Tuesday, January 25, 2011

Adverse Timing Kills Returns

The average return of an asset is merely looks at the price. One can argue whether geometric or arithmetic averages are best, but let's ignore that here. Look at the firm ESLRD, Evergreen Solar. They make solar power cells, so it's a sexy product, one very amenable to sales pitches of the next black swan that rubes do not fully appreciate. It came out in November 2000, with $112MM of insider money, and $42MM from outsiders via their IPO. The price (split adjusted) was $84.

The price fell over the next couple years, then rebounded in 2006 and 2007, and lately collapsed, and is now trading at $2.50. They have never made money. In 2006 revenue peaked at $103MM, up sharply from 2005, which instigated the bounce in stock around that time. Alas, it did not continue, and revenue has fallen, and it appears they will never be viable (current market cap is only $87MM).

Unfortunately for investors, their total of $446MM worth of injections were not random, but rather at relative stock highs, such as in 2007 and 2008. Look below and the red lines are investor contributions, and the black line at the end is their total value of those red lines today. If you look at the Internal Rate of Return, you get a very different picture than the top-line average return. The internal rate of return on an investment or project is the "annualized effective compounded return rate" or discount rate that makes the net present value (NPV) of all cash flows (both positive and negative) from a particular investment equal to zero.

It's especially useful when analyzing funds, because often they have incubation periods with little money, and they are presented only if successful, a clear selection bias. Then, they muddle along. But if they made 50% their first year with $10MM (their equally prescient yet less fortunate peers lost money and never went live), and then made 5% the following 6 years when they had $250MM under management, allowing them to look like all-stars based on top-line returns.

In contrast, most stock data merely looks at the return of stocks independent of cash flows, the annualized return is around -10%, consistent with a stock going to zero over 10 years. But the Internal Rate of Return was -47%, much worse. This is the average return to your average dollar that came in, weighted by how long it was there. It is a much more accurate picture at what an average investor endured.

Clearly in cases of failed companies, they generally will have much lower IRRs than 'average rates of return'. Ilia Dichev (2005) looked at a variety of indices, and equity net inflows, and found this adjustment reduces returns by 1.3% for NYSE/AMEX for the period 1926-2002, 5.3% for Nasdaq from 1973-2002, and 1.5% for 19 major international stock exchanges from 1973-2004. Applying this correction to the entire market is something that is never done in all those 'stocks for the long run' or Ibbotson analyses.

Why don't these experts recognize this bias? Well, economists hate to do anything that decreases their best example of the risk premium, the elusive thing that explains everything and nothing; it is implied by their basic conception of utility used everywhere, and so they can't simply turn it off. As to index providers, every one of them has a vested interest in indices they catalog, they are not disinterested providers of information. Just as Moody's used to selectively present their default data (note that munis and asset backed securities are separated from corporate defaults?), the major index providers have a strong vested stake in having their product look good.


Anonymous said...

See "Dollar-Weighted Returns to Stock Investors: A New Look at the Evidence".

Dichev (2007, American Economic Review), in an influential paper, examines the gap between the performance of major stock markets and the dollar-weighted performance of investors in these markets. He finds a significant gap of 1.3 percent per year for NYSE/AMEX and 1.5 percent internationally. We question these results. The NYSE/AMEX performance gap is actually negative in the last two thirds of Dichev's 1926-2002 period, while his international results are influenced by a dramatic increase in Datastream's coverage. When, instead of Datastream, we use a comprehensive share price database, the U.K. performance gap changes from 1.1 to -1.3 percent. In short, Dichev's findings are not robust.

Anonymous said...

Dichev wrote a recent paper on hedge funds which also looked at dollar wtd returns. He then compared those #'s to the time wtd returns on the S&P 500. This seems like a little bit of intellectual dishonesty as from his previous paper he was aware that there is the same behavioral problems with the long only investors.

Eric Falkenstein said...

Thanks for the link. I think they note Dichev's 5% Nasdaq estimate does not generalize, but no one, not even me, thought it did. As per looking at it by 25 year subsample, I suppose one could argue that a set of subsamples is better than the population, but that argument could go either way. As per the UK data point being lowered, ergo all 19 countries are not relevant, that's not a logical inference (possible, yes, but hardly definitive). I would have liked to see them do all 19 countries.

Aaron Brown said...

While this is a valid and important point, it applies only to non-investable indices.

Suppose a company is 1% of the S&P500 when its stock sells for $100/share. It issues new shares at that price to raise itself to 2% of the S&P500. The price falls to $50 and the company buys back half its shares, so it now represents 0.5% of the index. Its price climbs again to $100.

The stock returned zero over the period, but as a group investors in the company lost half of its original value.

But the stock does not have zero impact on the S&P500. It fell 50% while representing 2% of the index, so it reduced the index by 1%. It then rose 100% while representing 0.5% of the index, so it gave back 0.5%. Overall, the S&P500 is lower by 0.5% as a result of this stock, mirroring investors' losses of half the initial value of 1% of the S&P500.

This is one of several reasons that rigorous works usually emphasizes investable indices, preferably actual values (like the return to Vanguard Index 500) rather than theoretical investable calculations.

The same effect on the S&P500 itself plays out differently. It could be true that, say, individual investors buy more when the index is high and sell more when the index is low. In that case they will realize less than the stated index return. But that requires other investors, say institutional value investors, to realize more than the stated index return. So the index is stated fairly.

Eric Falkenstein said...

Aaron: I think you are correct for dealing with specific tickers. The issue then is whether all equities had that behavior. That is, if all stocks issued new shares at $x, each stock fell to 0.5*x, and all companies buy back half their shares, the dollar weighted returns are different. It's not straightforward because in the case of ESLRD, most of the issuance was internal--warrants, ESPP, restricted grant shares, eq. option exercises--and these issuances are different than dividends and buybacks.

bidrec said...

I believe that part of the collapse in stock price is due to the bankruptcy of Lehman Bros. Lehman lent them money with Evergreen stock as collateral. When Lehman collapsed Barclays got the stock and dumped it on the market. It is tough to support your stock price in the face of something like that.

Aaron Brown said...

That's a good point. It brings up a distinction between the index number versus the index portfolio.

If every company in the S&P500 does what I suggested at the same times, then their relative shares of the index don't change. The S&P500 would remain the same as companies representing 100% of the index fell 50%, then rose 100%. So S&P500 investors as a whole would have lost half their money while the S&P500 was flat.

However an investor in the S&P500 index would still break even. The loss has to be borne by people who bought the extra shares at $100 and sold them back at $50. These could be investors in individual stocks, or investors who timed their S&P500 index investments badly.

So the index is a good measure of what a buy-and-hold index investor earns. It is not necessarily a good measure of what an investor in the S&P500 portfolio earns.

Internal issuance only matters if we consider company insiders separately from the company. Otherwise, if the company gives the CEO a million shares, and she sells them, from investor accounting perspective it's the same as if the company issues the shares. If she doesn't sell them, it's the same as if the company issues the shares and the CEO buys them.

Anonymous said...

I haven't read this yet, but it might be of interest.

Measuring Investors' Historical Returns: Hindsight Bias In Dollar-
Weighted Returns

Simon Hayley
Cass Business Schoo1
17 November 2010

Dollar-weighted returns have been used as evidence that consistently bad timing drags investor
returns significantly below the buy-and-hold market return, and that consequently the equity risk
premium is overstated. In this paper we show that this approach is affected by hindsight bias in
the dollar-weighted return. We present an alternative method which quantifies and removes this
effect. The results show that bad timing has actually had little impact on investor returns from
mainstream US equities. We conclude that previous estimates of the equity risk premium remain