IS STOCK PICKING DECLINING AROUND THE WORLD?

IS STOCK PICKING DECLINING AROUND THE WORLD?*




Utpal Bhattacharya
Kelley School of Business
Indiana University
ubhattac@indiana.edu


Neal Galpin
Kelley School of Business
Indiana University
ngalpin@indiana.edu





November 15, 2005




Key words: modern portfolio theory; indexing; stock picking
JEL number: G11, G14, G15



* We are grateful for suggestions from Craig Holden, Chris Lundblad, and seminar participants at Indiana University.

---

IS STOCK PICKING DECLINING AROUND THE WORLD?


Abstract
We do three things in this paper. We first develop a metric to measure the maximum fraction of volume
explained by stock picking in a market. We then use our metric to measure stock picking around the world. We
find that though there is more stock picking in emerging markets than in developed countries, it is declining
everywhere. In the United States, for example, stock picking has secularly declined from a high of 60% in the
1960s to a low of 24% in the 2000s. Finally, as markets cannot be efficient if everyone believes that they are
efficient and, therefore, do no stock picking – the Grossman and Stiglitz (1980) paradox – we ask what is the
long-run steady state fraction of stock pickers? We develop a simple theoretical model, and calibrate this model
to the United States economy to conclude that stock picking will eventually settle at 11% of trading volume in the
United States.

---

IS STOCK PICKING DECLINING AROUND THE WORLD?
I. INTRODUCTION
“A small gamble in a large number of different companies where I have no information to reach a good
judgment, as compared with a substantial stake in a company where one's information is adequate, strikes me as
a travesty of investment policy”
John Maynard Keynes, 1883-1946

John Maynard Keynes was a good stock picker. From 1928 to 1945, the fund he managed for King’s
College, Cambridge, produced positive returns at a time when the U.K. stock market was declining by 0.5% per
year.1 The intellectual foundations of stock picking were laid out in the classic text on valuation by Graham and
Dodd (1934), who showed us how to figure out whether a stock was a “buy.” Many of today’s famous investors
like Warren Buffett have been influenced by their theories. Indexing, which is the practice of passively investing
in a portfolio containing a large number of stocks, is the philosophical opposite of stock picking. Instead of
picking winners and losers, indexing emphasizes diversification. The intellectual foundations of indexing are in
Markowitz’s (1952) paper on modern portfolio theory and Tobin’s (1958) paper on two-fund separation.2
Indexing also has its fans in the investment world, of which perhaps the most influential is the Vanguard group of
mutual funds. Ironically, the index funds that Vanguard popularized as an asset class now face serious
competition from Exchange Traded Funds that exchanges have introduced to cash in on the popularity of passive
indexing.3

The purpose of this paper is to find out which investment philosophy, stock picking or indexing, is
dominant in the stock markets around the world. This is an important research question because, though there is
much anecdotal evidence that the ideas in Markowitz (1952), a paper which led to the “birth of modern financial
economics” (Rubinstein (2002)), have permeated investment practice, there has been, to the best of our

1 This data and the quote above come from Wikipedia, an online user-contributed encyclopedia
(http://www.answers.com/library/Wikipedia)
2 The practice of diversification existed before Markowitz (1952). See Goetzmann and Ukhov (2005) for an empirical study
of British overseas investments in 1870-1913.
3 Wall Street Journal, November 5-6, 2005.

- 1 -

---

knowledge, no paper formally measuring this permeation.4 Our paper makes a modest first attempt to formally
measure how the investment community has accepted one of the ideas of modern portfolio theory – indexing.

The prevalence of indexing can only be measured if there exists a measure for indexing. No such
measure exists in the literature. So the first part of our paper develops a metric for the opposite of indexing –
stock picking. The idea behind this measure is inspired by a theoretical insight in Lo and Wang (2000). They
proved that, if the two-fund separation theorem holds, dollar turnover of a stock, which is defined as the dollar
volume of shares traded divided by the dollar market capitalization of the stock, should be identical for all stocks.

An empirical implication of the above theoretical insight is that if every person in the world indexes
between a risk-free portfolio and the market portfolio (or a value-weighted portfolio that is a proxy for the market
portfolio), trading volume in stock i should be explained completely by the market capitalization of stock i. This
would mean that (1-R2) of the cross-sectional regression of the log of volume against the log of market
capitalization would reflect the deviation from the indexing investment philosophy. This deviation will occur
because some agents pick individual stocks. This deviation will also occur if some agents index to portfolios
other than the value-weighted portfolio. This means that the (1-R2) of the above cross-sectional regression
between log dollar volume and log dollar market capitalization will be a measure of the maximum proportion of
trade that can be explained by stock picking.

