Nasdaq vs. safe investments

This is a follow up to my recent post on the tech bubble.

In my view, 3-month T-bills are the best asset for estimating nominal risk-free returns that can be earned at various moments in time.  During some periods of US history, it’s possible to earn very large nominal risk-free returns (real plus inflation).  During other periods, nominal risk-free returns are depressed by a variety of factors.  During these challenging periods, other investments may or may not struggle to earn high nominal returns.  Usually they will struggle.  (Consider 1929-45).  Nonetheless, it seems reasonable to compare the performance of other investments to this risk-free benchmark.  If you don’t see why, read my previous post, especially the example of why the 30-year Treasury bonds yielding 15% in 1981 was misleading. (A commenter pointed out that those interest payments could not be reinvested at 15%, but it was still a pretty impressive nominal investment, ex post.)

Unfortunately I don’t know how to find the figures I am looking for, but I’ll present my estimates, and then let commenters who work in finance correct my numbers:

1.  A series of investments in 3-month T-bills from the end of 1999:  Roughly a 35% to 38% total nominal return. (Based on a quick look)

2.  Investing in NASDAQ at the end of 1999 (at 4069), then reinvesting dividends:  Roughly a 121% total return.

3.  Investing in NASDAQ at the absolute peak in March 2000 (at 5048), then reinvesting dividends:  Roughly a 78% total nominal return.

My point is not that those Nasdaq returns are all that impressive (especially if you were unlucky enough to buy at the absolute peak), but rather that these returns were earned in a very challenging investment climate, where risk-free returns were quite low.

It’s not obvious to me that the difference between 121% and 37% isn’t enough to compensate investors for the extra risk from Nasdaq stocks  compared to long term investments into a series of T-bills.  As I mentioned earlier, the excess returns earned by stocks from 1926 to 2000 were probably excessive, in retrospect, even accounting for risk.

Here’s how I’d put it.  In 2002, the consensus view was that 2000 had obviously been a bubble, and that Nasdaq stocks were obviously grossly overpriced at that time.  Given what we know today, a fair-minded observer would say that Nasdaq stocks may have been overpriced in 1999-2000, but it’s no longer a slam dunk that an extreme bubble ever existed.

PS.  I tried to find total returns on Nasdaq, but could only find them for the past 10 years.  During that period, 10,000 invested in Nasdaq, with dividends reinvested, rose to 36646.  But the index itself (ignoring dividends) rose by a smaller amount.  Thus 10,000 invested 10 years ago would now be 32750, if you ignore dividends.  The total return figure of 36,646 is about 11.9% higher than the simple increase in the index, over the past 10 years.  So it seems reasonable that over 18 years the total return figure (including dividends) is about 20% higher.  Thus you’d want to calculate total returns as if Nasdaq were about 9000 today (not 7500) relative to investments back at the end of 1999.  Does that make sense?  That’s how I got my estimate of total returns.

Of course real returns would be lower, but they’d be equally lower for any alternative investments (such as T-bills), so it would be a waste of time to work with real variables, even if the inflation numbers were not unreliable.

PS.  Why don’t our history books call July 2002 to April 2003 a “negative bubble” for Nasdaq?  It ranged from 1100 to 1500.  More importantly, why isn’t there even a word for negative bubbles?

PPS.  Even more importantly, why isn’t there even a word for big plunges in NGDP growth?

Happy 4th of July to my American readers.




34 Responses to “Nasdaq vs. safe investments”

  1. Gravatar of Ironman Ironman
    4. July 2018 at 17:31

    It’s not NASDAQ, but I can put the entire history of the S&P 500 and its predecessor indices and components at your fingertips, including the rates of return between any two calendar months since January 1871, both with and without the effects of inflation and with and without the reinvestment of dividends.

    We update the tool monthly, after the previous month’s inflation data is available. It is current through May 2018 at this writing.

    As for terminology, there aren’t very many examples of what might legitimately be called a negative bubble – most such disruptive events in the market have been very short lived, which have been called flash crashes in recent years.

  2. Gravatar of dtoh dtoh
    4. July 2018 at 17:34

    Three quick points.

    1. A lot of people were stripping the long bond in the early 80S, selling the coupons and just keeping the corpus to create an instrument with a 30 year duration (i.e. effectively locking in the rate for the entire life of the instrument.

