Archive for the Category Efficient markets hypothesis


I don’t buy the Economist for investment tips

I view The Economist as the best magazine in the world, which is why I like to pick on them so much.  I recently came across a couple examples of faith-based reasoning at the Economist.  Here’s a piece on the recent recovery in Japan:

JAPAN’S economy has been so sickly for so long that many have stopped looking for signs of recovery. And yet, on close examination, they are there. Years of massive fiscal and monetary stimulus seem to be having some effect. Unemployment is below 3%—the lowest rate in 23 years—and wages are rising, at least for casual workers. Prices are creeping up, too, albeit by much less than the Bank of Japan’s 2% inflation target.

Yikes!  Massive fiscal stimulus?  In fact, Abenomics has involved a sharp contraction in fiscal policy, mostly due to a large increase in their national sales tax.  Budget deficits are shrinking dramatically.  Yes, fiscal policy was expansionary in the 1990s and early 2000s, but that corresponded to perhaps the worst 19 year performance in aggregate demand ever seen in a developed economy, with NGDP actually falling between 1993 and 2012.  It’s only since the beginning of 2013 that Japanese NGDP has shown signs of life:

Then I came across this headline on Bitcoin:

Manias, panics and Initial Coin Offerings

Crypto-coin mania illustrates the crazy and not-so-crazy sides of bubbles

In fact, it would be hard to find a more perfect refutation of the bubble hypothesis than Bitcoin. And yet somehow The Economist sees Bitcoin as a good example of a bubble.

Recall that back in 2012 when Bitcoin was trading at $12, the Economist was already calling it a bubble:

These curious capabilities make Bitcoins a combination of a commodity and a fiat currency (creating the coins is referred to as “mining” and they have value only because people accept them). But boosters inflated a Bitcoin bubble. Shortly after the currency launched, articles spread around the internet arguing that Bitcoins would protect wealth from hyperinflation and that early adopters would make a fortune. The dollar price of a Bitcoin currency unit climbed from a few cents in 2010 to a peak of nearly $30 in June 2011 (see chart), according to data compiled by Mt Gox, a popular online Bitcoin exchange. Inevitably, the currency then crashed back down, bottoming out at $2 in November 2011.

Inevitably?  Do the writers at The Economist have no sense of shame?  It’s bad enough that they prevented me from becoming filthy rich by purchasing bitcoin back in 2012, but after being spectacularly wrong about it being a bubble in 2012, they continue to make the same predictions over an over again, year after year.  Bitcoin could fall 99% tomorrow and the Economist would still be completely wrong about it being a bubble.  Please, I beg you, just stop trying to predict asset prices.  I don’t buy the Economist for investment tips.

As each day goes by the four anti-EMH arguments from 2009 (when I started blogging) look weaker and weaker.  NASDAQ 5000?  It just soared past 6700.  (Roughly 5000 in real terms)  Housing prices?  They’re soaring again.  Hedge funds and college endowments outperform?  Not any more.  Bitcoin is a bubble at $30?  Where can I buy some at that price.

But of course none of this will matter.  People don’t believe in fiscal stimulus and bubbles because of the facts, as the evidence strongly refutes these theories.  Rather it seems like soaring prices are a bubble, and it seems like big government spending programs should boost the economy.  Thus people will continue to believe these myths no matter how much evidence piles up against.

PS.  And it’s even worse.  When Bitcoin prices finally plunge (and they will at some point, as all highly volatile asset prices do) then the bubbleheads will think they were right all along, even as they’ve been wrong all along.

It’s hopeless.

One by one, the anti-EMH arguments collapse

When I started blogging in early 2009, the anti-EMH forces were riding high.  The previous decade had seen tech and housing “bubbles”, there were studies showing that hedge founds and elite college endowments outperformed the broader markets, there was the absurdly high price of Bitcoins, and there were academic studies finding market “anomalies”.  In the eight years since, all of these arguments have either mostly or entirely collapsed.