We run these cross-sectional regressions every month, for every country for which we have data, for as
long as we have the data. We plot the (1-R2) over time for 43 countries. We get two big results, and many small
results.

Our first big result is that, on an average, there is more stock picking in emerging markets than in
developed markets. As a matter of fact, the maximum fraction of volume explained by stock picking in emerging
markets in the 1995-2004 period is 63%, whereas the maximum fraction of volume explained by stock picking in
developed countries in the same period is only 45%. In our sample of 43 countries, the maximum fraction of

4 Rubinstein (2002), in his retrospective of Markowitz’s (1952) paper, states that “Markowitz’s approach is now
commonplace among institutional portfolio managers who use it both to structure their portfolios and measure their
performance. It has been generalized and refined in innumerable ways, and is even being used to manage the portfolios of
ordinary investors.” The website of Yahoo Finance (http://biz.yahoo.com/edu/bi/ir_bi5.ir.html), a popular site for individual
investors, states: “You can divide the history of investing in the United States into two periods: before and after 1952. That
was the year that an economics student at the University of Chicago named Harry Markowitz published his doctoral thesis.”

- 2 -

---

volume explained by stock picking in the 1995-2004 period is the least in the United States (29%) and is the most
in China (80%).

Our second big result is that, on an average, stock picking is declining around the world. Of the 43
countries under investigation, we record that for 38 countries, the maximum fraction of volume explained by
stock picking is lower in the last five years (2000-2004) than in the previous five years (1995-1999). The declines
in stock picking are quite dramatic, especially in the emerging markets. The most dramatic decline in the
popularity of stock picking is recorded in the United States, but that is probably because we have a longer time-
series data for the United States. In the United States, the maximum fraction of volume explained by stock
picking has secularly declined from a high of 60% in the 1960s to a low of 24% in the 2000s.
As we have a lot more data on the United States, we are able to get many small cross-sectional results for
the United States. We find that though stock picking is less in S&P 500 stocks than in non S&P stocks, the
difference seems to have disappeared in recent times. This fact shows that the though the actual mechanics of
indexing in S&P 500 stocks is easier, the mechanics do not matter much anymore because indexing seems to be
popular even in stocks not in the S&P 500 index. In terms of exchanges, there is more stock picking in AMEX
than in NYSE. Nasdaq starts out looking like the AMEX, but it looks like the NYSE today. In terms of size,
there is more stock picking in small stocks than in large stocks. In terms of age, there is more stock picking in
young firms than in old firms. In terms of industries, stock picking is highest in the telecommunication industries,
and lowest in the utilities industries. Our last cross-sectional result is that there is more stock picking in stocks
that are covered by fewer analysts than in stocks that are covered by more analysts. As analysts are the
quintessential stock pickers, it seems that investors who pick stocks avoid stocks that analysts pick. Finally,
whatever the above cross-sectional results, we record that stock picking is declining over time in each and every
category.
Our summary of findings from the first two parts of the paper is that stock picking is more pronounced in
emerging markets than in developed markets, but it is declining in nearly all stock markets of the world. Though
our paper is the first to formally document the declining popularity of stock picking around the world, indications
that this may be happening are in a paper by Fernando et al (2003), who document the explosive growth of mutual

