Arnold Kling  

Safe Investing

Hydrogen Cars... Price Discrimination and Infor...

What is a safe investment nowadays? Zvi Bodie argues for inflation-indexed government securities.

The probability that stocks will perform worse than risk-free investments, such as TIPS or I-Bonds, gets smaller the longer you hold them. After 30 years, there is a 95% chance that stocks will beat TIPS. But the severity of a possible shortfall gets bigger. For example, say you have $500,000 and are three years away from retirement. If you suffer three years of 15% losses, your savings will be almost cut in half -- and your retirement will be jeopardized.

Bodie is particularly hostile to the view that stocks are somehow less risky the longer the holding period. See this paper.

Alan Blinder argues for short-term cash instruments.

"There is … [the] quite large risk that people who are heavily into long bonds will be sitting on large capital losses," Blinder frets. If interest rates on 10-year Treasury notes rise from 4 percent to 6 percent, for example, the value of those notes will drop by almost 14 percent. "You can have a sizable loss, and that's not well understood," he says.

Blinder was interviewed in the AARP journal, which presumably explains why he was counseling such strong risk aversion. The recent behavior of bond prices makes his advice seem well timed.
For Discussion. Are investors in stocks and bonds under-estimating the downside risk?

Comments and Sharing

TRACKBACKS (3 to date)
TrackBack URL:
The author at Modulator in a related article titled Late Night Reading writes:
    If you are wondering whether the NY Times talks more about Michael Moore or Ann Coulter then head over to Virginia Postrel's place to find out. Arnold Kling, EconLog, asks 'What is a safe investment nowadays?" Have you done the boogaloo! Languagethat t... [Tracked on July 30, 2003 1:06 AM]
COMMENTS (46 to date)
Eric Krieg writes:

Regarding the riskiness of stocks, many people my age have run into huge stock losses early on in our investing lives. I started my 401k in 1997.

Because of the time value of money, isn't one strategy to avoid losses in your early years? With a significant negative return (real or not) over my (short) investing life, wouldn't I have been better off putting my money in the cash funds that guarantee a 5% return?

Of course, I have been dollar cost averaging, so my adjusted cost basis is low, and I have been making a killing lately. The power of DCA is when the market goes up, you make A LOT of money, and if the market goes down, you lower your ACB, helping you when the market goes up again. You can't lose.

Mcwop writes:

For most investors a diversified portfolio is the best way to go. Additionally, making regular investments (dollar cost averaging) is a good way to manage risk. Asset classes to include in a portfolio are TIPS, short term bonds, dividend paying stocks, junk bonds (good yield with upside in recovering economy), selective growth stocks. There are still many growth stocks, which are grossly overvalued.

I would argue risk in stocks right now is moderate, and low if you are selective. Merck is a good example. Currently, Merck's earnings yield (TTM earnings/current share price) is 5.86%. That is a good return on each dollar invested compared to interest rates. Merck's dividend yield is 2.61%, which compares well to cash yields (and many bonds). Merck is selling to a strong growing market. There are many stocks with similar numbers to Merck in solid industries.

Government long bonds do have significant risk. But junk bonds and TIPS appear to have some decent value thus lower risk. In summary, selectivity is key to managing risk right now. Rates may go up, but it is unusual to make a giant jump (2% or more) very quickly.

Bernard Yomtov writes:


If you promise not to respond by calling liberals names, I will do you the favor of pointing out that "lowering your average cost basis" in a stock is nonsense.

If you buy a stock at $50/share, and it drops to $20, you've lost $30/share. Buying more at $20 to lower the average cost of your shares doesn't change that.

Mcwop writes:

You are correct only assuming the stock stays at $20. If it goes up to $30 then your loss narrows to $10 anything above $35 a share is profit, versus having to hit $50 before seeing a gain. DCA is not nonsense. Of course if the stock goes below 20 your loss widens. If you want to really have fun, then explore Value Averaging.

[2 shares one bought at $50 and one at $20 - total invest $70 - average cost is $35 - $70/2 shares]

Value Averaging Link:

Eric Krieg writes:


What Mcwop said.

And I'm seeing the power of a lower ACB now. I think of my investments in terms of the Dow. Thanks to the dips in the market down to 7500, my ACB is like 8500 or so. So when the market hits 9250, I make real money. I'm at the point that every 100 point rise seems to make me $500.

