Powerful evidence that the big problem is demand, not structural

It’s easy to find a correlation between two macro variables, as the same sort of factors (inflation, output growth, interest rates) tend to affect all sorts of financial and macro variables.  For that reason, one of the most useful statistical tools is time-varying correlations.

For example, when I studied the Great Depression I recall reading a year end summary of the 1931 stock market in the New York Times.  They summarized the US stock market month by month, and mentioned German problems in all the summaries of the last 7 months, but none of the first 5 months.  And sure enough, during the last half of 1931 the price of German war bonds was highly correlated with US stock prices.

But how can one prove cause and effect?  Lots of things might have affected both German war bond prices, and US stock prices.  For example, signs of a strong economic recovery worldwide might boost the value of both assets.  On the other hand, if German problems were depressing US equity markets, then you’d expect a strong correlation between German war bond prices and US equity prices during periods when there were lots of problems in Germany (late 1931), but essentially no correlation during periods when German problems were not particularly significant (early 1931.)  And that’s exactly what I found.  Or to take another example, Steve Silver and I found that US industrial production became negatively correlated with nominal wages after 1933, precisely when New Deal programs began to artificially raise nominal wages.

Almost a year ago David Glasner mentioned that he was working on a similar study, this time estimating the time-varying correlation between US inflation expectations and US equity prices.  He has frequently sent me very significant results, but I held back from mentioning them in order to let him get the project completed before publicizing the results.  He has now placed the paper at the SSRN web site, where others can read it.  In the meantime I notice that others have observed this pattern, indeed the commenter Gregor Bush recently mentioned some similar results.

It is well known that there is normally little correlation between US inflation expectations and US stock prices.  Higher inflation might boost stock prices if associated with growing aggregate demand, but higher inflation can also lead to expectations of tight money, or higher taxes on capital, since capital income is not indexed.  Indeed the high inflation of the 1970s seems to have depressed real stock and bond prices.  In general, the stock market seems content with the low and stable inflation of recent decades, at least judging by reactions to changes in inflation expectations.

David looked at 8 years of data, from January 2003 until December 2010, and divided the sample up into 10 sub-periods.  He found almost no significant correlation between inflation expectations (TIPS spreads) and stock prices (S&P 500) until March 2008.  (Actually, there was a modest positive correlation during the first half of 2003, another period when people worried about excessively low inflation.)  After March 2008, the correlation was highly significant, and positive.  Right about the time where the US began suffering from a severe AD shortfall, the stock market began rooting strongly for higher inflation.  And it still is, even in the most recent period.  Money is still too tight.

There is no way to overstate the importance of these these findings.  The obvious explanation (and indeed the only explanation I can think of) is that low inflation was not a major problem before mid-2008, but has since become a big problem.  Bernanke’s right and the hawks at the Fed are wrong.

Arnold Kling noted that the AS/AD approach can sometimes verge on the tautological.  When inflation and output both fall, demand-side economists are likely to infer that AD is lower.  And they are also likely to claim that the fall in AD caused both the drop in inflation and the drop in output.  Of course that’s not the only evidence demand-side economists have, but in many cases it’s hard to find definitive evidence of causation.

In my view the time-varying correlations between inflation expectations and stock prices are one of the most important pieces of evidence we have that AD became a problem after mid-2008.  It will be interesting to see if those economists who are skeptical of demand-side explanations can come up with a plausible alternative explanation for this pattern.

PS.  If anyone knows how to estimate a continuous time-varying correlation, it would be interesting to find out precisely when in 2008 US equity markets started rooting for higher inflation.

PPS.  I’ve been asked to comment on the very weak performance of NGDP and the very strong performance in final sales during 2010:4.  If I had any confidence in the government numbers I would comment.  Let’s wait a few more quarters, look at lots of different data, and see where we are.  Again, my test of a policy is its effect on expected NGDP, not the actual movements in NGDP (which reflect all sorts of factors.)


Tags:

 
 
 

61 Responses to “Powerful evidence that the big problem is demand, not structural”

  1. Gravatar of Gregor Bush Gregor Bush
    31. January 2011 at 07:55

    Based on a 52-week rolling correlation of daily prices and yields it looks like the correlation was zero or slightly negative in 2004, 2005 and 2006 and turned slightly positive in the spring of 2007. The positive correlation rose sharply in 2008 and even more so in 2009 and 2010. Interestingly, the period of the highest correlation seems to be the last 10 months, ever since the euro crisis broke.

    Perhaps it would be more interesting to use the Cleveland Fed’s methodology and back out the expected inflation rate 2 years hence, thus removing the impact that large changes in oil prices have on the headline CPI number. I suspect that if you did that you would find a stronger correlation in early 2008.