- 3 -

---

funds around the world. Also, though our paper is the first paper to formally document the large cross-sectional
variation in the popularity of stock picking across countries, indications that this may occur are in a paper by
Khorana, Servaes and Tufano (2004), who document that the mutual fund industry is larger in countries with
better laws and regulations and more wealth. Interestingly, these are the countries with the lowest stock picking
in our sample. The above two arguments implicitly assume that mutual funds emphasize diversification over
stock picking. That may be a reasonable assumption, but as Wermers (2000) shows, stock picking is alive and
well in some United States mutual funds. Further, the explosive growth of hedge funds in recent years, also
means that stock picking may be making a comeback.
Our results show that modern portfolio theory has won. However, it is premature to write the epitaph for
stock picking. News of the death of stock picking will be an exaggeration. The reason no stock picking cannot be
an equilibrium strategy is because of the Grossman-Stiglitz (1980) paradox: if no one picks stocks, information
that stock pickers communicate with their trades cannot be impounded in prices, and so markets become
inefficient, and so develops an opportunity to gather information, pick stocks, and make trading profits. This begs
the question: how many stock pickers will exist in equilibrium? In other words, what is the long-run steady state
fraction of stock pickers?
We develop a simple theoretical model. Our theoretical model is based on a crucial insight that comes
from an early theoretical model by Treynor and Black (1973): in a mean-variance optimizing framework, even
active stock pickers would like to maximize their Sharpe ratio, which is the ratio of the risk premium of the
portfolio they choose divided by the standard deviation of the return of the portfolio (assumed to be the measure
of risk). This implies that if we restrict an investor to be either a passive investor in the market portfolio or an
active investor in a single stock, this would mean that this investor would be indifferent if the Sharpe ratios are the
same. The Sharpe ratio of the market portfolio is simply the market price of risk. The active investor, who we
allow to hedge the systematic risk of the single stock by taking an opposite position in the market portfolio, has
the following Sharpe ratio. It is his excess profit from his superior information divided by the cost of active
investing. The voluminous literature on market microstructure, which begins with Kyle (1985) and Glosten and
Milgrom (1985), tells us that the excess profit an insider obtains from his inside information is his profit based on

- 4 -

---

his superior information (Jensen’s alpha) minus his adverse selection cost and other transaction costs of trading.
We should also subtract his cost of obtaining the superior information from this number. Modern portfolio theory
tells us that the cost of active investing is the exposure to idiosyncratic risk. Equating the two Sharpe ratios
allows us to express the excess profit of the indifferent investor as a product of the market price of risk and the
idiosyncratic risk.
Note that the excess profit that makes an investor indifferent between stock picking and passive indexing
is the product of the market price of risk and the idiosyncratic risk. We call this value the “indifferent” excess
profit. This implies that stock picking would become less popular if the market price of risk is increasing and/or
idiosyncratic risk is increasing. The intuition is obvious. If the reward for holding market risk is increasing,
investing in the market portfolio is more attractive, and so stock picking is less popular. If idiosyncratic risk is
increasing, the cost of non-diversification, which is what stock picking entails, is increasing, and so stock picking
is less popular.
The market price of risk is time-varying. See Lettau and Ludvigson (2003) for an excellent survey of this
voluminous literature. There are many methods to estimate the market price of risk. We use the Whitelaw (1994,
1997) methodology as our primary method to estimate the market price of risk, though we report our estimates for
the other methods as well. Idiosyncratic risk is also time-varying. As a matter of fact, a growing literature has
documented that idiosyncratic risk has secularly increased over time in the United States (see Campbell et al.
(2001)) and all over the world (see Morck, Yeung and Yu (2000). We use the method of Morck, Yeung and Yu
(2000) to back out idiosyncratic risk for the United States. Our estimate of idiosyncratic risk tallies with their
estimate. It also tallies with the estimate of Campbell et al. (2001), who use a variance decomposition method of
estimating idiosyncratic risk. What is important, however, is that we get the same result: idiosyncratic risk is
increasing in the United States.
We then estimate the “indifferent” excess profit every year, which makes an investor indifferent between
stock picking and passive indexing, as a product of the estimated market price of risk every year and the estimated
idiosyncratic risk every year. We find that this “indifferent” excess profit is increasing over time. If we assume
that every agent in the economy has a different excess profit from stock picking, which depends on skill and/or

- 5 -

---

luck, it is reasonable to assume that this excess profit is drawn from a distribution that is stable over time. The
people who pick stock have a draw that is higher than the “indifferent” excess profit. So, if the “indifferent”
excess profit is increasing over time, this would imply that the proportion of stock pickers is declining.
We, therefore, conclude that stock picking is declining in the United States, because the total cost of stock
picking is increasing in the United States. This total cost is increasing because the direct cost of stock picking –
idiosyncratic risk – is increasing in the United States, though the indirect opportunity cost of stock picking – the
forgone market reward for risk – does not seem to have a trend.
As total risk is not changing (see Schwert (1989) and Campbell et al (2001)), idiosyncratic risk cannot
increase without bound. This suggests that idiosyncratic risk may asymptote to a steady-state. As the market
price of risk seems not to have a trend (see Lettau and Ludvigson (2003)), the product of idiosyncratic risk and the
market price of risk – the “indifferent” excess profit curve -- may asymptote as well. Therefore, if we assume that
the distribution of excess profits in the economy is stable, then the proportion of investors whose excess profit is
above the “indifferent” excess profit curve – the stock pickers – will stabilize where the “indifferent” excess profit
curve asymptotes. This is the steady-state proportion of stock pickers. For the United States, our model
estimates that stock picking will eventually settle at 11% in the United States.
The paper is organized as follows. We develop a metric for stock picking in section II. Section III
describes our data. Section IV is the main result of this paper. It documents that there is more stock picking in
emerging markets than in developed countries, but it is declining everywhere. As we have more data on the
United States, section V covers the United States in greater detail, and reports results for different categorizations
of stocks. Section VI explores the steady state proportion of stock pickers, and due to data availability, we restrict
ourselves to the United States Section VII concludes. In this section, we discuss extensions of our paper, of
which the most important extension is a deeper exploration of why there is so much cross-sectional variation in
stock picking.