My wife has an IRA that hasn't been DCA'd, and she has nowhere near these % gains.

On a certain level, I understand what you're saying, Bernard. DCA and ACB are in some way a psychological tools to make you feel better about your stock losses. It doesn't change the fact that for the stocks valued below the ACB, you're losing money. In the case of some of my funds it is a substantial amount of money.

My question about investing conservatively when you are young is still out there.

Bernard Yomtov writes:


If the stock goes back to $30 you have lost $20 on your original investment and made $10 on the new one. The arithmetic you cite is correct, but your interpretation is wrong.

Think of it this way. Suppose instead of buying more of the same at $20 you sold at $20, bought two shares of a different stock at $20, and saw your new investment go to $30. Your portfolio value would also be $60 a loss of $10. So the fact that you improved from -$30 to -$10 is a result of your decision to have $40 invested in the original purchase, not of some magical formula.

Bernard Yomtov writes:

"On a certain level, I understand what you're saying, Bernard. DCA and ACB are in some way a psychological tools to make you feel better about your stock losses."

In other words, self-delusion

"It doesn't change the fact that for the stocks valued below the ACB, you're losing money."

You are losing, or more precisely have lost, money on any lot that is below your purchase price, even if you subsequently buy a new lot of the same stock at a lower price and it rises from there.

You can't average consecutive bets like you are trying to do. If you bet on the Yankees yesterday and they lost, and then you bet on them today and they win, you can't claim you broke even on yesterday's bet.

Eric Krieg writes:

Bernard, you're being a little anal, and you're missing the forest for the trees. The goal is a nest egg at retirement. What's the best way to get there? You can't time the market, so DCA is a way to hedge, to make sure that SOME of your investments are well timed. Almost by definition, some of your investments will not be well timed, like just about everything I bought in 1999!

Is it self delusional? No more so than hedging by selling puts.

Like I said, now that there is a little bit of upside in my portfolio, I can see the power of DCA. Yes, it isn't as good as if I had sold when the Dow was at 11,500 and bought when it was at 7500. But how many people do that?

Ted Harlan writes:

A few points are being missed.

First, Eric, you wrote: "Because of the time value of money, isn't one strategy to avoid losses in your early years"

No, no, no. Assuming, say, 10 years of -10% return, 10 years of 10% return, and 10 years of 20% return, (use any numbers, these are easiest), if you are not adding or subtracting any money, the order of returns is irrelevant. If you are periodically adding, you want your worst returns in the early years, so you have the least money at risk in down years and the most money at risk in up years.

Second, DCA does not work! Or, more accurately, if you have the option, you're usually better off if you dump in a lump sum. Evidence:

Ted Harlan writes:

As far as the article's contention, I took the Ibbotson data a few years ago, and, if memory serves, there were 67 seven year periods in the data, and a stock investor would have doubled his money (from memory, here) 60 times, and done breakeven or better 62 times. So, one strategy is to buy a seven year bond ladder, and put 60% of the remaining money in the market. For a fixed-cost investor (one whose expenses aren't closely tied to inflation) , that's a 90% chance of (historical) success. Invest in cash, and, with a historic yield of 5% and historic inflation of 3%, an investor is guaranteed to fail, unless he or she considers 2% a good real return-- Generating $20,000 from a $1mm portfolio.

Eric Krieg writes:

Ted, thanks for the link. If I ever have a lump sum to invest, I'll follow the advice.

I think your link reinforces what Bernard said. A loss is a loss.

But we're talking about saving for retirement, right? DCA is by definition the strategy for us working stiffs with 401ks. Given that contraint, what is the best investment strategy? Even you answer to my question about avoiding losses early is not entirely focused on DCA. Does it still hold?

Eric Krieg writes:

Ted, I answered my own question. I made a spreadsheet, investing $100 per period, starting from nothing. There were 30 periods, broken into 3 phases, each phase consisting of 10 periods with one of your suggested interest rates (10, -10, 20) in all 6 combinations.

The -10, 10, 20 combination resulted in the highest end value. The 20, 10, -10 combination resulted in the lowest end value.

Mcwop writes:

I am still not sure what your point is. If you only look at each investment individually then sure on some lots you have a loss. Most people don't measure investment success in individual lots, they look at overall results.

