DIVERGENCE OF OPINION AND LONG‐RUN PERFORMANCE OF PRIVATE PLACEMENTS: EVIDENCE FROM THE AUCTION MARKET

• Author: Zheyao Pan,Jianlei Han,Guangli Zhang
• DOI: http://doi.org/10.1111/jfir.12170
• Date: 01 July 2019
• Published date: 01 July 2019

DIVERGENCE OF OPINION AND LONG-RUN PERFORMANCE OF PRIVATE

PLACEMENTS: EVIDENCE FROM THE AUCTION MARKET

Jianlei Han and Zheyao Pan

Macquarie University, Australia

Guangli Zhang

Central University of Finance and Economics, China

Abstract

In this article, we propose and construct a direct measure of investors’divergence of opinion

based on auction bids data from private placements in China. We find that firms with higher

bids dispersion generate lower long-run stock returns after the issuance of private

placements. This effect is economically significant and robust when controlling for market

discount, earnings management, alternative dispersion measures, and self-selection bias.

Moreover, this negative relation is stronger for stocks with more stringent short-sale

constraints. Our findings therefore provide strong evidence for the overvaluation hypothesis.

JEL Classification: D44, G12, G14

I. Introduction

Miller (1977) hypothesizes that divergence of opinion can lead to asset overvaluation and

subsequent market underperformance, when pessimistic investors do not take adequate

short positions, for institutional or behavioral reasons.

1

In contrast, Williams’s (1977)

risk theory introduces heterogeneous beliefs into the capital asset pricing model (CAPM)

and predicts a positive relation between divergence of opinion and expected returns.

2

We are grateful to Erik Devos (editor) and an anonymous referee for their constructive comments. We also

thank Daoquan Chen, Albert Chun, Tao Gao, Xing Han, Priya Garg (discussant), Petko Kalev (discussant), Allan

Kleidon, Buhui Qiu, Byoung-Kyu Min (discussant), Krishnan Murugappa (discussant), Mohammed Sharaf

Shaiban (discussant), Tom Smith, Clara Zhou, Elizabeth Zhu, and seminar and conference participants at Xiamen

University, University of Queensland, University of Otago, University of Tasmania, Macquarie University, 2017

IFABS Asia Ningbo China Conference, 2017 FIRN Annual Conference, 2017 SIRCA Young Researchers

Workshop, 2017 Australasian Finance and Banking Conference, 2018 Financial Markets and Corporate

Governance Conference, 2018 Vietnam International Conference in Finance, and 2018 Financial Management

Association Annual Meeting for comments and suggestions. Han and Pan acknowledge funding from the

Accounting and Finance Association of Australia and New Zealand 2018 Research Grant. Zhang acknowledges

funding from the Project of Humanities and Social Science from MOE of China (18YJC790214). All remaining

errors are our own

1

As summarized by Hong and Stein (2007), three mechanisms drive divergence of opinion: gradual

information flow, limited attention, and heterogeneous priors.

2

Varian (1985) reaches the same conclusion that heterogeneous beliefs are a risk factor in the Arrow–Debreu

framework. More recently, Veronesi (2000) shows that higher information quality increases expected returns,

supporting Miller’s (1977) hypothesis; although consistent with Williams (1977), Epstein and Schneider (2008)

prove that there is an information ambiguity premium for stocks with low information quality.

The Journal of Financial Research Vol. 0, No. 0 Pages 1–32 2019

DOI: 10.1111/jfir.12170

1

© 2019 The Southern Finance Association and the Southwestern Finance Association

Vol. XLII, No. 2 Pages 271–302 Summer 2019

271

Using different measures of divergence, empirical studies have not generated convincing

evidence for or against Miller’s hypothesis in different settings. In this article, we shed

new light on this debate by proposing a novel measure of divergence of opinion based on

auction data from private placements. With this new measure, we document strong

evidence in support of Miller’s hypothesis.

One of the main reasons for the ambiguity am ong studies is that they use only

indirect measures for divergence of opini on (Garfinkel 2009), which might overlap

with risk factors and contain substan tial measurement errors. The most co mmonly

used measure is analyst forecas t dispersion. Consistent with Mill er’s (1977)

hypothesis, Diether, Malloy, and Scherbina (2002) find that st ocks with highly

dispersed analyst forecasts have lower future ret urns than stocks with less disperse d

analyst forecasts. However, there ar e shortcomings to using analyst forec ast dispersion

to measure investors’divergence of opinion. First, analyst dispersion is based on

forecasts of earnings rather tha n valuations. Second, investors ’decisions may not

follow analyst forecasts; thus, an alyst dispersion may not fully reflec t market

participants’divergence of op inion. Third, analyst dispersion i s contaminated by the

effect of uncertainty in analyst for ecasts (Barron et al. 1998; Sheng and Thev enot

2012). Fourth, analyst forecasts could be bias ed because of agency or behavioral

reasons, and investors often fail to co rrect for bias.