    2. You need to look at the after tax return (difference in timing and rates of taxes) for Tbills versus stocks.

    3. From a really long term perspective, capital gains above the risk free return can be completely shielded from estate tax so this further alters the effective after tax return.

  3. Gravatar of Michael Sandifer Michael Sandifer
    4. July 2018 at 18:03


    I would point out that the S&P 500 crashed right along with the Nasdaq in 2000. Why would that be the case if there was a tech bubble? The correlation between GDP and the S&P 500 level is over.97 since 1950.

    Also, given that the whole stock market is worth only about 5% of the value of future expected GDP, is it more likely a GDP shock caused the Nasdaq and S&P 500 to crash, or the reverse?

    The null hypotheis concerning the cause of stock crashes should be GDP shocks. This being the case, one should also give weight to an avoided recession for the ’87 crash, but I digress.

    Markets have to be inefficient enough to provide enough arbitrage to keep them relatively efficient, so it doesn’t seem I find markets quite as efficient as you do, but I do think they’re more consistently efficient than you do.

  4. Gravatar of Matthew Waters Matthew Waters
    4. July 2018 at 18:55

    The most direct data is from historic Treasury auctions. The absolute safest investment is having funds directly with the government in “book entry” form. There is no intermediary broker. Then you buy 13 week T-bills at auction. You roll over with non-competitive bids.

    Treasury Direct has auction data going back to at least 1997, but it’s not in a convenient form. You have to download each one individually. For auctions before 2008, only PDF’s of the result are available.

    For example, the 1/20/2000 auction gave non-competitive bidders 91-day T-bills at 98.648.

    You can chain to the next auction after redemption on 4/20/2000, which was 4/24/2000. Treasury Direct currently allows roll-overs of T-bills up to 2 years. The only action would be required when that 2 years is up, to schedule a new 2 year roll-over. There are no fees either.

    That’s the very pedantic way to look at 3-month T-bill yields. I downloaded the FRED data set and took rolling 90 day compounded calculations.

    Here are the calculations from 3000, 4000 and 5048 peak. This is annualized returns for NASDAQ vs T-bills:

    – 11/3/1999 – 4.98% vs 1.86%
    – 12/29/1999 – 3.4% vs 1.88%
    – 3/10/2000 – 2.19% vs 1.9%

    Out of curiosity, I added in QQQ since you can’t exactly track the NASDAQ index weights and selection. Strangely, the QQQ outperformed the NASDAQ. I’m not sure how to explain this.

    The 1999-2000 still augers against the EMH. The outperformance here isn’t particularly good. Equities have historically outperformed T-bills by 6.4 percentage points (9.7% to 3.3%), going back to 1927.

    Also, the stock market added 33% in about two months. Then it added 20% in four months. These movements definitely were not due to some unanticipated news which affected honest fundamental valuation.

    In the 1999-2017 period, even very naive buying on fundamentals outperformed either stocks or bonds alone. T-bills yielded 5.5% in 11/1999 – 3/2000 period, while stocks earnings yield went from 4% to 3%. Earnings would have to increase at rates far higher than historical averages to justify the 11/1999 valuation (much less the 3/2000 valuation).

    On 1/1/2010, earnings yield was 4.8% with past 12 months earnings. T-bills yielded 0.1%. In many ways, the prolonged “negative bubble” in stocks for 2009-2010 is more inexplicable than the short tech bubble for 11/1999-3/2000.

  5. Gravatar of Matthew Waters Matthew Waters
    4. July 2018 at 19:02

    “I would point out that the S&P 500 crashed right along with the Nasdaq in 2000. Why would that be the case if there was a tech bubble? The correlation between GDP and the S&P 500 level is over.97 since 1950.”

    The S&P 500 includes NASDAQ stocks. The 500 stocks are also weighted by market capitalization. Stocks like Cisco and Microsoft reached huge valuations in 1999/2000, which weighted the S&P 500 more in technology.

  6. Gravatar of Michael Sandifer Michael Sandifer
    4. July 2018 at 19:32

    Matthew Waters,

    So, it was only tech stocks that crashed in 2000? I think you better check the record.

  7. Gravatar of ssumner ssumner
    4. July 2018 at 22:09

    Ironman, Thanks for the info. I’d say 2002 was far more a negative bubble than 2000 was a positive bubble. The 1987 crash also produced a negative bubble.

    dtoh, All good points, and all support my argument. I had forgotten those zero coupon bonds from 1981, but used to teach that example to my students back in the 1980s. I seem to recall that at the period of highest interest rates you could buy $1,000,000 for about $20,000, if you were willing to wait 30 years!

