendix D, where they tried to make up for the differences in the treatments due to chance, the treatments differ with respect to experimental earnings just on a 10%-level. In my eyes, they should definitely run more sessions per treatment and maybe use the same signals in both treatments in order to make sure the possible differences truly reflect differences in behaviour.” Economics would like to publish the results of this research. However, at this time, I do share a bit of the concern voiced above. I strongly encourage and hope that the authors will consider running a few additional sessions in order to substantiate their interesting insights. This would then make an outstanding contribution to the experimental literature on cascades.

Andrea Morone - Response to the Associate Editor
October 07, 2007 - 14:19 | Author's Homepage
Dear Professor Jeitschko, We appreciate very much your useful comments and it is our firm intention to reinforce our results enlarging our sample size. However, we would be very grateful if you could provide us with further details to accomplish this task. As a first step, we would like to agree on how many sessions for treatment you could consider valuable. Moreover, we agree with the referee on the opportunity to use the same signal set for both treatment. At the moment we have T1 (based on one set of signals , S1) and T2 (based on a second set of signals, S2). We are thinking to run a repetition of T1 using the signals set S2 and to run a repetition of T2 using the signals set S1. Would this be sufficient? If not can you tell us how many repetition per treatment you would consider sufficient? We thank you for your attention. Best Regards, Andrea Morone and Annamaria Fiore

Thomas D Jeitschko - Associate Editor´s Reply
October 11, 2007 - 09:32
I'm happy that you are willing to obtain additional data to strengthen your paper! As for your particular questions, unlike the referee, I would prefer that you generate new signals to assure true randomness and independence across the treatments. However, I am hopeful that one more round of each type would suffice -- provided, of course, that this yields a substantiation of the previous results.

Andrea Morone - new version
December 20, 2007 - 15:04 | Author's Homepage
Dear Editor, Following your advise we have run two new sessions of the experiment (one for each treatment). This new data confirms our previous results. We are attaching the new version of the paper. Thanking you in advance for your consideration, Sincerely, Andrea Morone and Annamaria Fiore

Andrea Morone - new version
January 15, 2008 - 09:55 | Author's Homepage
Dear Thos, I would like to thank you once again for your comments. We have now addressed them in the attached version of the paper. Sincerely Andrea