Using your Yankees example:
-If I only bet on one Yankees game (for a win), then I will either win or lose that bet.
-If I bet on every Yankees game this season, and they win the majority of games, then OVERALL I am ahead even though I lost some individual bets.

DCA is a legitimate way to avoid certain investment mistakes. Like buying the market high, selling low, then repeating that process, and many people do repeat that process. I never claimed it was a magical formula, but rather a risk mangement technique.

Again, I will stress that Value Averaging (different from DCA) is a better strategy to pursue:

Eric Krieg writes:

Mcwop, how would a 401k investor implement VCA?

Part of the appeal of DCA is automatic investing. It's a no brainer. VCA seems like it requires knowledge and analysis.

Ted Harlan writes:

McWop's link can also be found here:

I'll read it now.

Also, anyone here taking the CFA? Just took LIII, and dynamic allocation is well-burned into my mind. Here's something "lifted" from one of the CFA texts. You may be able to find even more.

mcwop writes:

VCA is more complicated to implement. I have been thinking about how to implement in a DC plan account. I suppose you could use a money fund as a staging point. I will get back on this. Could be an inetersting service that providers could create for DC plan investors.

Bernard Yomtov writes:


There is nothing wrong with investing a fixed amount on a regular basis. Indeed, that's most people's only choice - take a part of your income each month, say, and invest it. It's a sensible habit.

What I object to is the notion that DCA in a specific stock is a particularly clever way to do it. One version of this is the idea brokers push, of "averaging down," essentially the same thing. Take the example above. You bought a stock at $50 and it went to $20. Now you're going to buy another share at $20.


You don't reduce what you paid for your initial investment. You just invest another $20 in the company. Might work out, might not. (though the desirability of diversification suggests it's a bad idea).


In other words, in deciding whether to put another $20 in the same company you should consider its future prospects, and the fact that you already have $20 (NOT $50) invested there. You should disregard what you paid for the share you own.

Bernard Yomtov writes:


The best investment strategy is to determine your risk tolerance and allocate your assets between risk-free(treasury securities) and risky (Stocks, corp bonds, etc.) assets accordingly.

In choosing risky assets you should diversify your portfolio among companies and industries, and maybe even countries.

You should strive to minimize transaction costs by not trading any more than absolutely needed and by paying close attention to the expenses of any mutual funds you buy.

If you were talking about a taxable account, rather than a 401K then tax minimization would become important, and again would suggest minimizing trading.

Eric Krieg writes:


Thanks for the clarification. Your explanation parallels what Ted posted from the JFP.

My logic for DCA would be that I like a stock, even though it has decreased in value. I would never invest more in a stock simply because I already own it. I would still have to believe in the long term worth of that company.

I generally only invest in index funds. While I only invest in stock funds, I diversify by owning a S&P 500 Fund, a Russell 2000 fund, a Wilshire fund, and an international fund (non-indexed).

Brian writes:

Ted, thanks for the DCA article. It looks like DCA is being explained to the public the wrong way. If you invest out of employment income, you have no choice but to make periodic investments. (I'm going to overlook the market timing strategy of hoarding cash in order to make a lump sum investment at a later date.) If you have a lump sum presently, DCA should be suggested as a risk reduction strategy that works very well: risk reduction by diversification across time. Since the market's trend is up, it makes sense that lump sum investing beats DCA on return. It makes sense that DCA has a worse return because it is less risky. The Williams,Bacon report you linked shows that the standard deviation of DCA is lower. The lower standard deviation is the benefit of DCA. In retrospect this all looks very obvious so I guess we shouldn't be surprised "little empirical evidence exists to support the DCA strategy". We don't need evidence that DCA maximizes return because DCA is not a return maximizing strategy.

Come to think of it, the only time I used DCA is when I was unemployed and risk averse.

BTW: I did my CFA recently too (2000,01,02).

Ted Harlan writes:


Have you been able to find suitable work, lately?

Bernard Yomtov writes:

Actually, it doesn't really work as risk reduction either, except in the sense that you have less money invested on average.

Suppose you get a lump sum bonus of, say, $20,000 and decide to invest it in the market over two periods. You can either (a) invest it all at once or (b) invest $10K now and another $10K at the start of the second period.

