Wednesday, October 3. 2012
The Behavior Gap
With the recent surge in the stock market, S&P 500 with all dividends reinvested has reached an all-time high.
Numerous articles in the newspaper and on the web typically show retail investors as a whole are "dumb" and imply that they took money out of the stock market near the bottom and are now putting money into the stock market just as it makes new highs.
No doubt such behavior exists. How bad does it hurt the investors' return when they buy high and sell low? The most widely cited study is probably DALBAR's Quantitative Analysis of Investor Behavior. This study compares the investors' returns against market returns. Mutual fund tracking company Morningstar also calculates investor returns for every fund and compares them against the fund returns.
Investor return is a dollar-weighted return (or more generically "money-weighted return"). It takes into account the size and timing of investors' purchases and sales. If investors put a lot of money into a fund and the fund does poorly after that ("buy high"), the investor return will be low relative to the fund's published (time-weighted) return. Similarly, if investors pull a lot of money out of a fund and the fund does well afterwards ("sell low"), investor returns will lag.
The latest DALBAR study shows the investor return in all equity funds in the 20 years ended in 2011 was 3.49% a year while the S&P 500 returned 7.81% a year during the same period. This suggests that the average equity investors' poor market timing cost them 4.32% a year for 20 years - a gap in performance due entirely to their own harmful behaviors.
I've seen this interpretation in many books, including books by respected authors Burton Malkiel, Larry Swedroe and Rick Ferri. While the intention is good -- warn investors against buying high and selling low -- the interpretation is wrong because comparing investor returns against index returns is comparing apples to oranges. The 4% a year number is so incredible that it makes the number not credible. The so called "behavior gap" isn't as high as the DALBAR study says.
Flawed Methodology For Measuring The Behavior Gap
The reason the DALBAR study is incorrect is that while "buy high sell low" will make investor returns lower than market returns, it's not the only factor. When using dollar-weighted returns, the pattern of market returns over time also plays a big role. Consequently, when you see the investor return is lower than the market return, you can't necessarily attribute all (or possibly even some) of the difference to "buy high sell low."
Let's look at two hypothetical examples.
Suppose the stock market doubled in year one and then stayed flat for nine years. Over the 10-year period, the market return is 7.2% a year ("rule of 72"). If an investor invested $1,000 every year in an index fund that exactly matched the market, the investor would have $11,000 at the end of 10 years. Only the first $1,000 doubled. The other $9,000 had a zero return. As a result, the investor's dollar-weighted return is only 1.7% a year for 10 years.
The big difference between the market's 7.2% per year return and the investor's 1.7% per year dollar-weighted return isn't caused by any performance chasing or bad market timing. The investor is just faithfully investing in an index fund for the long term. When the market did well in year one, the investor simply didn't have much money invested to catch the good return.
Now suppose the stock market stayed flat for nine years and then doubled in year 10. Over the 10-year period, the market return is still 7.2% a year. If an investor invested $1,000 every year in an index fund that exactly matched the market, this investor would have $20,000 at the end of 10 years, resulting in a dollar-weighted return of 12.3% a year for 10 years. It's higher than the market return because in the year when the market return was high, the investor had $10,000 invested versus only $1,000 invested in the previous example.
Depending whether the market had higher returns in the beginning or in the end, investors are seen either as dumb or smart even when they made no effort to time the market.
That's exactly what happened lately. Morningstar shows some mutual funds have investor returns much higher than the fund returns. Here are some examples:
|Vanguard Target Retirement 2045 (VTIVX)||1.09%||4.35%||+3.36%|
|Fidelity Freedom 2045 (FFFGX)||-0.08%||5.43%||+5.51%|
|T. Rowe Price Retirement 2045 (TRRKX)||1.52%||5.53%||+4.01%|
* Source: Morningstar. Data as of Aug. 31, 2012.