    Michael, Bubble supporters would claim the broader indices were also a bubble, which I find hard to believe.

  8. Gravatar of Matthew Waters Matthew Waters
    4. July 2018 at 22:10

    I didn’t say only Tech stocks crashed. Tech stocks did crash much worse though. NASDAQ lost 70% while S&P 500 lost 42%. Also, the NASDAQ reinvests dividends while S&P 500 does not, furthering how NASDAQ crashed more than the S&P.

    A general stock market overvaluation does not look much better anyway for the EMH. NGDP did not change enough 2001-2002 to warrant a ~40% drop generally.

  9. Gravatar of Michael Sandifer Michael Sandifer
    4. July 2018 at 23:21

    Matthew Waters,

    Exxon fell about 32%, Pfizer fell about 36%, and Citigroup fell about 50%. So, you’re claiming there was a general stock market bubble? And how is this consistent with the .97 correlation between GDP and the S&P 500?

    Show me your calculation, because the crashes for both indexes are consistent with trend GDP growth path that fell 2.8%. The unemployment increase is also consistent with 2.8%.

  10. Gravatar of Benjamin Cole Benjamin Cole
    5. July 2018 at 05:30

    Ditto on reverse bubbles. Invent some buzzwords.

    A money vacuum. Low tide. Irrational gloom.

    It may be that “bubbles” burst quickly, but investing opportunities tend to play out over a few years.

  11. Gravatar of Majromax Majromax
    5. July 2018 at 05:32

    > In my view, 3-month T-bills are the best asset for estimating nominal risk-free returns that can be earned at various moments in time.

    I don’t think this is an appropriate portfolio to compare against a NASDAQ portfolio. The portfolios have very different durations, so the different total returns reflect different sensitivities to the change in the risk-free rate.

    Your NASDAQ scenario is apparently “buy the index in 2000, hold until 2018, reinvest all dividends, and sell.” Your bond portfolio should be similar. I’m not a finance expert so my prior would be to suggest a similar sort of bond-weighting, by investing in a variety of bond terms. If I had to pick just one term, it would probably be the 10-year just because it’s a common benchmark for “long-term” investments.

    If we allow ourselves to look at the NASDAQ index in 2000 to better-inform our “bond-equivalent” comparison, then it only strengthens the conclusion we should look at long-term bonds. As you note, NASDAQ valuations were based on expectations of both company success and strong market growth, but the current price brought forward a lot of the “long tail” of returns to the present. That’s an argument that we should be looking at bonds that did similar things.

  12. Gravatar of Majromax Majromax
    5. July 2018 at 05:59

    After some further thought, I think I can simplify my immediately-previous comment:

    Your comparison portfolio is inefficient if the EMH is true. Your scenario is buy, hold-for-18-years, then sell. By the EMH, you should have been looking for a zero-coupon, 18-year bond. Since that doesn’t exist you’d have to synthesize it, but that synthesis looks nothing like reinvested 30-day treasuries.

  13. Gravatar of Ironman Ironman
    5. July 2018 at 06:24

    Michael, Matthew – Some more background on the dot com bubble:

  14. Gravatar of bill bill
    5. July 2018 at 08:16

    People will go to great trouble to justify their bubble predictions, though only a few actually short the asset in a bubble. At the end of any investment period, something will have had the worst performance. Calling that proof of a bubble is just hindsight bias. Going forward, let’s see some real time bubble predictions that include a real trading position. Let’s see how those do.

  15. Gravatar of Justin D Justin D
    5. July 2018 at 08:36

    I think some of the best analysis on equity markets comes from the blog Philosophical Economics.

    Historical returns look excessive to us today, but make sense considering three important aspects of market history: 1) The rise of mutual funds/indexing 2) Pension funds plowing into equities and 3) Lack of good long term data on stocks several generations ago (and the relative impact on investor confidence)

  16. Gravatar of sean sean
    5. July 2018 at 10:39

    3 month-tbill isn’t the right comparison. I think you need to compare to a 10 year treasury or maybe a 20 year to calculate the risks free rate.

    You also had something happen over those 18 years….fed losing control of ngdp causing tbills to be depressed for most of the time period. 10 year treasuries would not have expected that and thus offered a much higher return than tbills. long term treasuries would have factored in expected future rates while tbills give you realized.