Now suppose the market return differs in the two periods. For example, it's -3% in one and +5% in the other. If you invest the lump sum you don't care about the order. Your return over the two periods is 1.85% regardless, and you have $20,370 at the end of the second period.

But if you follow the DCA strategy it does matter. If the loss period comes first you end up with $20,685. But if it comes second you end up with $19,885. Your average is $20,285, less than the lump sum strategy gives. More risk, less return.

The lump sum strategy works better any time the return in the first period is positive, and I believe similar results will work out for more periods.

The problem is that, far from "diversifying across time," you are doing the opposite. By following DCA you are putting more weight on some months - the later ones - than others. To take a more realistic example, suppose your year-end bonus is $12,000 and you invest it over the following year using DCA, e.g., at $1000/month beginning Jan. 1.

January returns will not influence your returns very much, while December returns will influence them greatly. A bad month in November or December can more easily wipe out a decent year with DCA than if you had just invested it all in January. Of course, if January is bad and December good, the DCA'ers will be ahead, but that adds risk, rather than subtracting it.

The lump-summers don't care whether Jan is good and Dec bad or vice versa, as long as the year is good.
They are the ones "diversifying across time."

Eric Krieg writes:

Two things, Bernard.

One, what is the money doing before it is invested. Does it make any difference to your example?

Two, for your specific analysis, you say "The lump sum strategy works better any time the return in the first period is positive". Isn't that an assumption that makes your point meaningless? Assume that the first period is negative!

Bernard Yomtov writes:


The money is somewhere before it is invested. For these purposes I assumed it was earning no return. Assuming a small return, in a money market fund for example, doesn't change things much. I suppose, though I haven't worked it out, that in that case lump sum is better if the first period return exceeds the money market return.

A positive first period return is not an assumption, it is a statement about the condition under which the lump sum strategy works better. If the first period return is negative then DCA works better, obviously, because you avoid the first period loss.

I don't think that makes my analysis "meaningless." First of all, it sure doesn't prove DCA is better. Second, presumably you're investing in the market because you think that, on average, it's going up. So it's more likely to be up than down in the first period. Finally, what I was trying to show was that DCA does not reduce risk and that it is not diversifying over time. I think those conclusions hold.

Eric Krieg writes:

I revised my spreadsheet to add lump sum to the analysis. My conclusion is that lump sum is clearly less risky.

I added the situation where you invested $3000 ($100 x 30) right at the beginning.

Of the 6 interest rate combinations, 2 of them result in DCA outright beating LS in the final nest egg (the ones where the first period is a negative interest rate). But in only one combination does DCA substantially beat LS (-10, 10, 20).

4 of the combinations result in LS beating DCA. But in all 4, LC beats DCA substantially.

Most frighteningly, the delta between the best final nest egg and the worst final nest egg for DCA was 81%! In my $100 per month example, the best nest egg DCA got me was $23483.95, but the worst was $4460.73!!! Investing $3000 right at the beginning got me $16799.09.

Like Bernard says, the risk of DCA is that you are going to get a negative interest rate in the final periods. When that happens, you get hammered. Even when the negative interest phase is in the middle periods, and positive interest occurs in the final periods, you still get hammered.

But DCA does have the potential to give superior returns WHEN the first phase of your investing life is negative. Good news for Generation X.

You all should make a spreadsheet to prove it to yourselves.

Brian writes:

Ted, I was employed in the software/computer industry working in development while I took the CFA program. I quit recently to go for a degree in economics. My feeling is that this career change will require quite an investment of my time and energy. I know there are analysts and fund managers who have a science/engineering background so I'm probably not taking the most direct approach to maximizing lifetime income. However, for the time being I haven't been dissuaded from jumping through some more hoops if it will make the employment search much easier.

Bernard Yomtov writes:


Good luck.

Brian writes:

Thanks Bernard.

You're right, I shouldn't have said DCA is "diversification over time".

I don't know how the standard deviation was calculated in the Williams,Bacon report but I do nonetheless accept that DCA is less risky for the reason you pointed out: "lower average investment". It's just an axiom that less return implies less risk even if I wasn't able to 'nail' the reason the way you did.

Bernard Yomtov writes:


OK. But for someone looking for less risk by keeping some money in money market funds, for example, it still makes more sense to make the risky investment all at once.