Are investors in these target date funds geniuses in timing the market? In addition, are investors in the actively managed Fidelity and T. Rowe Price target date funds smarter than investors in Vanguard funds because they beat the fund returns by a bigger margin? I don't think so. The measured 5-year period starts in 2007, before the crash. When the fund return was bad, investors didn't have much money in these funds. As more money came into the funds, the market had better returns. That's all.
DALBAR Results More Sequence Of Returns Versus Bad Timing
When you see that big positive difference between investor returns and fund returns can be caused by when the market had good returns, you know big negative difference can be an accident of history as well. It just so happens DALBAR's study period begins in 1992 and ends in 2011. The stock market had good returns in the first decade and bad returns in the second decade. No wonder there is a big negative difference.
Because DALBAR sells the study to financial advisors to show how investors do poorly on their own, DALBAR has an incentive to exaggerate the effect of poor investor behavior. The 2012 edition of the study also includes some contrived arguments against the fiduciary standard. It's off topic for this post but it also reflects where DALBAR's allegiance is.
To its credit, the latest DALBAR study also shows investor returns of a dollar cost averaging investor. If an investor invested a fixed amount in equity funds every year, the investor return would have been 3.17% a year for 20 years, compared to the actual investor return of 3.49% a year. The average investor beat dollar cost averaging! Why isn't that the headline?
- DALBAR: Quantitative Analysis of Investor Behavior, Advisor "Free Look" Edition, 2012
- Russel Kinnel, Morningstar: Bad Timing Eats Away at Investor Returns
- Morningstar: Investor Returns Fact Sheet
Tracked: Oct 03, 10:21
Tracked: Oct 03, 11:45
Tracked: Oct 03, 14:00
Tracked: Oct 06, 05:43
Tracked: Oct 08, 07:04
Tracked: Oct 15, 02:16
Tracked: Nov 17, 13:04
Probably like most advisors I see a few people bail at market extremes but most do nothing. Some are forced to get out at the bottom because of job issues etc. Emotions play a role but it seems they may have been over emphasized.
Indeed, clearly some people time volatile markets badly. We've all seen them as advisors.
However, the implication here is that we're seeing a non-representative sample. We see the people who are panicked, and/or bailed out. We don't see the people who were nervous and sold early (when the market was still high) and bought in after the crash, navigating the markets successfully - as those people are probably much less likely to seek out an advisor given that they're already doing well.
Clearly there are issues comparing DW and TW returns in studies like these. Couldn't much of your question be answered by examining the total fund assets at each interval? I'm sure that fund assets have grown over the time period studied, but can the study be controlled for that? I would also note that DALBAR has conducted this study for many years (and thus several rolling periods) and the results have held.
Given the sequencing issue, where returns were great in the 1990s, peaked in 2000, and declining ever since, adding more time to the end of the study means little - every year they continue to add since 2000, where the market has not made material new highs, will simply reinforce the time-weighted versus dollar-weighted gap that's already there.
And notably, DALBAR to my knowledge has not looked at DIFFERENT rolling time periods. They keep taking the SAME time period and just adding bad years onto the end (given mediocre cumulative returns since 2000), which just builds on and exacerbates the issue.
I think you are mistaken about the time periods - I believe each study has always been a 20-year rolling period, so the earliest studies would have significantly different time periods than the most recent. So we aren't talking about adding time to the end of the study, but moving the 1990's closer to the middle of the study. Can anyone find out who is right on this one?
About 18 months ago, I wrote a blog post on this topic and I highlighted what I think is an even more basic methodological weakness of the historical DALBAR studies. I tried contacting DALBAR a couple of times to get their input but received no response.
My post is linked below. I'd be interested in knowing if the methodology description is very different in the current report as in the older report cited in my article.
My best interpretation is that they are calculating dollar-weighted returns only on the flows, ignoring the beginning balance. In other words, if I start out with $1 million, add or take out $1k, $2k here and there along the way, my DALBAR return is calculated only based on my $1k and $2k trades. I could screw up my $1k and $2k timing really badly but my overall return isn't that bad because it's predominately determined by the $1 million that doesn't move.
