Retail Investor .org

The Research on Retail Investors' Returns

(return to Active vs. Passive Investing Returns page)

What are the Returns Retail Investors Earn?

You would think that with all the computing power of today's brokerages, it would be a simple thing to find out factually what individuals are earning on their accounts. Not so. When the question is raised, ALWAYS it is stated for a fact that retail investors underperform the benchmark indexes. When you ask for references or proofs you are assured that there have been a zillion studies and they all prove the same underperformance. Funny but all those studies have gone missing.

Barber and Odean (2011)  have published a summary of past papers dealing with retail investors' behavior and the attributes of their trading and holdings, etc. Most all these papers are irrelevant to the issue of investor returns vs. index returns. Instead they compare investor subgroups. E.g. they measure the 'return' for people who trade frequently vs. those who don't. Table 1 lists a variety of sources, but a closer look show only three to be relevant.  The Cohn, et al paper cannot be found on the web.  The others are discussed below.

Broker Commissions

Schlarbaum, Wilbur and Lease (1978) measured returns with and with transaction costs. They added State taxes and odd-lot price differentials to the actual broker commissions. Median investment returns were reduced by 25%.

Coval et al. (2005) ignored these costs in the metric they used for calculating 'returns'.

Barber and Odean (2000) used data from 1991 through 1996. This era was pre-WWweb, pre-home computers, pre-online trading At that time there was only one index ETF trading in Canada. Index mutual funds were available in the US, but not Canada. Barber and Odean reported (Table 1) mean commissions of 1.5%. On the mean trade value of $12,500 the commission averaged $187.   Today we trade for $5 or $10. The reported returns 'net after transaction costs' no longer has any relevance and should be ignored. Only their results measured as gross returns BEFORE transaction costs are relevant today.

Bauer et al. (2008) measured returns on both a net basis and gross - with commissions added back into the month-end portfolio's value. They found that commissions reduced returns by 1% per month across the dataset (Table 2, page 30). This in spite of their reporting that 75% of investors traded less than once every two months and 75% of investors had monthly turnover of 2.4% or less, of their account's value each month (Table 1, page 29). And their data was from a discount broker for the period 2000 to 2006 when commissions had already fallen far. How is this possible? You just have to question their data.

Who is Analyzed?

The conclusions reached in these papers are generalized to the public who are saving (or have saved) for retirement. Does the sample data represent those kinds of investors? No. The results of these papers are excessively weighted to small accounts of investors with little experience.

Barber's sample has 80% with accounts smaller than US $150,000. Bauer's sample has 90% with accounts smaller than about US$110,000. Coval's sample had a median account of only US$4,500. Barber's sample shows the median trade value of $5,500. If he used the same data set as Coval then the average trade equaled the average total portfolio size. Bauer's sample had 90% with less than 6.3 years of investing experience.

The size of these portfolios will not fund anyone's retirement. They do not represent the knowledge that comes with experience or the risk-aversion that comes when you invest large amounts relative to your income. Bauer's Dutch investors have comprehensive state pensions that leave personal investments free for speculative long-shots.

Risk Adjusted Benchmarks

Only Schlarbaum benchmark against unadjusted index returns. They find the investor's choice of stocks had a Beta of 1.38.

No other papers compare investors' actual returns to the actual returns of large-cap indexes. Instead they compare a metric they create for investor returns, with a synthetic benchmark. The point of this is to normalize for the different risk levels of different portfolios. Whether you SHOULD normalize for risk is debatable.

These studies create synthetic benchmarks using math formulas, usually along the lines of the Fama-French model. So the validity of the comparison depends on your acceptance of their math construct. It can be argued that replacing actual market returns with these math constructs is airy-fairy academic nonsense. One paper adjusts French investors' returns using a wider variety of risk-adjusting-methodologies (Magron (2013). His Table 6 shows that by one metric (the F-T(1.5-2) ratio) investors out-performed on average across the four years of the study.