In the $12,000 example, you may only be comfortable putting $6,000 in the market. But do it all at once and leave the other half in cash equivalents. That's a perfectly sensible policy.

Brian writes:

Bernard, I disagree. I think there are (at least) two ways to lower risk. First risk is lowered by having a money market fund allocation. If the time horizon is 2 years, of the $12,000 capital, $6,000 can be put in a money market fund and of the remaining $6,000 the DCA strategy might be used to build equity exposure. Suppose $3,000 is invested after 6 months and another $3,000 after another 6 months. Equity exposure at historical cost will be 0k, 3k, 6k, 6k, 6k at 0,6,12,18,24 months. Thus we have (3k*6)+(6k*12)=90,000 dollar-months of equity exposure. Investing a lump sum creates 144,000 dollar-months of equity exposure. In other words, more risk -- presumably unacceptable. One could argue that if only 90,000 dollar-months of equity exposure was desired, then DCA can be abandonded because this is equivalent to a lump sum investment of $3,750 for 24 months.

Brian writes:

Bernard, I realize now that you wrote that having $6,000 in the market was supposed to be acceptable. So my suggestion that the "true" acceptable level being $3,750 is not really relevant. However, I think the risk reduction of DCA has still been illustrated in my previous post. Assuming no transaction costs, the DCA strategy can be used to mimic the lower risk of a larger money market fund allocation. Since the market trend is up, and that there are transaction costs in reality, DCA is not a good strategy. Implicit in the use of DCA, then, would be the investor's belief that he has market timing skill that outweighs the additional transaction cost.

Bernard Yomtov writes:


It's true that DCA can only be justified on a market-timing basis. But this introduces two further problems:

First, timing the market is virtually impossible.

Second, if you are one of the very rare people who can do it, DCA would not automatically be the right strategy, since it works only if returns are negative early and positive late.

Eric Krieg writes:

>>if you are one of the very rare people who can do it, DCA would not automatically be the right strategy, since it works only if returns are negative early and positive late.

Brian writes:

Bernard, if an individual has market timing skill, DCA lowers risk regardless of the order of returns. In my view, DCA always works. It simply cannot fail to give less expected risk and lower expected returns.

Eric, the answer to your question is because skilled people will hedge and diversify because their skills are profitable on average but not always profitable.

Bernard Yomtov writes:


I don't think so. Suppose your market timing skill tells you prices are exceptionally low on January 2. Doesn't it make sense to invest a lot of your capital then, rather than 1/12 of it? And if prices are high a month later shouldn't you sell rather than investing another 1/12 of your capital?

Like Eric (ta-da!) I think that taking advantage of market-timing skill is quite a different thing than using DCA.

I also think the discussion is moot, since virtually no one can in fact time the market.

Eric Krieg writes:

Brian, what does it mean to reduce risk?

To me, DCA increases risk. It decreases risk in the short term (a short term loss is minimized), but greatly increases it in the long term (if returns are negative in later periods, huge losses incur).

If you are investing for the long term anyway, what's the point of DCA?

Eric Krieg writes:

This is the best I could do, in terms of posting my spreadsheet.

First column is the time period (1 to 30). 100 is your DCA payment (100 per period).

Second column starts with the 3 rates to be used. In the first case, we start at time periods 1 to 10 with -10%, 11 to 20 with 10%, and 20 to 30 with 20%. The rest of the column is the period ending DCA balance.

Third column is lump sum analysis for $3000 over the same time periods over the same interest rates.

Fourth column is the interest rates.

And so on for the 6 possible interest rate combinations.

Again, see that DCA only beats LS 2 of 6 times. LS beats DCA substantially 4 of 6 times.

Brian writes:

Bernard, you are right and I was wrong on market timing. Implicit in the use of DCA is that the investor is risk averse and has no market timing skill.

I would still assert that DCA always works. It cannot fail to produce lower expected risk and lower expected returns.

Eric, I think your view of DCA increasing risk in the long-term is wrong. A LS investor will also have a large amount invested in the long-term.