3

In line with these concerns,

Doukas, Kim, and Pantzalis (2006) docu ment a positive relation between divergen ce

of opinion and future returns after rem oving the effect of uncertainty in analyst

forecasts. Johnson (2004) shows tha t the findings in Diether, Malloy, and Scherbi na

(2002) can be explained by the effect of financial lever age. Other widely used

measures, such as idiosyncratic volat ility, turnover, unexplained trading vo lume, and

breadth of mutual fund ownership, are also ind irect proxies, which are potentially

contaminated by other return predict ors. The empirical evidence that relies on t hese

measures is mixed and inconclusive.

4

These drawbacks make it difficult to conclude whether evidence rejecting (or

supporting) Miller’s (1977) hypothesis is due to the theory itself or the proxy used. In this

article, we revisit Miller’s hypothesis and propose a direct measure of investors’opinions

regarding firm value based on their trading activities. Our measure is constructed using

auction data from private placements in China. There are two pricing schemes for private

placements in the Chinese market: a fixed price set by the board of directors and a market-

clearing price from the uniform sealed auction. The former is used when the issuance

targets are internal investors, that is, controlling shareholders and blockholders. The

latter is used when the issuance targets are external investors, that is, institutional

investors including mutual funds, trusts, private funds, asset management companies,

3

See Anderson, Ghysels, and Juergens (2005, p. 899) for a review of analyst dispersion biases. In addition,

Banerjee (2011) shows that the predictive ability of analyst dispersion is sensitive to the choice of scaling factor.

4

For example, Ang et al. (2006) and Guo and Qiu (2014) find a negative relation between idiosyncratic

volatility and returns. However, Bali and Cakici (2008) find no robust relation between idiosyncratic volatility and

stock returns, based on different tests. Fu (2009) even documents a positive relation between these two variables

when expected idiosyncratic volatility is used. Goetzmann and Massa (2005) and Garfinkel and Sokobin (2006)

reach different conclusions when trading volume is used. There is also a debate on whether turnover is a measure of

investor disagreement (Harris and Raviv 1993) or liquidity risk (Avramov and Chordia 2006).

2The Journal of Financial Research

272

and individual investors.

5

In the auction scheme, all bids information is publicly released

in private placement completion announcements. As suggested by Cammack (1991) and

Liu, Wei, and Liaw (2001), the divergence of auction bids directly reflects investors’

heterogeneous beliefs about valuations.

6

Therefore, this measure largely overcomes the

problems plaguing existing measures. Our measure is in line with the measure proposed

by Garfinkel (2009), who uses proprietary data from investors’limit and market orders in

individual stocks to directly measure their private valuations. However, in contrast to

Garfinkel’s measure, our auction-based measure is publicly available for a much longer

period.

7

To empirically measure investors’divergence of opinion, we manually collect

10,425 bid records from 411 private placement auctions, from 2007 to 2015. For each

auction, we construct the dispersion of bids by two measures: quantity-weighted standard

deviationand quantity-weightedabsolute distance.We then test Miller’s (1977)hypothesis

by investigating the relation between the divergence of bids and private placement

investors’holding-period returns.Our sample fits Miller’s theoretical assumptionsbecause

most Chinese firms face stringent short-sale constraints (Chang, Luo, and Ren 2014).

In this article, we report four main empirical findings. First, consistent with

Miller’s (1977) hypothesis, we find a negative and significant relation between the

divergence of bids and subsequent one-year stock returns for all four return measures:

two buy-and-hold abnormal return (BHAR) measures following Barber and Lyon (1997)

and Daniel et al. (1997) and two calendar-time abnormal return measures using the

CAPM and Fama and French (1993) three-factor model as benchmarks. This negative

relation is robust to controlling for the discount rate of issuing price to market price, scale

of private placement issuance, number of bidders, firm size, market-to-book ratio, cash

holding, return on assets (ROA), firm age, book leverage, earnings management, year

fixed effect, and industry fixed effect. In addition to statistical significance, we find that

the effect of bids dispersion on long-term returns is economically significant: a one-

standard-deviation increase in the bids dispersion—quantity-weighted standard

deviation (quantity-weighted absolute distance)—decreases the 11-month-ahead Barber

and Lyon matched BHAR, Daniel et al. matched BHAR, CAPM-adjusted return, and

Fama and French (1993) three-factor model adjusted return by 5.05% (4.98%), 3.40%

(3.58%), 4.11% (4.29%), and 3.28% (3.30%), respectively.

Second, we test whether the predictability of bids dispersion overlaps that of the

other divergence of opinion measures in the literature. We find that controlling for

analyst forecast dispersion, idiosyncratic volatility, and abnormal turnover does not

affect the significance of the bids dispersion, both qualitatively and quantitatively.

Hence, our measure of bids dispersion contains new information regarding future long-

run stock returns.

5

See Section II for more details on the institutional background of private placements in China.

6

Milgrom and Weber’s (1982a, 1982b) generalized auction model proves that bidders have incentives to

gather extra information to increase their profits in a sealed auction.

7

Garfinkel (2009) uses data from a much narrower sample period (January 2002–March 2002) and does not

study the relation between his divergence of opinion measure and future stock returns.

Divergence of Opinion 3

273