Reader Question
How important is it to ramp up your investments over time? I started from a non-investing position and am slowly investing everything in 2 week intervals over the course of the year. The goal is to normalize risk over time but I worry that I'm missing opportunities to invest as they are simply sitting in savings accounts. What is your perspective on this?
My Reply
This is much more a psychology question than a finance question. On the finance side things are pretty clear: because all investments are expected deliver positive returns over time, the sooner you invest the higher your returns can be rationally expected to be.
Unfortunately, "expected to be" is not the same as "will be". This is where the psychology comes in. Bear markets do happen and sometimes those expected returns do not materialize -- at least not for many years or even decades. If we happen to be heading into a bear market you will certainly be glad that you waited. But there is no way of knowing if that is the case. Obviously, plenty of people do claim to have this knowledge of future bull or bear markets. But for some reason these people also tend to work 9-5 jobs and spend a lot of their time convincing others of this special ability of theirs instead of cruising the Tropics on their mega-yacht. Draw your own conclusions on their market-timing skills.
If some investments were obvious screaming bargains, then things would be simpler. But unfortunately that is not the case. Bonds range from somewhat to very pricey, Stocks might possibly be fairly priced, but certainly nowhere close to cheap (cheap was late 2008 and early 2009). Gold is obviously not cheap after its tremendous run of the past decade plus. Even cash is "expensive" as it's likely getting not too much more than 0% return for you, so it's not like you are getting compensated for deferring your investments.
I think of spreading out potential one-time investments over time as a way to minimize regret. Regret is a powerful and unpleasant emotion. Few people are immune from it completely. In extreme cases, regret of past actions can lead you into making new mistakes. It's very rational to minimize regret as long as the cost of doing so is not excessive. By spreading your investments over a year you are extremely unlikely to ever have to regret investing at the worst possible time. You will always be able to point to one date and say "well, at least I only bought a little bit there". About other dates you will be able to say "I wish I bought more then, but at least I got something on the cheap". And so on. Obviously all of your purchases could still end up under water after some time as a bear market accelerates. But there's no escaping bear markets completely. At least you will have some relative highlights to think back to instead of the crushing regret of buying everything at the top.
But make no mistake. All you are accomplishing by spreading out your investments over time is minimizing potential regret. If it ends up a profitable strategy it will be purely by luck -- not by design. By design this strategy will lose a bit (but likely only a little bit) relative to the buy-it-all-at-once approach. If you are finding yourself regretting the very choice of trying to minimize regret by spreading out your investments, then perhaps it's easier to just pull the band-aid off all at once and immediately commit all your funds earmarked for investing. Only you can know which choice will be easier on you psychologically or whether the psychological gain is worth the (small) expected loss of missing out some time in the market. Obviously, whichever way you go, make sure you are sufficiently educated on how to invest.
Dear LTR,
ReplyDeleteI am curious to hear what you say about the following thought experiment.
Say you invest only once in your lifetime. You do it immediately when the money is available because you have a positive expectation to make money. Let's say there is a fixed sell date then on that date the expected value is some positive outcome but the actual realized outcome may vary greatly.
Now contrast that with taking that big amount you had above and say spread the buying spree over a single year. First let me make an incorrect analogy. Thing about it as repeating an experiment very often. You would expect by some law of large numbers to get closer to the actual true expected value than if you repeat the experiment only once. Of course it is easy to point out flaws in this analogy since the the experiments are not independent. Also one has to take into account that money that you invest last has missed out on possible returns.
I wonder though if there is something to the moral of the story above. As in, if spreading out investments reduces risk in the sense that you are more likely to be closer to the actual expected value of the market but at the cost of having money sit "dead" for a bit. Maybe there is some cost-benefit analysis here?
Perhaps the length of the investment horizon plays a part too? For example I could imagine on a 10 year horizon then buying at the bottom of the crash versus the highest point within a year around the crash could have some actual effect on the net result. On the other hand I could see that effect essentially disappearing if the investment horizon is 40 years.
What are your thoughts on this LTR?
Very thought-provoking comment. I would need to think about this awhile longer...
DeleteMy initial reaction is that yes, undoubtedly, by spreading out your investments over time you would average out market's estimates of fair value so in that sense you would capture a more "fair" valuation. But, at the same time, the market's estimate of fair value is expected to go up over time. So you do still sacrifice expected return in exchange for a more "fair" price.
An inverse way of looking at it is that cash (and only cash) always has absolute fair price. If you never invest your cash, then you can be assured of always getting perfectly fair value for your money. Viewing our thought experiment from this point of view, by spreading out your investments you simply choose to hang on to some of the cash for longer, increasing the average "fairness" of the price you pay for your portfolio holdings. But, clearly, taken to the extreme -- never investing any cash and getting the ultimate fair value in return -- is not a good strategy at all. Is there a reason to believe that a scaled-down version of this get-most-fair-value-possible principle is useful where the full-blown version fails? I can't think of one.
My approach would still be to pick asset allocation that is psychologically tolerable in the inevitable ups and downs of the market and, ideally, contains some less-than-perfectly-correlated asset class (e.g. both stocks and bonds, both in substantial amounts) so some of the risk is diversified away without sacrificing expected returns as would happen with a spreading-investments-over-time approach.