In answer to your question: what does it mean to reduce risk? There are two parts to the answer. First, is likelihood of not achieving investment objectives. Second is that the likelihood is usually measured by standard deviation or the square root of variance. It can be measured by negative variance below a target threshold as well, which would seem to be more appropriate given that we are concerned about the likelihood of not achieving objectives. However, there are reasons why this is not usually done. Standard deviation is just as good in most cases.

You also asked what's the point of DCA if you are investing for the long-term? The answer is the lower standard deviation. In the long-term the standard deviation of DCA and LS will converge. In my example where the DCA investor was fully invested after 1 year and the time horizon is 2 years, the partially invested period is one-half as large as the time horizon. This is a large fraction so the benefit of DCA is significant. If the long-term is 40 years and the partially invested period is 1 year then there's not much difference between DCA and LS.

Brian writes:

Of course what I wrote earlier about the smaller lump sum investment still applies. A smaller lump sum investment will have lower transaction costs than DCA and also offer lower risk.

Brian writes:

It's interesting that we cannot come up with something better than standard deviation or negative variances to measure risk.

Since we accept volatility as our risk measure some investors will avoid certain asset classes or certain stocks. Yet for most people standard deviation is only a proxy for the true risk of failing to have assets to match liabilities when they are no longer working. In other words, the true risk may have nothing to do with the volatility in the interim. The volatility in the interim may be the result of actions of investors with a shorter time horizon. Because volatility may be due to actions of long-term investors who know something fundamental, we are somewhat willing to accept volatility as a risk measure.

Eric Krieg writes:


Just in my simple example spreadsheet, the deviation from the high to the low on DCA is huge. The best you could do with DCA was over $23K, the worst around $4K, for a delta of $19K.

Now, I would have to do some research to put that in statistical terms, but I would bet that that is a huge variance! And the average payout from DCA is 4000 less than LS, so you don't even make out there.

I just don't see how you can say that it is less risky.

Brian writes:


The standard deviation should be calculated on the portfolio as a whole. This means the allocation to a money market fund has to be considered part of the portfolio.

Bernard Yomtov writes:


For a given average investment, DCA is riskier than the lump sum approach. One way to think about this is in terms of diversfication.

Treat each month as a stock. Then DCA allocates your portfolio 1/78 to Jan, 3/78 to Feb (roughly- since there will be Jan returns), etc. It is clearly less risky to allocate 13/156 to each month - the lump sum approach.

You are correct that risk needs to be calculated for the portfolio as a whole. It is important to understand that, when considering the purchase of an individual stock, one should consider the marginal risk it adds to the portfolio, NOT the risk of the individual stock. This is (partly) the case for diversification - adding a stock adds less risk, as a general proposition, to a diversified portfolio than to an undiversified one.

This is also why, without getting too far into it, you want to measure NOT the total risk of the stock, but its systematic risk - that portion that is not "diversified away" by the rest of the portfolio. Standard deviation doesn't tell you that. Beta is an attempt to do so.

Brian writes:

Bernard, I know about beta. I suggest we simplify the comparison of DCA and LS by saying that for the purpose of discussion the DCA is applied to a portfolio that has two asset classes: money market fund and equity market index with a beta of 1.0.

I'm beginning to think that given the same average investment, the risk is the same between DCA and LS. Of course with DCA risk is concentrated in later months but then it is "diluted" in the earlier months. Wouldn't the two effects cancel?

equity=$3,000 and later equity=$6,000
mmfund=$3,000 and later mmfund=$0

equity=$4,500 and later equity=$4,500
mmfund=$1,500 and later mmfund=$1,500

Looks to me like it may be a tie with respect to risk exposure.

I don't know why people want to compare DCA to lump sum investing. Most of the people that are doing DCA are doing it in a 401K, and don't have the option of a lump sum.
I think it would be much more interesting to compare DCA to (VA, or VCA, or invest%, or Synchrovest. It is my opinion that Synchrovest is the best of the bunch.

Heifer writes:

I have an IRA that has done nothing but lose money, and feed the broker. I don't know what I'm doing, and really could use some advice. My husband and I will be retiring probably in the next 12 years, or maybe never if we don't start saving. I would really love some good sound advice about what to invest in, where to invest it, and how not to get taken to the cleaners by brokers. The IRA we have now is high risk growth. Is that the best thing for us to be in at this point in our lives?
Thanks so much for your consideration. Heifer

Comments for this entry have been closed
Return to top