1. Remember those people who told you not to buy Bitcoin at $30 because it was a wildly inflated bubble?  They stopped you from becoming filthy rich, as it’s now at over $2400.  Yes, it could collapse by 90%, but it would still be 8 times higher than when anti-EMH pundits were calling it a bubble.  Alternatively, if 98% of Bitcoin-type investments fell in value to zero, it would still be a good idea to invest in all of them as long as one in 50 went from $30 to $2400.  Yes, the anti-EMH argument is that weak—even if they were right 98% of the time on bubbles bursting, they’d be wrong in their broader argument that markets are not efficient.

2.  Hedge funds have done poorly since I started blogging (Buffett won his bet that they would not continue outperforming the S&P500.)  College endowments haven’t even been able to beat index funds.

3.  House prices are back up to the peak, and NASDAQ is almost 24% above the 2000 peak.  In fairness, in both cases the real price remains below peak levels.  But there is no longer a serious argument that these markets were “obviously” ridiculously overvalued, especially given that so many other foreign housing markets are now far above 2006 levels. Back in 2002, when NASDAQ was at roughly 1100, people were claiming that 5000, and even 4000, had been an absurdly overvalued level.  Now it’s over 6200.  And yet most of these pundits seemed to have no problem with a NASDAQ of 1120 in October 2002.  Don’t let anti-EMH people tell you how to invest.

4.  Alex Tabarrok linked to a recent academic study by Kewei Hou, Chen Xue and Lu Zhang, which looked at a large number of market “anomaly” studies, and found that the results were heavily influenced by data mining (aka p-hacking):

The anomalies literature is infested with widespread p-hacking. We replicate the entire anomalies literature in finance and accounting by compiling a largest-to-date data library that contains 447 anomaly variables. With microcaps alleviated via New York Stock Exchange breakpoints and value-weighted returns, 286 anomalies (64%) including 95 out of 102 liquidity variables (93%) are insignificant at the conventional 5% level. Imposing the cutoff t-value of three raises the number of insignificance to 380 (85%). Even for the 161 significant anomalies, their magnitudes are often much lower than originally reported. Out of the 161, the q-factor model leaves 115 alphas insignificant (150 with t < 3). In all, capital markets are more efficient than previously recognized.

In retrospect, 2009 was “peak anti-EMH”, and it’s been all downhill from there.

PS.  Just when you think the GOP and its fake news co-conspirators can’t get any further down into the gutter, they hit a new low.  Remember the saying; “The fish rots from the head down”?  A pro-Trump Republican assaulted a reporter in Montana, while running for Congress.  There was a Fox News reporter standing three feet away.  But during the next few hours, Fox News reported the candidate’s pathetic lie (notice how these bullies don’t even have the courage to stand up for their beliefs?), but failed to report its own reporter’s eyewitness account–which contradicted the candidate.  Perhaps Fox is spending too much time trying to find the real killer of Seth Rich.

And of course GOP politicians just run and hide when asked to comment.

Update:  The editors of Bloomberg want the Bank of Canada to stop trying to stabilize the economy and shift over to trying to control asset prices:

Canada Must Deflate Its Housing Bubble

Asset prices are always wrong, in retrospect

One argument against market efficiency is that asset prices are subject to speculative bubbles, where prices rise (or perhaps fall) more than can just justified by fundamentals. Today I’ll discuss why this theory is so hard to evaluate.

During my lifetime, I’ve seen three major asset price movements that looked, in retrospect, rather irrational:

1. The 1987 stock market boom and crash.

2. The late 1990s NASDAQ boom and crash

3. The housing “bubble” of 2006.

During the first 8 months of 1987, the Dow soared by about 40%, and then fell by a roughly equal amount late in the year, with a notable 22% decline on a single day (a record, by far).