- 6 -

---

II. A METRIC FOR STOCK PICKING
Lo and Wang (2000) proved that, if the two-fund separation theorem holds, dollar turnover of a stock,
which is defined as the dollar volume of shares traded divided by the market capitalization of the stock, should be
identical for all stocks. The intuition behind their result was simple. If an agent just invests in the risk-free asset
and the market portfolio, and if the market portfolio has $85 million of stock A and $15 million of stock B, then if
the agents buys (sells) $100 of the market portfolio, she buys (sells) $85 of stock A and $15 of stock B. If prices
do not change between trades, the share turnover of a stock, which is defined as volume of shares traded divided
by the number of shares outstanding of the stock, should also be identical for all stocks if the two-fund separation
theorem holds.
We run the following cross-sectional regression model every month for every market
ln (volume of shares)i = a + b ln (number of shares outstanding)i + ,i (1)
where
Volume of sharesi is the monthly trading volume of stock i,
Number of shares outstandingi is the number of shares outstanding at the end of the month for stock i, and
,i is the volume of shares of stock i that cannot be explained by the number of shares outstanding at the end of the
month for stock i.

If every person in the world indexes between a risk-free portfolio and the market portfolio (or a value-
weighted portfolio that is a proxy for the market portfolio) with little error, trading volume in stock i should be
explained completely by the market capitalization of stock i. So we should obtain the following estimates in our
above regression equation:
a = ln (turnover),
b = 1, and
R2 = 1.
If we do not obtain 1 as our estimate for R2 in the above regression, this would mean that there is
deviation from the indexing investment philosophy. This deviation will occur because some agents pick

- 7 -

---

individual stocks. This deviation will also occur if some agents index to portfolios other than the market
portfolio. As a matter of fact, for K factors, as Lo and Wang (2000) showed, indexing will occur in K funds.
This means that the (1-R2) of the above cross-sectional regression between log volume and log number of shares
outstanding will be a measure of the maximum proportion of trade that can be explained by stock picking.
So our metric for stock picking every month in a market is the (1-R2) of a cross-sectional regression of
log volume of stock i against the number of shares outstanding of stock i in that market. To be precise, the
estimate of (1-R2) gives us the maximum proportion of trade that can be explained by stock picking that month in
a market.
This (1-R2) metric has the following advantages. First, it is simple to estimate. Second, the data
requirement is minimal. Volume and shares outstanding data are publicly available for nearly all stock markets of
the world. Third, the definition of a market is flexible. Markets could be the various country stock markets,
which would allow us to compare stock picking across the world. Markets could be local markets within a
country, like NYSE, AMEX or Nasdaq, which would allow us to compare stock picking across these local
markets. Markets could be defined by different types of stock categorization like size, age of firm, industry, etc.,
which would allow us to compare stock picking across different sizes, different ages, different industries, etc. The
fourth and the biggest advantage of this metric is that the cross-sectional regression can be estimated at different
points in time, which would allow us to detect time-trends, if any, in the popularity of stock picking.
This (1-R2) metric has the following disadvantage. It does not give us an estimate of the proportion of
trade that can be explained by stock picking; it gives us an estimate of the maximum proportion of trade that can
be explained by stock picking. Our estimate of stock picking is, therefore, biased upward. We can decrease this
bias by introducing additional independent variables in our cross-sectional regression. These variables could be
factors that mimic possible factor portfolios that investors also index to, or it could be variables that have
appeared in the volume literature (see Lo and Wang (2000) for a description). These additional explanatory
variables would reduce 1-R2, thus reducing the bias in our estimate. We, however, refrain from this exercise for
two reasons. First, it complicates a simple metric. Second, as these additional independent variables are quite ad
hoc and are not motivated by well-established theories, we would not know how to interpret our new results.

- 8 -

---