Stock-pickers try to find mis-priced securities. Since these are more likely to be found in smaller stocks not covered by analysts, they end up owning smaller-caps. If their strategy is to buy low and sell high, they end up with value stocks. If they choose to do no research and tag along with the market, they end up with momentum stocks. Yet returns are discounted for exactly these factors. To dismiss their returns because of the strategy chosen, is to guarantee the conclusion that they under-perform.

Worse still, cash in the account is ignored. It is impossible to measure risk without including cash. Active management's greatest benefit comes from simply exiting the market, into cash, during major downturns. The arrogance of academics pretending they are correctly adjusting for risk, without measuring cash, is unspeakable.

Consider also that any individual's risks come from more than his investments. E.g. If he is an oil worker, and oil stocks are booming, is it appropriate to claim he under-performed the market because he held no oil stocks? No. His portfolio choice was correct. It reduced his risks. But the researcher would have no way to know that, or correctly risk-adjust his returns.

The use of risk-adjusted metrics is valid only if the investor changes his debt-equity asset-allocation to compensate for the added risk of his securities. In real life investors make no asset allocation adjustment. Investors accept the difference in risks as immaterial - whether it is or not. The asset allocation decision is made before deciding which stocks to buy. The stocks chosen are the ones with potential profits. All profits derive from some 'reason'. If you refuse to recognize a profit because it has a 'reason' that you call 'risk', you are simply refusing to recognize profits.

Investor Returns

None but Bauer measure portfolio performance the way common sense would dictate. (Ignore transactions. Measure the change in the portfolio's value between the beginning and end of each year. Factor in cashflows, in and out of the account. Calculate the IRR for each year, or better use a per-unit-return. Then compare the results to the relevant large index benchmarks. This would be the way you calculate your own return. This would certainly be the simple method. It is the method that generates meaningful results with the fewest arbitrary modifications. But academics won't use it.

Schlarbaum measure only the completed round-trip stock trades. They ignore cash and the 40% of trades in bonds, options, mutual funds, etc. They ignore the 20% of one-sided trades for positions open at the beginning and end of their study. They use actual transaction prices and calculate the holding period's IRR for comparison to the benchmark over the same period.

Barber and Odean are most frequently quoted, but their methodology has problems. Their results are used to 'prove' that active stock-picking underperforms the benchmarks, yet their methodology by its definition creates its own conclusion. They treat all trades during a month as if they happen at month end. This has the effect of ignoring all gains within the month of stocks bought as they rise in price. This has the effect of adding additional losses on stocks sold as they fell in price. Since most people buy rising stocks and sell falling stocks, this methodology very effectively reduces their measured returns. This methodology explicitly presumes there is NO value to stock picking in their research to determine IF there is any value to stock picking.

Because their methodology does not use actual transaction prices (because they use the month end) they invent another way to reduce their measured returns. They assign a 0.6% cost for the bid-ask spread on sales and a 0.3% cost for purchases (Table 1), so for each round trip returns are reduced almost 1%.

Another problem is that they report returns as a monthly average. It is only year by year results, compared to each year's benchmark that are meaningful. Underperformance during market bubbles is not a bad thing if your strategy delivers big-time in years of market declines. Then also consider that they do not include cash in their performance measures, etc, etc.

Bauer has produces the most correct measure. They measure the account as a whole, just like you would measure your own returns. They seem to include cash in the monthly valuation. A possible bone of contention is their decision to date all cash inflows at the beginning of the month, and all withdrawals at the end of the month. This increases the amount considered to have been invested during the month and reduces the calculated return. They also measure and report averaged monthly returns, instead of yearly results to compare with index returns.

The Papers' Results

Schlarbaum, Wilbur and Lease (1978 Realized Returns on Common Stock Investments: the Experience of Individual Investors) used 7 year's of data from 1964 through 1970. They found that investors' gross returns without risk adjustments outperformed the index by an average 5.3%, mean 2.8%. Short-duration trades had higher excess returns, but less impact on the total results because of their short duration.