After the crash, the pre-crash prices were widely viewed as a bubble, and not justified by fundamentals. Today those prices seem quite reasonable, and if anything it’s the post-crash levels that seem too low. On the other hand, the size of the price change, particularly the 22% decline in a single day, is hard to square with fundamental theories of asset prices, where stocks move only on new information. Nothing occurred on October 19, 1987, that would justify such a large price move. So in one of two respects, 1987 still looks bad for the efficient markets theory.

2. In the late 1990s, the tech-dominated NASDAQ soared from below 1000 to a peak of 5048 in March 2000. Then it fell 1114 in October 2002. During the 21st century, the March 2000 levels have been almost universally viewed as an insane bubble, whereas the October 2002 levels have received little comment. And yet a good case can be made that it is the October 2002 levels that are far out of line with fundamentals, at least based on today’s NASDAQ (over 5800 as I write this post.)

In fairness, the (PCE) price level is up 35% since March 2000, so the real NASDAQ is still considerably lower than at the 2000 peak. Still, even in real terms the NASDAQ is higher than it was just a couple months before or after the March 2000 peak. It should also be noted that the dividend yield on NASDAQ was lower than the real interest rate back then, so investors considerably over-estimated the returns they could expect from buying tech stocks at those lofty levels. I’m not claiming that the March 2000 prices look correct, in retrospect. But then if you go back in time and cherry pick the day when valuations were at their absolute peak, then of course it’s not going to look like the optimal time to buy that asset.

I’d also point to the 1114 low in October 2002, which looks, in retrospect, even more “wrong”. Obviously if I wrote this post in 2002, I would have reached very different conclusions about the 2000 “bubble”. The point here is that in retrospect, previous asset prices will almost always look “wrong”. And since we never reach the end of time, we don’t ever get a definitive reading on what asset price level would have been correct, at any given point in time.

3. As far as the 2006 housing bubble, two subsequent events have cast doubt on whether the prices actually were irrationally high in 2006. First, housing prices in many other countries with similar price run-ups, such as Canada, Australia, Britain and New Zealand, did not crash, and indeed are as high as in 2006, or even higher, even in real terms. (Ireland did crash like the US, to round out the English speaking countries.) Second, home prices in many coastal cities like New York, Boston and coastal California have soared back up to “bubble levels”, and even higher. Many central US regions like Texas never saw a price bubble. Yes, some cities never did recover, but Kevin Erdmann has many posts that provide further evidence that the so-called bubble was not as irrational as it now seems, given what people knew at the time.

This post is also very provisional. In two years, asset prices might be much higher than today and the idea of a 2000 NASDAQ bubble and a 2006 housing bubble may be almost completely discredited. (Just as the 1987 stock bubble was later discredited.) Or asset prices may fall sharply and this post may seem mistaken. That’s why I always try to take a pragmatic approach to bubble theory. What’s in it for me? How do bubble theories help me to live my life more effectively as an investor, as an academic, and as a voter? So far I don’t see much use for bubble theories, but I’ll keep an open mind.

Off topic:  I watched part of Trump’s press conference this afternoon, which reminded me the the “strawberries” scene in The Caine Mutiny.  I’ve got news for Trump—this is your honeymoon period.  It will get far worse.  If Trump’s already showing signs of being mentally unstable after a few minor flare-ups, what’s it going to be like when his administration gets into serious trouble?  Anyone who watched the press conference and still doesn’t understand why I think Trump is a spoiled, immature brat, then, well then I have nothing to say to you.

(Not that I could care less, but the rest of the world is laughing at us.  We elected a right-wing version of Chavez, or if you prefer a Duterte or a Berlusconi.  I feel bad for the reporters who had to sit through that clown show.  And thank God that there are a few GOP senators who are willing to tell the truth.)

PS.  I also recommend this FT article on bubbles:

The background history to these booms confirmed what historians of bubbles had already shown: that they always have at least some backing from the fundamentals. Bubbles may end up being irrationally expensive, but they are not stupid. They arrive when an exciting new development — canals, railways, the internet — creates confusion over the future value they will create. As he puts it, “there was at least some method to the madness of investors”.