The fixed income returns, however, are inexplicable both for one-year and longer horizons. Those numbers make no sense at all.
So it seems that not only is DALBAR not comparing apples to apples (i.e., time-weighted vs. dollar-weighted), they aren't calculating anything meaningful at all! It appears to be some bastardized version of dollar-weighted, but actually has no informational content.
What you say may be correct because the gaps shouldn't be so wide as time goes on and assets grow and net flows shrink as a % of assets.
But looking at this again it appears as if DALBAR is lumping an awful lot into this 'gap'. Their description today sounds more like they are calculating investor returns in funds. But they are not comparing against the potential returns on the same funds (via buy and hold or DCA).
They appear to be comparing investor returns in funds vs. benchmark returns for U.S. stocks and bonds. And that gap includes things like fund expense ratios and manager value added/detracted. It also includes performance differences attributable to any investing outside of the benchmark that funds do. In other words, lots of noise that has nothing to do with investor behaviour.
As a test, if you know anyone at DALBAR, ask them if they can provide their return data for just the 1990's. Given the significant growth in mutual fund investing and hearty returns at the end of the cycle, I would expect to see the dollar-weighted investor returns exceed the indexed time-weighted returns.
I'm not sure I follow your point.
The problem is not that it's a rolling 20-year period, per se.
It's that the implied "behavior gap" is largely an artifact of what happens using the DALBAR methodology over a 20-year period when dollars from into markets over time.
The fundamental point is that, even with DALBAR's own data on rolling 20-year periods, THERE IS NO BEHAVIOR GAP if we merely assume investors dollar-cost-average over time! No one said the MF investor had a great run and then walked away; in fact, the problem is the opposite, that DALBAR's methodology is distorted by the buy-and-hold investor compared to the (arguably more realistic) systematic dollar-cost-averaging investor who saves whatever he/she can year by year over time.
The point here is that the claim "investors pile into investments and markets after they have risen and exit en masse after they fall" actually doesn't really hold up with the data after all.
Yes, it holds up if we assume all our clients never earn a dollar of employment income and never add to their savings.
But if we assume that they do in fact save systematically over time from their earnings by contributing to IRAs, 401(k)s, etc. - which is effectively dollar cost averaging - then what the results show is that investors don't actually move in NEARLY the herds that we imply. In fact, they actually time the markets BETTER than just blindly, passively saving!
Obviously, SOME investors are at the wrong end of that trade, but the point made here is that, when you control for the fact that most people are ALREADY adding to their savings over time with contributions, that in the AGGREGATE investors are not the bad market timers we make them out to be.
Or viewed another way, most of what we imply is "investors chasing returns" is actually just "investors redirecting their new savings" and they're not hurting themselves nearly as much as we suggest.
Again, that's not to say that there aren't some who do - and those are disproportionately the ones who seek out advisors, so out perspective is very biased. But in the aggregate, the results suggest that investors may be BEATING passive dollar-cost-averaging strategies with their new savings, not underperforming as a herd.
So interpreting what was written in their QAIB report, I surmised that they weren't really calculating investor returns and the gap they reported wasn't the real gap to begin with.
That said, dollar cost averaging (DCA) is relevant because that's a more realistic scenario for investors at large. In other words, most people simly invest when they have the money not in a giant lump sum with no subsequent flows.
As Stephen Nesbitt did in his terrific (and simple) paper, he compared DWR using actual flows to a DWR of investors doing regular fixed contributions - i.e. DCA. That's what he defined as the behavioural gap.
Virtually every fund category had a TWR that was higher than the DCA-DWR and that is attributed simply to the fact that there is a drag in a rising market when you don't invest all of your money on day one - so let's not count that as a behavioural weakness of investors.