    It appears very much as if the correlation between TIPS spreads and stock prices is a function of the output gap – or, more likely, the expected future output gap. It likely reflects a lack of confidence that the Fed will quickly restore inflation its pre-crisis rate (or, if you prefer, NGDP to its pre-crisis trend). So I think the correlation tells you something about expected future AD. If the correlation is strongly positive, it’s likely that expected future AD is too. If the correlation were strongly negative I think it would be a signal that expected future AD is too high.

    Interestingly, the correlation between equity prices and breakeven inflation spreads has been much lower over the same period in Canada (where there is a formal inflation target) and the increase in the correlation has been much less.

  2. Gravatar of Gregor Bush Gregor Bush
    31. January 2011 at 07:58

    Sorry, that line in the last paragraph should say:

    “If the correlation is strongly positive, it’s likely that expected future AD is too low.”

  3. Gravatar of Gregor Bush Gregor Bush
    31. January 2011 at 08:18

    On a monthly basis in 2008, the correlation was negative in January, strongly positive in February and March (Bear Stearns), mildly positive in April and May, negative in June and July flat in August and then surged September and remianed very high in October and Novemeber.

    It was also very high in March and April of 2009 (QE 1 was annouced on March 17th).

  4. Gravatar of Ken Ken
    31. January 2011 at 08:27

    I think stock prices have been rising with inflation expectations because people are worried about holding onto cash balances that will lose real value. They view stocks as a safer holding in that environment than cash. (Right or wrong, I am one of those investors!) Is this logically equivalent to “the big problem is demand” and/or “rooting strongly for higher inflation”? I can tell you that many investors are concerned about inflation and they don’t know where to put their money… I’m not sure they are rooting strongly for higher inflation. Perhaps Scott or other economists can tell me how these feelings map to their models.

  5. Gravatar of Gregor Bush Gregor Bush
    31. January 2011 at 09:12

    Ken,

    It’s true that there are some investors who are concerned about high inflation. But people betting real money in the bond market are, on average, expecting lower inflation over the next 2, 5 and 10 years than they were before the financial crisis. So in the aggregate, people are no more worried about high inflation than they were in 2006. In fact, they’re worried less. And what the correlation tells us is that, over the past two years, whenever the average expected inflation rate has fallen, stock prices have tended to fall with it. Under normal circumstances this tends not to be the case because the ex ante equity premium is positively correlated with inflation (investors demand a higher risk premium when inflation is high). But these are not normal circumstances. Inflation expectations are stock prices are responding to changes in some variable with which they are both positively correlated. And the only variable that could be is expected future aggregate demand.

  6. Gravatar of Scott Sumner Scott Sumner
    31. January 2011 at 09:15

    Gregor, Thanks for that monthly data, it suggests to me that the inflation surge in mid-2008 may have reduced fears of recession. But it also may have led to worries that higher expected inflation numbers would prevent the Fed from easing. So it’s not easy to figure out the precise reason for the correlation.

    I’m not sure if the output gap is the only factor. If we had had TIPS markets in the 1970s, I’d guess that the correlation might have been negative, even during periods where there was an output gap.

    Ken, But inflation expectations were much higher before the recession, and there was no correlation. That’s the beauty of time-varying correlations, it allows us to rule out many alternative hypotheses.

    By the way, look at real stock prices during 1966-81, stocks are not a good investment during high inflation.

  7. Gravatar of John Hall John Hall
    31. January 2011 at 09:36

    Scott,
    There are several techniques used to capture time-varying correlations. Many have already been developed as matlab packages. Bollerslev did the first paper on multi-variable Garch. Dynamic Conditional Correlation from Engle and a paper from Tse and Tsui are other recent contributions.

    Regime-switching would be another way to examine the problem. For instance, if you took a vector auto-regression of a few of the key variables and allowed for regime-switching in the means and covariance matrix, then you could find the regime-switching correlations. In the matlab implementation it would take a while to estimate, but is pretty easy to set up.

  8. Gravatar of Matt Matt
    31. January 2011 at 09:36

    Scott, have you compared the change in inflation expectations from the 2009 stimulus vs. QE announcements?

  9. Gravatar of Matt Matt
    31. January 2011 at 09:37

    Let me rephrase: “…the change in inflation expectations resulting from…”

  10. Gravatar of marcus nunes marcus nunes
    31. January 2011 at 09:39

    Scott
    I did the exercice using 20 day rolling correlation. The difference is that strongly positive correlations begin one year later – March 09 – when QE1 was implemented. It dissapears for a short period – March/April 10 – when QE1 was ended and goes on strongly positive since then.