He found 72 cases of a market doubling in a year. In the following year, six doubled again, and three halved, giving back all their gains: Argentina in 1977, Austria in 1924 and Poland in 1994.

For doubling in three years, he found 460 examples. In the following five years, 10.4 per cent of them halved. The possibility of halving in any three-year period, regardless of what had come before, was lower than this but not dramatically so: 6 per cent.

On this basis, arguments made by many (including me) that central banks should concentrate more on pricking bubbles before they get too big begin to look threadbare.

My critics in 2012: “See Sumner, housing prices were way too high in 2006”

Real house prices are still well below the peak, but nominal prices hit a record high in September:


The 2006 period may have been a bubble, but as of today is seems far less irrational than it seemed in 2012.  (This recovery also makes Kevin Erdmann’s arguments look even stronger.)

The world’s full of uncertainty and markets are volatile.  Get used to it.

PS.  UK house prices (green) took a dip after 2006, but are now well above 2006 levels (and equal to 2006 prices in real terms).  Australia (blue) and Canada (pink) are much higher in both real and nominal terms.

What goes up must eventually . . . go up even more!


So the US is down in real terms, the UK is even, and Canada and Australia are up in real terms.  Isn’t that sort of consistent with the EMH?

Sure, those prices will dip at some point in the future.  That’s what efficient markets do, they go up and down.


Rational expectations and betting markets vs. polls and models

I don’t have strong views on who’s going to win the election.  Clinton seems more likely to win, but by how much?  Earlier today I defended 538, which gives Trump a (fairly good) 35% chance.  That’s more than the betting markets.  Now I’ll present the opposite argument, and then try to tie it in to monetary policy, which is what this stupid blog is supposed be about—right?

To see the argument for Trump having a good chance, check out this map, from RealClearPolitics.  It shows all states colored, based on poll averages, no matter how narrow the margin:

screen-shot-2016-11-05-at-8-23-13-pmThat seems like a pretty decent margin for Hillary–so why do I say it’s good news for Trump.  Because (blue) Florida is really close, has 29 electoral votes, and would put Trump up to 270 if it flipped.  That brings back memories of 2000, when a very close Florida put Bush up to 271.  And just as in that case, if Trump narrowly wins with 270 electoral votes, he’ll likely lose the popular vote—too many wasted Hillary votes in California and New York.  My hypothetical is identical to 2000, except Iowa, Virginia and Colorado flip.

I think this path to victory is semi-plausible, which is why Nate Silver’s 538 gives Trump a 35% chance.  But if he loses any of these states, no matter how small, then he falls short unless he can pick up another state.  And that’s where things get tougher. He’d need Pennsylvania, Colorado, Michigan or some place like that.  Those are tougher than Florida.  (I’m from Wisconsin, and have confidence in my fellow cheeseheads.)

To me, the most interesting event in the past 24 hours is the sharp fall in Trump contracts in the betting markets, to 22%.  AFAIK, that’s 10 points down from a couple days ago.  What’s going on?  The polls have not changed dramatically.  I’m not sure, but I suspect Nevada:

And now that Nevada early voting has come to a close, Ralston isn’t mincing words about how he sees Trump’s prospects. “Trump is dead,” Ralston tweeted Saturday. He elaborated on his blog that from the early voting numbers so far, the GOP nominee would need a “miracle” to win Nevada at this point.

The polls have tended to put Nevada as a pure toss-up state, and a few recent ones have even shown Trump ahead there. Accordingly, it hasn’t generally been considered part of Clinton’s swing state “firewall.”

But Nevada is a famously difficult state for national pollsters to get right. Its population is transient and many work at night. Furthermore, its population is over one-quarter Hispanic, and it’s often challenging for English-language polls to sample Hispanic voters accurately.

Does anyone know where Nevada was two days ago in the betting markets?