  17. Gravatar of Scott Sumner Scott Sumner
    5. July 2018 at 14:25

    Majromax and Sean, I already explained why I disagree with using long term bonds, in my previous post.

  18. Gravatar of Benjamin Cole Benjamin Cole
    5. July 2018 at 20:05

    Bubbles vs. Negative Bubbles:

    After profound contemplation, I think it goes like this:

    Obviously, with the benefit of 20/20 hindsight, there are times when any particular market is heftily above or below future values. Gold was in a “bubble” in 1980 and then again in 2012, for example. (I love gold “bubble” stories, as most bubble-mongers never seem to think gold is in a bubble).

    But for whatever reason, rapid declines in market values are part of the historical record, such as the stock market 1929, the post-2000 Nasdaq bust, or US property values after 2008.

    In contrast, even strong bull markets tend to play out over years. The recovery from the 2008 property debacle has played out over 10 years.

    This gives the general public, and perhaps even astute observers, the impression that there are only “bubbles” that pop, and not reverse bubbles, and in one sense they are right. There are rapid market-wide and large declines in value, from time to time, but no equivalent on the upside,

    Still, I think EMH rules the roost. No one seems to correctly call bubbles, or reverse bubbles, or beat the market in general.

  19. Gravatar of Philo Philo
    6. July 2018 at 02:53

    “{W]hy isn’t there even a word for big plunges in NGDP growth?” Because very few economists share your belief in the central importance of NGDP (expectations) for macroeconomics. But one could adapt the vocabulary used for prices, and, say, talk about “N-crashes,” along with “N-surges” and “N-bubbles.” There remains the question why this vocabulary doesn’t contain ‘reverse-bubbles’, so that we have to make up the term.

  20. Gravatar of Majromax Majromax
    6. July 2018 at 03:37


    You explained why you disagree with using long-term bonds, but with all due respect it’s a poor argument for two reasons:

    *) First, you’re missing the term premium. Even if NGDP growth exactly met expectations and the real interest rate never changed, a long-term bond would be better-yielding because of time preference. Your comparison portfolio gives you the implied option to take gains at any three-month interval along the way, but you make no use of this option with buy-and-hold.

    *) Second, you’re offering a post-facto justification for the prices, as “prices were rational after controlling for NGDP falling short of expectations.” This is the equivalent of bubble-mongering, as “housing prices are only high because they’re being supported by .”

    The NGDP expectations were obviously (implicitly) built into the price, so the price could be irrational if those expectations were irrational. An 18-year zero-coupon bond would have zero correlation to NGDP surprises, of course.

    If you insist on controlling for NGDP expectations, then your comparison portfolio needs to behave in a similar way towards NGDP surprises. The S&P500 index is one portfolio that we would expect to be positively correlated with NGDP surprises, and using this as a control still doesn’t look good for you. Based on the handy link provided by Ironman above, the annualized nominal growth from 1999-12 to 2018-05 was 5.49%, for a total return of 167.6% over the period versus your estimate of 121% for the NASDAQ portfolio.

  21. Gravatar of Majromax Majromax
    6. July 2018 at 03:39

    Addendum to previous:

    The quote at the end of my third paragraph should read “… only supported by [foreign money / reckless loans / low interest rates].” I had originally used angle-brackets, but the blog software ate it under the assumption it was an HTML tag.

  22. Gravatar of Scott Sumner Scott Sumner
    6. July 2018 at 07:08

    Majromax, You said:

    “First, you’re missing the term premium. Even if NGDP growth exactly met expectations and the real interest rate never changed, a long-term bond would be better-yielding because of time preference.”

    That’s so obvious I didn’t think I needed to point it out. And no, I did not ignore that fact in my analysis. I clearly said the expected return on Nasdaq should be much higher than on T-bills.

    As far as your second point, when thinking about whether an asset is overvalued, you’d want to see how it did, given the general environment for returns on risk free assets. That’s the T-bill yield. Using T-bills means you don’t have to adjust for unexpected changes in NGDP growth, or ANYTHING ELSE, the T-bill market does that automatically. It’s the right risk free benchmark. Then you add a premium for how much Nasdaq needed to outperform due to risk maturity, etc. Maybe it didn’t outperform enough–that’s the best argument against me.