  11. Gravatar of Benjamin Cole Benjamin Cole
    31. January 2011 at 09:50

    A very important post by Scott Sumner.

    Kudos to David Glasner and Gregor Bush as well.

    I’ll say it again–there is no more important economic blogger, or commentator period–than Scott Sumner right now.

    Any economist not reading Sumner is simply not in the game–you can disagree with Sumner, but you should have an explanation of why, other than your politics or shibboleths.

    Bernanke, I hope you are reading. Pur it on, Ben, pour it on, spike the punch bowl, crank up the music, blow the doors off and set the house on fire. Ben Bernanke, you need to get an open-chest shirt, some gold chains, and tight bell-bottoms. Let it rip, but a stiff drink in both hands.

    There ain’t no inflation Jack, and there is gobs of unused capacity, and the USA economy imports goods, labor, services and capital freely, tamping down inflation (should we ever see any again).

    You see Niagara? That is a stream next to the river of money the USA economy can easily absorb. The Mississippi is a rivulet next to what is needed.

    Have I made my point?

  12. Gravatar of StatsGuy StatsGuy
    31. January 2011 at 11:45

    Gregor:

    “But people betting real money in the bond market are, on average, expecting lower inflation over the next 2, 5 and 10 years than they were before the financial crisis.”

    Ssumner:

    “look at real stock prices during 1966-81, stocks are not a good investment during high inflation.”

    Broad statements like this leave me puzzled. Gregor, when you say “people betting real money”, you have to consider that the level of Fed monetization is about equal to the rate of deficit spending right now. Average Primary Dealer takedown in recent auctions which is then flipped back to the Fed in <30 days is quite high. In essence, primary dealers are buying bonds for the commission and then reselling them back to the Fed at close to purchase price. Does that count as "real people making real bets?" I dunno. At the margin there are some real people, certainly, but that's a margin which does not define a Fed-less equillibrium.

    Note that I _agree_ with the Fed's actions, but I don't know that the price of 10 year bonds is real proof of inflation expectations right now, since you do not have an equal amount of private money taken opposing sides in the trade. Predictive markets be damned. I suppose you could make an argument about private money arbitraging the Fed, but wow – that's a lot of money, especially if the Fed appears committed.

    Ssumner – when you say stocks are not a good investment during inflation, I have no idea what you mean. Seriously. Do you mean relative to cash? Do you mean international or domestic stocks? And doesn't it matter what is causing the inflation (supply shock, or money printing)? If it's money printing, one would imagine stocks are a better investment than cash – indeed, corporations benefit in multiple ways, first by seeing the cost of their debt decline, second by seeing AD increase, and third by reducing real wages. All around, it's a winner. But if it's a supply shock (aka, oil prices in the 70s), then it seems a different story, right? It seems you could just as well argue that "low stock prices tend to increase inflation expectations".

    I'm sorry, but you are always telling us never to argue from price, and yet…

  13. Gravatar of StatsGuy StatsGuy
    31. January 2011 at 12:11

    Regarding the Glasner paper, I’m not sure I know what it means.

    Glasner writes:

    “In this simple model, expected inflation ought to occasion precisely offsetting effects on expected cash flows and (via the Fisher effect) on rates of discount, leaving present values, and hence asset prices, unchanged.”

    Um, yah, this is like, totally true in a closed economy. But today, some 50% of corporate profits come from outside the US, compared to… uh, 1985? The impact of the exchange rate alone on profits from exports could be driving this correlation. If you break this down by exposure of domestic stocks to output international markets (and domestic production costs + dollar denominated debt), you would get… oh, yah, an even higher correlation. Go figure.

    Moreover, short term rolling correlations (DAILY)? This could easily be the result of the rising dominance of statistical arbitrage by trading desks. I have no idea what the data look like for inflation vs. other asset classes over the same periods. I suspect it’s high, because the ENTIRE FINANCIAL PRESS has been observing that correlations across ALL ASSET CLASSES have spiked. This has been the source of the complaint that there are fewer opportunities to diversify, that international and domestic stocks don’t move opposite each other anymore like they used to, that the entire trading regime right now is bimodal (risk on/risk off)…

    Perhaps the effect is resulting from the loss of the dollar’s reserve currency status and the internationalization of the financial markets. Notably, for various periods of time, we’ve been observing that Bond AND Stock prices are correlated – this could result from money flowing into or out of US dollar denominated assets as a whole. But is this daily correlation strong evidence of a breakdown in the Fisher Effect? I dunno. Maybe, sure. I agree with the conclusions.

  14. Gravatar of Doc Merlin Doc Merlin
    31. January 2011 at 12:47

    Its been 2 and a half years Scott. By now any AD wage-stickyness argument should be inoperative

  15. Gravatar of Greg Ransom Greg Ransom
    31. January 2011 at 12:53

    StatsGuy, stop wrecking their fun!