Barber and Odean (2000) Trading is Hazardous to Your Wealth ) (page 786) found that the gross returns (before transaction costs) averaged 18.7 percent for investors vs. 17.9 percent for the synthetic benchmark. Investors outperformed the indexes but that may have been due to their tilting toward small cap and value stocks. Even the net returns were excellent - a 16.4 percent mean. Considering that the 1.5% trading commissions no longer exist that brings the mean net return exactly equal to the indexes.

When their sample was sorted by portfolio size, the smallest portfolios showed the greatest outperformace, with lower returns as the size increased (Table 3 market adjusted returns). There was a wide variance between returns. "25 percent of all households beat the indexes, after transaction costs, by more than six percent annually" and "25 percent of all households underperform the market, after transaction costs, by more than eight percent annually" (page 790).

When their sample was sorted by frequency of trading, those trading frequently outperformed (gross returns) the US indexes and those trading infrequently (Table 5). The authors dismissed these results in favour of talking about the underperformance relative to their risk-adjusted benchmark.

Coval, et al (2005) Can Individual Investors Beat the Market?  ask "whether some set of real investors have demonstrated abnormal skill in generating abnormal trading profits". Their "results suggest that skillful individual investors exploit market inefficiencies to earn abnormal profits, above and beyond any profits available from well-known strategies based upon size, value, or momentum." In other words, they outperformed risk-adjusted benchmarks.

They also found strong persistence in investor's performance. "We find that trader performance, regardless of measurement horizon or risk adjustment, is consistently correlated across the two sample halves" (page 3). "Investors classified in the top performance decile in the first half of our sample (4 years) subsequently outperform (in the subsequent 3 years) those in the bottom decile by about 8 percent per year. "Traders in the top decile (based on the performance of their other trades) buy stocks that earn 0.12% to 0.15% per day during the following week. Trades in the bottom decile lose between 0.11% and 0.12% per day" (page 4).

But their assertive conclusions are not so clear. Table 5 shows that the claimed 8 percent outperformance by top-tier investors comes from the 3.978 percent difference in holding period return multiplied by 2 because the mean holding period was half a year. But between the results for the best and worst performers the relationships break down. Absolute returns are high at both ends. Risk-adjusted excess returns vary widely. Trading frequency seems to vary with absolute returns, but not risk-adjusted returns (more frequent trading at both ends). It is only the top decile performers from the first 4 years that seem to show clear outperformance in the subsequent 3 years. Still if 10% of investors have IT then is it not worth finding out if you are in that camp?

Bauer et al. (2008) Option Trading and Individual Investor Performance were a complete disappointment for what they did NOT report - even though they took all the trouble to do the measuring. They reported returns with a positive alpha when risk adjusted (Table 2). Their results for returns net of commissions is not credible, as discusses above. They chose to not report on the individuals' yearly returns or the country's benchmark index. They chose to not report the distribution of returns by decile. They chose to not report the distribution of returns by age, or portfolio value, by frequency of trading, etc.

No Skill, Mere Luck? by Hackethal (2012) tries to separate investors' excess returns between luck and skill. To accept their results you have to accept that their 'bootstrap zero-alpha benchmarks' are correct. They found that investors' lack of skills cost them 7.5% annualized. They also found that overall investor returns are not statistically different from their model's expected returns. Given those two results it must be that luck add a whopping 7.5% gain.

What is never discussed in the paper is the common sense objection that, since luck is by definition random, how can it benefit 8,600 portfolios over a 4.5 year period so highly? Would not the luck run out over time. Would not good AND bad luck even out over such a large sample?

Conclusion

These papers have not addressed head-on the issue of retail investors' portfolio returns. The results generated are not conclusive but only indicative because of the failure to measure actual portfolio returns. If you believe that benchmarking should be done against risk-adjusted indexes, then maybe the results can be seen as showing underperformance. But if you will not be changing your asset allocation because the specific stocks you choose are either smaller-cap or value stocks, then it is clear that retail investors outperform. There is evidence that the top 10% of investors show an ability that continues over a 7 year period at least.

(return to Active vs. Passive Investing Returns page)