    T-bonds are not helpful, as they clearly performed far better than expected in 2000, in a relative sense. Bonds were yielding 6%, and the short rate quickly dropped to 1%, and mostly stayed very low throughout the past 18 years. So it’s not a good benchmark. T-bond returns tell you about much more than Nasdaq, they tell you about how T-bonds were GREAT investments, ex post.

  23. Gravatar of Daniel Morgan Daniel Morgan
    6. July 2018 at 07:58

    It’s regularly accepted in business valuation circles that returns were excessively high in the late 20th Century due to a one time increase in price to earning ratios. This effect is often removed when estimating the “equity risk premium.”

  24. Gravatar of H_WASSHOI (Maekawa Miku-nyan lover) H_WASSHOI (Maekawa Miku-nyan lover)
    6. July 2018 at 08:59

    I think it should be from some Japanese word. Mizore or hisame? (Freezing rain) > a word for big plunges in NGDP growth

  25. Gravatar of H_WASSHOI (Maekawa Miku-nyan lover) H_WASSHOI (Maekawa Miku-nyan lover)
    6. July 2018 at 09:01

    well, Nadare is better (Avalanche)

  26. Gravatar of Benjamin Cole Benjamin Cole
    6. July 2018 at 18:03

    OT but in the ballpark.

    Gagnon says QE lowers interest rates 40-50 basis points.

    Big whoop?

    Well, better than nothing.

  27. Gravatar of bill bill
    7. July 2018 at 12:05

    I know how to lower IOR by 40 to 50 bps. Just do it. I wonder what would have happened with all that QE if the Fed had lowered IOR to minus 50 or minus 75 right after Lehman failed instead of raising it to more than 1.25%. It was almost Christmas before they lowered it to 25 bps – still higher than it had been from 1913 to Oct. 8th, 2008. For these reasons, we’ll never know what QE really could do since it was done simultaneously with the intro of IOR.
    Sorry for the rant.

  28. Gravatar of bill bill
    7. July 2018 at 13:55

    Also, a better sign that QE was “working” would have been rising interest rates. Working should have meant rising NGDP expectations and interest rates should have just fallen wherever the market took them.

  29. Gravatar of Benjamin Cole Benjamin Cole
    7. July 2018 at 20:54


    Good rant.

    As I say, no one is ever wrong in macroeconomics.

    Does QE work? Of course it does, and of course it is inert.

    Genuflect to the totem of your choice.

  30. Gravatar of Mike Rulle Mike Rulle
    8. July 2018 at 09:30

    It is better to look at NDX, or Nasdaq 100, rather than Nasdaq 5000. Futures are the easiest way to look at total return. The NDX 100 (once you get past the top 40 or 50 it is irrelevant) futures contract is first quoted on 6/23/99. The annualized total return (assuming target Fed Funds as risk free rate) to July 5, 2018 is 6.4%. Worst drawdown was 83%. The annualized return from 1/1/2009 to July 5, 2018 was about 21%. The return from 6/23/99 to 12/31/2012 was zero percent. The Sharpe ratio for the entire time frame was .17. For fun, the rolling two year futures contract outperformed the NDX until 2015 at 4 times the Sharpe ratio. Not sure what I am getting at. But it’s best to stick with a diversified portfolio and not try to guess the twists and turns. Even better is to use intertemporal risk parity approach. NDX at constant volatility of 20 has four times the Sharpe ratio, 2.5 times the annual return, and .5 the Drawdown of simply rolling the NDX contract. Bubbles are hard to determine. I do think they exist, but they are rare and they exist when you can predict them using the implication of their value.

    My real point was to give you more data to work with.

  31. Gravatar of Jacob Egner Jacob Egner
    8. July 2018 at 19:43

    Scott, you said:

    > the excess returns earned by stocks from 1926 to 2000 were probably excessive, in retrospect, even accounting for risk.

    As in we should used to a CAPE that is higher than the historical average?

    Also, thanks for the posts on bubbles; you have swayed me some on the matter.

  32. Gravatar of ssumner ssumner
    12. July 2018 at 10:52

    Daniel, That sounds right.

    Bill, I agree.

    Thanks Mike.

    Jacob, Yes, a higher CAPE.

  33. Gravatar of SG SG
    16. July 2018 at 06:40

    So I asked my friends on social media what they would suggest as a buzzword for a negative bubble and my favorite was “dimple”.

  34. Gravatar of ADifferentAnonymous ADifferentAnonymous
    27. July 2018 at 08:19

    Isn’t a negative bubble traditionally called a ‘panic’?

Leave a Reply