    The closed economy model is as addictive for economists as is Modern Warfare for boys playing the XBox.

    This problem has been widely acknowledged for more than 50 years — and it hasn’t stopped an economist yet.

    The closed economy model is that official academically sanctioned game.

    Let the boys play!

  16. Gravatar of spencer spencer
    31. January 2011 at 13:27

    One factor causing inflation to depress stocks in the 1970s was the impact of inflation on reported versus actual profits. In the high inflation 1970s standard depreciation measures caused firms to massively understate the replacement cost of capital so that reported profits were artificially inflated.

  17. Gravatar of Morgan Warstler Morgan Warstler
    31. January 2011 at 13:45

    “Gregor, Thanks for that monthly data, it suggests to me that the inflation surge in mid-2008 may have reduced fears of recession. But it also may have led to worries that higher expected inflation numbers would prevent the Fed from easing.”

    Inflation surge of 2008 reduced fears of recession?
    Increased fears the FED wouldn’t ease?

    I don’t remember it that way in either regard, I remember exactly the opposite:

    Inflation (Gas prices) were killing everyone, and as the dollar plummeted, the Saudi Oil minister was arguing forcefully, they had nothing to do with it. He said, every percent the dollar falls, oil goes up $4 per barrel.

    All of this led smack dab into a slow motion car crash that everyone was predicting all summer (Nancy Peolosi was screaming all through 2008, about lost jobs) until right before the election, when the whole country knew for sure we were in a recession, and it is looked like the Fed was going to print money FOREVER.

    —-

    I mean why don’t we ever focus our time talking about when raising rates earlier could have staved off this thing?

  18. Gravatar of Gregor Bush Gregor Bush
    31. January 2011 at 15:10

    Statsguy,
    There are still huge volumes of nominal Treasuries and TIPS being traded at current market prices between private investors. And among those private investors there is always a marginal investor – someone who is just indifferent between holding a nominal bond and holding a TIPS bond. So current market prices must approximately reflect inflation expectations of investors – even if the Fed is a large player in the market. And right now those spreads are still a bit lower than where there were during the 2004 to mid-2008 period. There was a better case to be made for worrying about high inflation in the middle of 2006 than there is today.

    Yes, there is anecdotal evidence that some investors are worried about inflation. But that’s almost always the case. The information in market prices is much more useful for policymakers than anything that you’ll hear from a particular fund manager or trader.

    Also, I’m not sure that I understand your argument. Are you saying that the Fed’s QE program is resulting in TIPS spreads being LOWER than they otherwise would be?

  19. Gravatar of Ryan Ryan
    31. January 2011 at 16:30

    I don’t understand. If it is true that inflation and stock prices are uncorrelated in the long run, why do we attempt to say they are correlated in the short run. There is a real methodological problem with this kind of approach.

    The correct approach is to start with a theory, such as, “I think inflation has a big impact on stock prices.” Then you test your theory to see if it holds up. If no good correlation can be found with a bajillion historical data points, then the evidence is weak.

    If correlations pop up after you farm out years that argue against your theory, then you’re fudging the data.

    If you take any subset of the data and come up with a plausible inflation story to explain what happened, that is quite a bit better. However, I don’t understand why it would be a superior approach to taking any other economic statistic, such as consumer confidence or mortgage foreclosures or the MAT q of horseshoes sold…

    In other words, I guess I need more underlying theory to buy into the correlation. I don’t think we’ve moved beyond the stichomancy phase at this point.

  20. Gravatar of StatsGuy StatsGuy
    31. January 2011 at 18:30

    @ Gregor

    The marginal buyer argument you’re presenting is a very strong version of wisdom of crowds. Consider the market a median dollar estimator. There are dollars that vote, and a well behaved market settles near 50%. Now imagine there’s a massive position on one side. You’re assuming the dollars will decide among themselves the right answer, and THEN they will settle in the same place (arbitraging away the position). This assumes unlimited liquidity, but the point of Fed intervention is precisely because liquidity is limited.

    Thus, what we have instead of a median estimator, is an 80% quantile estimator… or something. I don’t even know what. The bias in TIPS spread depends on what bill issues are monetized, and in what sequence. Indeed, your assumption that the private markets will attempt to arbitrage away the Fed action is tenuous, because many private actors will attempt to _anticipate_ Fed actions, and will bid up the prices of issues they think the Fed will monetize. (This has become a multi-billion dollar trade.) TIPS is a small and illiquid market relative to the rollover of T-bills, which are being monetized at a rate of ~7 billion per trading day. (BTW, don’t get me wrong, I’m all for the monetization, but let’s call it what it is.) In practice, the TIPS spread could be biased in whatever whatever direction the Fed wants it to be. If the Fed wants to narrow the spread, it monetizes non-TIPS, collapsing the rates. Does it want the yield curve to rise? It monetizes near term issues. Does it want the 2/30 spread to narrow? It monetizes the 30 years.

