Post on 24-Jan-2016
description
Voluntary Disclosure of Firms as a Function of Industry Correlation: An
Experimental StudyGabriel D. Rosenberg
Motivation
• U.S. securities markets are based mainly on mandatory disclosure.
• Mandatory disclosure is expensive – will voluntary disclosure work just as well?– Are there different circumstances under which we
need mandatory vs. voluntary disclosure?
Different Industries
• Firms are not all the same. Firms in the same industry may have a common component to their value – correlation between firms in an industry.– “Disclosures by one firm in an industry may alter
investors’ beliefs about the profitability of other firms in the same industry, and thereby change their market value.” (Dye, citing Foster)
Question
• Do firms’ voluntary disclosure choices change as the correlation between firm values change?
Hypotheses
• Public goods hypothesis: – “Voluntary disclosure will necessarily be incomplete, will not
be as informative as it potentially could be, and might be very wasteful. Disclosure involves information, which is a free good and is difficult for those who produce it to capture the full gain from the cost of disclosure (public good). Thus, there is underproduction of information. There is a free-rider effect for similar companies.” [paraphrasing Judge Ralph Winter, Yale Law School class on Securities Regulation]
• Alternatively, disclosure decision might just be based on value.
Experimental Method
• Common Weighting % randomly chosen• Value = (Common Weighting %)*(Common
Component) + (100–Common Weighting %)*(Individual Component)
• Firms decide whether to disclose (cost of 10)• Investors bid on firms
Total DisclosuresCommon
Weighting %Total Number of
Disclosures
Round 1 16 1
Round 2 12 1
Round 3 74 1
Round 4 27 3
Round 5 47 2
Round 6 22 0
Round 7 16 2
Round 8 62 2
Round 9 32 2
Round 10 12 1
Total DisclosuresCommon
Weighting %Total Number of
Disclosures
12 1
12 1
16 1
16 2
22 0
27 3
32 2
47 2
62 2
74 1
Total Disclosures
Disclosure as a Function of Value
Disclosure as a Function of CommonValue
Disclosure as a Function of Independent Value
Logit Model
• Used to predict a binary event
Pr(DisclosureChoice = 1|Var1, Var2, Var3 …)
= f(β0 + β1Var1 + β2Var2 + β3Var3 …)
Logit Model: Disclosure Choice as a Function of Value
DisChoice Coef. Std. Err. Z P>z[95% Conf.
Interval]
Value .0876932 .0283026 3.10 0.002 .0322211 .1431652
_cons -5.807228 1.849376 -3.14 0.002 -9.431939 -2.182517
Logit Model: Disclosure Choice as a Function of Value
DisChoice Coef. Std. Err. Z P>z[95% Conf.
Interval]
Value .0876932 .0283026 3.10 0.002 .0322211 .1431652
_cons -5.807228 1.849376 -3.14 0.002 -9.431939 -2.182517
Logit Model: Disclosure Choice as a Function of Correlation,
CommonValue, and IndependentValue
DisChoice Coef. Std. Err. z P>z [95% Conf. Interval]
Correlation -.0147318 .0276366 -0.53 0.594 -.0688985 .0394349
Common Value
1.516459 2.45782 0.62 0.537 -3.300778 6.333697
Independent Value
8.519157 2.497779 3.41 0.001 3.6236 13.41471
_cons -5.965454 2.1691 -2.75 0.006 -10.21681 -1.714097
Logit Model: Disclosure Choice as a Function of Correlation,
CommonValue, and IndependentValue
DisChoice Coef. Std. Err. z P>z [95% Conf. Interval]
Correlation -.0147318 .0276366 -0.53 0.594 -.0688985 .0394349
Common Value
1.516459 2.45782 0.62 0.537 -3.300778 6.333697
Independent Value
8.519157 2.497779 3.41 0.001 3.6236 13.41471
_cons -5.965454 2.1691 -2.75 0.006 -10.21681 -1.714097
DisChoice Coef. Std. Err. z P>z [95% Conf. Interval]
Correlation -.0147318 .0276366 -0.53 0.594 -.0688985 .0394349
Common Value
1.516459 2.45782 0.62 0.537 -3.300778 6.333697
Independent Value
8.519157 2.497779 3.41 0.001 3.6236 13.41471
_cons -5.965454 2.1691 -2.75 0.006 -10.21681 -1.714097
Logit Model: Disclosure Choice as a Function of Correlation,
CommonValue, and IndependentValue
DisChoice Coef. Std. Err. z P>z [95% Conf. Interval]
Correlation -.0147318 .0276366 -0.53 0.594 -.0688985 .0394349
Common Value
1.516459 2.45782 0.62 0.537 -3.300778 6.333697
Independent Value
8.519157 2.497779 3.41 0.001 3.6236 13.41471
_cons -5.965454 2.1691 -2.75 0.006 -10.21681 -1.714097
Logit Model: Disclosure Choice as a Function of Correlation,
CommonValue, and IndependentValue
Logit Model: Disclosure Choice as a Function of the Components Value
DisclosureChoice
Coef. Std. Err. z P>z [95% Conf. Interval]
CommonTimesCorr
.0687366 .0309282 2.22 0.026 .0081184 .1293548
IndTimesWeighting
.1231108 .0380677 3.23 0.001 .0484994 .1977221
_cons -6.709012 2.110985 -3.18 0.001 -10.84647 -2.571558
Logit Model: Disclosure Choice as a Function of the Components Value
Disclosure Choice
Coef. Std. Err. z P>z [95% Conf. Interval]
Correlation -.0125845 .0269673 -0.47 0.641 -.0654395 .0402706Common
Value1.463747 2.504474 0.58 0.559 -3.444932 6.372426
IndependentValue
8.300579 2.616179 3.17 0.002 3.172963 13.42819
Previous Profit
-.0192508 .0335655 -0.57 0.566 -.085038 .0465364
_cons -5.925656 2.289073 -2.59 0.010 -10.41216 -1.439156
Logit Model: Disclosure Choice as a Function of the Components Value
Disclosure Choice
Coef. Std. Err. z P>z [95% Conf. Interval]
Correlation -.0125845 .0269673 -0.47 0.641 -.0654395 .0402706Common
Value1.463747 2.504474 0.58 0.559 -3.444932 6.372426
IndependentValue
8.300579 2.616179 3.17 0.002 3.172963 13.42819
Previous Profit
-.0192508 .0335655 -0.57 0.566 -.085038 .0465364
_cons -5.925656 2.289073 -2.59 0.010 -10.41216 -1.439156
Conclusion
• Firms seem to make decision based on value (mainly independent value) rather than correlation– No visible public goods problem
• In the future, would be better to pick certain correlation levels and randomize within those rather than completely random