Context: I got this advice from the intelligent asset allocater book (by william bernstein)
DeleteDear LTR,
ReplyDeleteI am glad to hear that the comment was thought-provoking. For full disclosure then I already had one reader question and am now an avid reader of your blog.
http://www.longtermreturns.com/2012/09/financials-valuations-buffett.html
To summarize my comment above then the idea is that spreading investment may lower the expected return of an investment but may improve the probability distribution of returns. Another analogy is that you recommend, as I agree with, that one you should invest in the entire market instead of making bets on very particular parts of it. The idea is to improve the probability distribution in the sense of getting more close to the expected value of the entire market instead of having a more volatile distribution (as in you could potentially gain more from a particular sector but also might lose a lot more).
Also, let me be very clear, that I so appreciate the advice you give on this site and I am not trying post any of this as trying to counter any of your previous advice. I am just trying to get a good conversation going because from an outside point of view you seem like a person that analysis all options thoroughly so any new point of view could be fun.
I absolutely appreciate your comments, so no need to apologize for presenting a different point of view (especially one advocated by William Bernstein whom I respect greatly). People can disagree while still respecting and learning from each other.
DeleteWhere I see the analogy between diversifying in holdings versus diversifying in time break down is that diversifying in stock holdings does *not* decrease your expected return while decreasing risk. It really is a free lunch.
Diversifying in time also decreases risk but at the expense of decreased expected return (just as holding cash does, going back to my previous analogy). It may be that some people would find it a worthwhile trade-off, but to me it still seems to achieve primarily psychological benefits. That's not to discount the importance of psychology -- if you invest in a way that is fundamentally incompatible with your temperament you are likely setting yourself up for some impulse bad decisions down the road (e.g. selling out of stocks in a bad bear market).
One good project would be to backtest whether and by how much lump-sum investing bested or lost to spreading-out investing. E.g. investing in S&P500 as a lump sum January 1st of each year vs investing 1/12th of the lump sum on the first of each month of each year. If I ever get some more free time I'd run the numbers, but free time has been hard to come by lately.
Yes, that would be a very interesting project. Where would you recommend looking up this sort of information? (In case I have time over the holidays.)
ReplyDeleteThere are quite a few free data sources for these types of projects.
DeleteIt's always best to use real-world investment data (as opposed to synthetic indices, especially ones constructed after the fact) whenever possible. For this approach it would hard to best Vanguard S&P 500 Fund, VFINX, which goes back to mid-1970s.
Unfortunately Yahoo Finance only seems to have downloadable data for it going back to 1987: http://finance.yahoo.com/q/hp?s=VFINX . On the plus side Yahoo has built-in dividend adjustment (the "Adjusted Close" column) and lets you download either daily or weekly or monthly data. Yahoo Finance has pure S&P 500 price data going back to 1950 but that's next to useless since it omits dividends.
Next possible source of data is Morningstar which is great for charting but probably useless for this project because it does not have downloadable source data. Instead you'd have to scrape it manually from its excellent charting tools. It would be accurate going back to the creation of VFINX but far too time-consuming to be practical.
To go back even further in time we can rely on Professor Shiller's S&P 500 synthetic data available at http://www.econ.yale.edu/~shiller/data.htm (there's a link to XLS near the top). I would not rely on these data for absolute numbers (especially going back to early 20th century and earlier), but since we are comparing two investing strategies instead of looking for absolute returns I don't see any reason why these data wouldn't be acceptable.
Finally, MSCI Barra has excellent indices for almost every country and region going back to 1970 for developed markets available for download as price-only and as gross total return (meaning with re-invested dividends -- what we need for this project) at http://www.msci.com/products/indices/country_and_regional/dm/performance.html (you might have to click around to go through their disclaimer but that site is well worth exploring). This would let you run the same analysis for different markets, not just S&P 500.
Dear LTR,
ReplyDeleteI haven't gotten around working on that project but I did find a nice paper that I think addresses most of the points that were raised in the comments.
https://pressroom.vanguard.com/nonindexed/7.23.2012_Dollar-cost_Averaging.pdf
First of all then the LSI (Lump Sum Investment) method delivers on average higher returns as we expected because of the general upward drift of the market and because of the idle cash in the DCA (Dollar Cost Average) method. What I did find interesting though, as seen in Table 3, then the probability distribution for DCA is narrower in some sense than the distribution for LSI. In other words, one can expect wilder outcomes with the LSI method. Perhaps that is a very natural thing to expect.
I wish the authors of the paper had calculated the expected return for each so that we would get a better feel for what the price was for reducing the risk. Of course they try to address it with the Sharpe ratio calculations but what I found interesting in that table is that 100% bonds has a better Sharpe ratio than 100% equity. Is that to indicate that in some sense 100% bonds is better than 100% equity when it comes to return vs risk? Would one expect that?
This whole exercise then made me think about the return vs risk for different equity distributions, such as 90-10 vs 50-50 models and how the return vs risk there compares to the return vs risk that is incurred based on the choice of DCA or LSI decision. I am interested in it because I feel the DCA vs LSI decision is often described as being just about sentiments while I feel people take the 90-10 vs 50-50 (equities vs bonds) question more seriously.
Thanks for taking time to follow up on this. That LSI on average beats DCA and that DCA's return range is narrower than LSI's are both natural conclusions. It is, I suppose in some ways similar to how stocks will on average beat bonds, but bonds returns will be less variable over time.
DeleteYou can view DCA as choosing to stay in stock/bond/cash allocation for some time instead of diving 100% into stock-bond allocation. Cash is safer than bonds and, on average, will return less than bonds. That makes DCA both safer and expected to return less.