    Moreover, the TIPS spread relies on BLS estimates of inflation, which are increasingly widely challenged.

    Again, don’t mistake me – I think inflation is necessary to whittle down the debt (and “wealth”) of a generation that ran up too many IOUs for its children, and to make US wages more competitive – but I do not believe the numbers coming out of BLS reflect true cost of living for staples. In short, I don’t believe adding 200 channels of cable TV at the same price compensates for the price of apples going up, nor do I believe that food inflation has been tame for the last 2 years. Then again, I don’t watch 200 channels of cable TV, so maybe I’m wrong.

    Frankly, I don’t know very much at all.

  21. Gravatar of StatsGuy StatsGuy
    31. January 2011 at 18:31

    @ Greg

    From now on, my monicker is FunWrecker

  22. Gravatar of Mark A. Sadowski Mark A. Sadowski
    31. January 2011 at 19:24

    Scott,
    You wrote:
    “PS. If anyone knows how to estimate a continuous time-varying correlation, it would be interesting to find out precisely when in 2008 US equity markets started rooting for higher inflation.”

    What is this? A homework assignment?

    Seriously though, I was intrigued by the results enough to fiddle with it myself. I estimated 6 month rolling correlations but excluded the $/euro coefficient (because frankly I got lazy).

    I’m finding statistically significant and positive correlations between the log of S&P500 and the 10 year TIPS spread first difference in six month periods overlapping 7/01/2006-3/01/2007, 2/01-2008-09/01/2008 and 3/01/2010-12/01/2010. The first period marks when the yield curve turned negative, the second period marks when inflation expectations plummeted and the third period marks when European public debt anxiety spiked.

    So despite the fact I excluded the $/euro coefficient my extremely preliminary results seem to be corroborating some of Glasner’s results and what I’m reading here.

    On the other hand my intuition tells me that these results may not be very robust as they could be conditional on the choice of variables as well as the length of the periods.

    For what it’s worth, my favorite part of Glasner’s paper was the following:

    “If so, the expressed rationale for the Fed’s quantitative easing policy (Bernanke 2010), namely to reduce long term interest rates and stimulating spending on investment and consumption, reflects a misapprehension of the mechanism by which the policy operates, increasing expectations of inflation and future profitability and, hence, of the cash flows derived from real assets, increasing asset values along with both with inflation expectations and real interest rates. Rather than a policy to reduce interest rates, quantitative easing appears rather a policy for increasing interest rates, though only as a byproduct of increasing expected future prices and cash flows.”

  23. Gravatar of scott sumner scott sumner
    31. January 2011 at 19:48

    John, Thanks, The monthly data Gregor Bush provides above gives me some idea. The price spike in mid-2008 made things a bit more complicated than I expected. It seems like there were several switches; worry about recession in early 2008, then worry about high inflation, then recession worries returned with a vengeance.

    Matt, Great question, but it’s hard to do because we don’t know when the markets became aware of fiscal stimulus. To give the Keynesians their due, I do recall one stock market increase that was attributed to a strong Obama announcement on stimulus. But the period as a whole was quite bleak for markets, so I doubt the stimulus had a major positive effect.

    Marcus, That’s interesting, but both Gregor and David Glasner got strong positive correlations at certain times in 2008. What maturity of TIPS spread did you use?

    Statsguy, I think you are missing how much evidence we have:

    1. It’s a huge Treasury bond market out there, the Fed’s role is modest. There are also other types of safe bonds, that sell at a predictable premium to T-bills. There are bank CDs.

    2. Going beyond the bond market we have CPI futures markets, an entirely different way of measuring inflation expectations.

    3. The we have actual core inflation trending lower, quarter after quarter.

    Looking at the entire picture, I’m comfortable asserting inflation expectations are low.

    Regarding stocks and inflation, I don’t think they did well during high inflation, because it sharply raised the tax on capital. That may be less true now that tax rates on capital are lower. But I don’t think it was just supply shocks, stocks didn’t do all that well (in real terms) in 1966-72, before the supply shocks.

    Yes, cash was even worse. But the Dow went from 990 in 1966 to 850 in 1982, and the price level probably tripled. That’s not good. Gold or real estate or art were probably batter during that period.

