Market efficiency, trading institutions and information mirages: evidence from a laboratory asset market

Similar phenomenons are herd behaviour (Banerjee 1992; Morone and Samanidou 2008; Morone 2012) and informational cascades (Bikhchandani et al. 1992; Fiore and Morone 2008; Morone et al. 2009).

Market 2 (DA treatment) was run over 11 trading periods.

Session 2 (DA treatment) and session 1 (CM treatment) were run with 7 and 8 participants respectively.

Session 2 (DA treatment) and session 1 (CM treatment) were run with 4 insiders, because of the lower number of participants with respect to the other sessions.

In line with Plott and Sunder (1982), we ensure that, in periods with inside information, at least half of the traders are fully informed about the realized state.

These particular cases are discussed in detail in Appendix C.

Reference to this branch of literature can be found in Milgrom and Stokey (1982) and Tirole (1982).

It remains open the question why, in these experimental markets, there is so much activity. One possibility is that hinted at by Hung and Plott (2001): subjects are bored and preferred to do something than to do nothing. Angrisani et al. (2011) tested the no-trade theorem in a laboratory financial market where subjects could trade an asset whose value was unknown. They found that market feedback matters: trading losses encouraged more conservative strategies, thereby reducing trade. Subjects lost money more frequently when there were no gains from trade, so feedback tended to drive these markets in the direction of no-trade. Overall, they concluded that their results “offer some vindication to both detractors and defenders of the no-trade theorems empirical validity. For the latter, the fact that trade is virtually wiped out in a setting where traders must learn a noisy information structure on the fly demonstrates that the no-trade logic is not as fragile as sometimes supposed. The former can counter that when the level of noise rises, informational trade is driven out much more gradually, and we cannot conclude that it will ever cease entirely”.

The high dividend price will be attained only if informed traders have enough cash to bid up the prices. If not enough liquidity reaches buyers’ hands and if sellers aggressively compete to sell out the asset, then insiders will be able to buy at prices even below the high dividend value (20).

Again, depending on market liquidity and on demand/supply conditions.

We recall there are 5 insiders out of 9subjects, with a capital and asset endowment of 200 ECU and 10 units of asset each.

Two limit cases (period 6 of session 2 and period 1 of session 3) occurred in the call market treatment. According with the previous discussion, one period (period 1 of session 3) has been considered as a mirage and the other one (period 6 of session 2) as a non-mirage case.

Out of the 12 mirages detected, 3 reflected the good state price and 9 reflected the bad state price.

Out of 9 mirages, 3 reflected the good state price (20) and 6 reflected the bad state price.

95% Confidence interval.

In each market, the frequency of information mirages is measured as the ratio between the number of mirages and the number of non-insider periods.

The econometric results hold for any specifications of Model (2). See regression output in Table 3, column (3).

Insider premium averages over periods at market level have been taken into account. Similar results have been achieved using MSE market medians.