In the past, polls in Nevada have been less accurate than in other states:

So in previous years, analysts like Ralston have found success in reading tea leaves from Nevada’s early voting numbers instead. And all week, Ralston has been warning of danger signs for Trump. The partisan and geographic breakdown of early voting turnout has looked similar to 2012, when Barack Obama won the state by 6 and a half points. But the final day of early voting Friday was, Ralston writes, “cataclysmic” for Republicans.

.  .  .

Though the statewide early voting numbers aren’t yet finalized, Ralston estimates that registered Democrats will have a 6 point lead on registered Republicans among early voters. Since registered partisans tend to overwhelmingly vote for their own party, Trump probably either needs to dominate among early voters associated with neither party or else make up the gap on election day.

But how important is the early turnout?  This important:

But ballots equivalent to well over two-thirds of the total 2012 turnout in Nevada have already been cast. So if Trump has indeed fallen significantly behind in the early vote, it will be very challenging for him to catch up.

Experts warn that early vote totals can be misleading.  But they are not meaningless.  At a minimum, we actually have some HARD DATA.  We know that Dems in Nevada are turning out in large numbers.  That doesn’t mean Hillary will win the state; I could image Trump doing well among working class Dems.  But it tells us something–maybe that GOTV is working.  It’s no longer just polls.  My hunch is that 538 is ignoring this data, because early voting isn’t always reliable, but the betting markets are concluding that Hillary is very likely to win Nevada. Indeed while RCP (i.e. polls) give Nevada narrowly to Trump, the betting markets show a very strong 77% for Hillary.  I know of no other state with such a huge gap.

Rational expectations say that investors look at everything when making a forecast, including decisions on asset valuation.  Thus while experts might have said (in the weeks after Brexit) that “it’s too soon to say how it will impact the UK economy” the markets sniffed out that the UK was holding up better than expected, and UK stocks rallied quite a while ago.  Economists wait for data like GDP and employment, which comes out with a lag.  As another example, I think the markets sniffed out that slow RGDP growth and “lower for longer” interest rates were the new normal long before the Fed figured that out.  The Fed relies on models where 3% trend RGDP growth and 5% T-bill yields are “normal”, and it took them a long time before they downgraded those forecasts.

So while 538 is a great site, and I love their thoughtful statistical analysis, you could argue that it’s more driven by “models” than a pure rational expectations forecast would imply.  That is, it might not put enough weight on tea leaves like the Nevada early vote, because of its unreliability in other contexts.

I don’t want to make too much of this difference, as 22% isn’t that different from 35%.  Even after the election we won’t know for sure which approach was “right”, especially if Hillary wins.  I suppose if Trump wins then 538 will look good, as its numbers for Trump have been higher than elsewhere.  But even then it won’t be a crowning victory, after all, 538 is predicting a Hillary victory.  Michael Moore won’t be impressed.  You’d need lots of repeated tests to establish whether rational expectations beats really good models.

I believe it does, which is why I prefer NGDP futures markets to Lars Svensson’s suggestion that the Fed target its own internal forecast, based on structural models.  So there, a Trump Derangement Post that actually had implications for monetary policy!

PS.  Here is the betting market map:screen-shot-2016-11-05-at-8-58-28-pm

Notice that they have Hillary winning New Hampshire and North Carolina too.

PPS.  Not enough derangement in this post?  I love this Matt Yglesias post on that disgusting illegal immigrant Melania:

So there’s really nothing so surprising about the Melania story. Trump doesn’t like immigrants who change the American cultural and ethnic mix in a way he finds threatening and neither do his fans. Europeans like Melania (or before her, Ivana) are fine. I get it, David Duke gets it, the frog meme people get it, everyone gets it.

But it does raise the question of why mainstream press coverage has spent so much time pretending not to get it. Why have we been treated to so many lectures about the “populist appeal” of a man running on regressive tax cuts and financial deregulation and the “economic anxiety” of his fans?

Slovenian models include anorexics, prima donnas, former porn stars, and some, I assume, are good people.