    I don’t necessarily agree with Glasner’s comments on the Fisher effect, I found his inflation expectations findings to be the most interesting. As far as it being a result of innovation–keep in mind the correlation suddenly appeared in 2008. A point I should have mentioned is that David would have found the exact same correlation during the Great Depression. I should probably add an update to that effect. To summarize, we get stocks strongly and positively correlated with inflation in the 1930s, then not much correlated for many decades, then they suddenly become highly correlated again during 2008. That seems pretty suggestive to me.

    Doc Merlin, So all the unemployment is now voluntary?

    Greg Ransom, I’m going to take a wild guess, go out on a long limb, and assume that David would have found the same correlation using an index of domestic-oriented stocks. And then there’s the question of why the correlation didn’t occur before 2008. People are good at spinning alternative stories for the correlation, not so good at explaining why it’s time varying.

    Spencer, I agree.

    Morgan, Yes, maybe it was a stretch to assume it reduced fears of recession. But I stand by my other point–I think it led to a belief that the Fed would not ease.

    Ryan, You said;

    “I don’t understand. If it is true that inflation and stock prices are uncorrelated in the long run, why do we attempt to say they are correlated in the short run. There is a real methodological problem with this kind of approach.”

    You misunderstood, it’s not long run vs. short run, it’s some time periods and not others. That’s different. Of course stocks respond positively (in nominal terms) to changes in the price level during any period, that’s a given.
    The model is that low inflation is a problem in a disinflationary recession and zero interest rates, but not otherwise.

  24. Gravatar of scott sumner scott sumner
    31. January 2011 at 19:56

    Mark, Thanks for completing your homework on time. 🙂

    Your findings seem a bit different from both Gregor and David, who find the correlation beginning in 2008, or at least the strongly positive correlation beginning in 2008.

    I also liked that Glasner paragraph you cited.

  25. Gravatar of Dirk Dirk
    31. January 2011 at 20:05

    From above: “If so, the expressed rationale for the Fed’s quantitative easing policy (Bernanke 2010), namely to reduce long term interest rates and stimulating spending on investment and consumption, reflects a misapprehension of the mechanism by which the policy operates… Rather than a policy to reduce interest rates, quantitative easing appears rather a policy for increasing interest rates…”

    That is exactly what I was scratching my head over when I read the Fed minutes from December:

    “Participants pointed to a number of factors that appeared to have contributed to the significant backup in yields, including an apparent downward reassessment by investors of the likely ultimate size of the Federal Reserve’s asset-purchase program, economic data that were seen as suggesting an improved economic outlook, and the announcement of a package of fiscal measures that was expected to bolster economic growth and increase the deficit over coming quarters… A number of participants indicated that, because the backup in rates appeared to importantly reflect changes in investors’ expectations about the size of Federal Reserve asset purchases, the backup was consistent with purchases
    helping to keep longer-term yields lower than would otherwise be the case.”

    I was a bit shocked reading that, because from reading this blog I had assumed that QE should raise long-term rates as a result of increased expectations and that the Fed would be thinking along those lines. They seem to be confusing the mechanism of buying bonds, which might cause high-frequency type blips up in bond prices, with the greater force of the policy itself. So does the Fed know what they are doing or don’t they?

  26. Gravatar of Mark A. Sadowski Mark A. Sadowski
    31. January 2011 at 20:08

    Scott,
    Yes, evidently just under the wire.

    Gregor noted a positive (although not necessarily statistically significant) correlation in the spring of 2007. The only thing I think that I’ve done different from Glasner so far is that I’ve excluded the $/euro coefficient.

  27. Gravatar of Morgan Warstler Morgan Warstler
    31. January 2011 at 22:12

    “It seems like there were several switches; worry about recession in early 2008, then worry about high inflation, then recession worries returned with a vengeance.”

    Ok, now we’re cooking with gas. During 2008, I spent far to much time staring at Forex screens attempting to grapple with some insane hedging strategy that gives me the willies just recounting.

    In fact the entire swing we saw where EVERYONE thought there would be inflation… as the dollar continued to plunge.

    What Scott says happened is that the Fed didn’t ease – but what I remember was everyone couldn’t believe that people sought the “safety” of the US Treasuries. There was much grumbling….

    http://finance.yahoo.com/echarts?s=EURUSD=X+Interactive#chart2:symbol=eurusd=x;range=5y;indicator=volume;charttype=line;crosshair=on;ohlcvalues=0;logscale=on;source=undefined

    So Scott, you idea is that when people try to find safety in treasuries, that’s exactly when to try and make them seem less safe?
    —–

    But, it does raise an interesting question: has Scott already ruled out the Oil Boom as the reason for the global crash?

    http://zoniereport.com/?p=1735

    I mean sure the loose monetary policy of 2004-2007 could have led to oil prices. But maybe it was oil prices that seized the heart of the patient.

  28. Gravatar of Morgan Warstler Morgan Warstler
    31. January 2011 at 22:18

    Meanwhile, we see another prof point that it is ALL about housing:

    http://www.cnbc.com/id/41355854

    Houses are 11% empty! Apartments are full!

    Now couldn’t this be considered proof that what we really needed are a complete liquidation of the insolvent banks with a flood of MBS properties ONLY FOR the guys with 30% down on the $1 auctions they won?

    Isn’t there a model where home vacancies are only 8%, and 3% of total houses were mercilessly torn from the dying clutches of the zombie banks – so that they might be sold to triumphant Austrians and rented cheap to the deserving foreclosed?

    Is there no way to model that outcome?

  29. Gravatar of Dtoh Dtoh
    1. February 2011 at 03:38

    Scott,
    I have two questions.

    First, what is the mechanism by which inflation expectations translate into equity prices.

    Second, I think there are some very significant correlations in exactly the same period between equity prices and expectations on tax rates. Have you looked at this. From a causality perspective, I think it tax rate expectations is an easier argument.

  30. Gravatar of nanute nanute
    1. February 2011 at 04:30

    http://www.clevelandfed.org/research/commentary/2009/0809.cfm
    This approach by the Cleveland Fed shows that inflation expectations for the next 10 years is about 2%. On the question of the Feds intent regarding QEII: If the Fed was overly concerned with deflation as opposed to inflation, would it not be the desired policy outcome to lower long term real interest rates? I can’t see how longer term real rates decline without an significant increase in the correlating rate of inflation. I would argue that inflation is what is needed to get the economy back on track in both the jobs market and the output gap production side. Current political reality will make this approach all the more difficult for the Fed. The mindset taking hold in Congress will more than offset any possible positive effects of Fed monetary policy.

  31. Gravatar of bertusmaximus bertusmaximus
    1. February 2011 at 04:41

    statsguy seems to know what he is talking about … not sure about the rest. like maybe you guys need to get out of your ivory towers.

    amazing that the words “credit migration” haven’t been mentioned or “cost of carry” as that would reflect issues pertaining to most investors or allocators of capital.

  32. Gravatar of “Fill the hole with (not by) workers” – Spend! « Historinhas “Fill the hole with (not by) workers” – Spend! « Historinhas
    1. February 2011 at 04:47

    […] Scott Sumner argues from a different perspective that the problem is demand, not […]

  33. Gravatar of marcus nunes marcus nunes
    1. February 2011 at 05:32

    Marcus, That’s interesting, but both Gregor and David Glasner got strong positive correlations at certain times in 2008. What maturity of TIPS spread did you use?

    I used the 5 year TIPS spread. The high frequency (20 day) RC for March-Dec 08 had a mean of 0.16 (insignificant). There were a few instances when the Cor was >0.8. For the March 09-Dec 10 period, the mean Cor was 0.6 and there were many instances of Cor>0.8.
    Later I did a “low frequency” RC (52 week). The Cor becomes positive (and stays positive) after 9/11/08! The mean Cor from then until dec 10 is 0.7

  34. Gravatar of Gregor Bush Gregor Bush
    1. February 2011 at 05:45

    Statsguy,
    “You’re assuming the dollars will decide among themselves the right answer, and THEN they will settle in the same place (arbitraging away the position). This assumes unlimited liquidity, but the point of Fed intervention is precisely because liquidity is limited.”

    “If the Fed wants to narrow the spread, it monetizes non-TIPS, collapsing the rates. Does it want the yield curve to rise? It monetizes near term issues. Does it want the 2/30 spread to narrow? It monetizes the 30 years.”

    I don’t agree with this. Liquidity doesn’t have to be unlimited. Liquidity just has to be sufficient to allow market prices reflect the marginal investor’s expectation about the future. And in both the nominal and TIPS markets this is approximately the case (with the possible expectation of September to December 2008 but even there I’m not so sure). To see why, take you example to an extreme. Suppose that investors true expectation was for 2% inflation over the next ten years but the Fed wanted TIPS spreads to “appear” to expect 0% inflation . According to your theory, the Fed could do this by buying nominal Treasuries and selling TIPS in sufficient quantities. And what I’m saying is that it will be impossible for the Fed to “collapse” that spread to 0%. That’s because investors all over the world that expected +2% inflation would be buying TIPS and selling nominal Treasuries like crazy. Why would anyone hold a nominal Treasury with the same yield as a TIPS bond if they expected 2% inflation over the life of the bond? This process will continue until the spread roughly reflects investor’s expectations – even if the Fed is a very large player in the market.

    All of this is to argue that TIPS spreads approximate inflation expectations as measured by the CPI. Your second point about the CPI not measuring inflation correctly is really a separate issue. We can have that argument another day.

  35. Gravatar of StatsGuy StatsGuy
    1. February 2011 at 06:29

    @Gregor

    “Why would anyone hold a nominal Treasury with the same yield as a TIPS bond if they expected 2% inflation over the life of the bond?”

    I’m not sure who “they” is here. The first portion of the argument above is to indicate that there’s enough liquidity to allow arbitrage, but then the second point involves “they” expecting 2% inflation. But that’s not how the market works as an information aggregation mechanism.

    “They” is actually a distribution of expectations. The market draws a cutpoint in this distribution, at approximately the median voting dollar. Roughly HALF the dollars (or people if everyone is equally wealthy) expect more than 2% inflation, and roughly half expect less.

    If you load one side of the bet, then you no longer have a 50/50 split, and the median estimator becomes an arbitrary quantile estimator. Imagine 2% is the median estimate of a market with 2 trillion in liquidity over a 6 month period (the portion of money that actually moves around to arbitrage things – who knows what that portion actually is…) Imagine 2% is the point where 1 trillion dollars thinks the future rate will be lower, and 1 trillion thinks it will be higher. Now you have a spare 600 billion join 1 side of the bet. Let’s say the market clears with 1.3 trillion on each side of the bet. Now, the number of private dollars on one side of the bet is 1.3 trillion, and that leaves 0.7 trillion of private dollars on the other. There’s no reason at all to expect that the cutpoint which achieves this settles at the same target as the cutpoint which achieves an even 1 trillion vs. 1 trillion split.

    The fallacy in the original argument is assuming that IF it were true that the un-biased market were settling at 2%, then that means that “investors all over the world” all homogenously expect 2% and would all be willing to arbitrage any Fed action back to 2%. Not true. The market is not monolithic, and individuals are not composed of aggregates (rather, aggregates are composed of individuals).

    We can leave the inflation issue for another day.

  36. Gravatar of StatsGuy StatsGuy
    1. February 2011 at 06:55

    @ ssumner

    “A point I should have mentioned is that David would have found the exact same correlation during the Great Depression. I should probably add an update to that effect.”

    That’s a much better argument.

    “Yes, cash was even worse. But the Dow went from 990 in 1966 to 850 in 1982, and the price level probably tripled. That’s not good. Gold or real estate or art were probably batter during that period.”

    Over the 18/19 year period from 66 to 82, you need to consider dividends. Recently, the portion of dividends in total equity returns has lowered, but in earlier decades was higher – and, indeed, we as large as the cap gains component. See figure 1 here. So Cash actually did MUCH worse than stocks.

    http://www.tweedy.com/resources/library_docs/papers/highdiv_research.pdf

    [Also, by picking ’66 and ’82, you’re top-ticking and bottom-ticking the index. Alas, I know you believe in the EMH.]

    On the bulk of evidence for low inflation, I generally agree that 2008/2009 saw a price collapse. What I am disagreeing with is the latter comment that “people betting real money in the bond market are, on average, expecting lower inflation over the next 2, 5 and 10 years than they were before the financial crisis”.

    If you compare this to 2007 early 2008, perhaps – notably, the commodity ‘bubble’ in early 2008 agrees. But 2004/2005, not so sure. Looking at forward expectations, the Core CPI metric evidence vanishes (it’s the future, man). That leaves points 1 and 2. HOWEVER, it’s wrong to assume the CPI futures market and the TIPS spread are independent, since it’s easy to lock up an arbitrage trade. If CPI futures were predicting 3% inflation, and the TIPS spread were predicting 1.5% due to Fed intervention, savvy investors would buy sell inflation protection in the futures market and buy TIPS instead of tbills, narrowing that gap.

    So looking at future expectations, we have 2 pieces of evidence instead of 3, and those 2 pieces are not independent.

    The Glasner argument, on its own, is not convincing enough – I’m not saying its wrong, but that there are other (equally simple) explanations. If you extended the data back 100 years, then yes, I’d believe it. Hands down.

  37. Gravatar of david glasner david glasner
    1. February 2011 at 08:16

    First of all, I want to thank Scott for first encouraging me to publish my results after I first mentioned to him that, inspired by his regressions on daily changes in stock prices in his work on the Great Depression, I had been playing around with data on inflation expectations and stock prices, generously sharing his insights along the way. I originally started collecting the data and running the regressions out of sheer curiosity and have been consistently surprised at the strength of the results. I also was sure that someone else would discover the link way before I could ever do so, and had it not been for Scott, I would not have written a nearly 20 page paper about what I found. In rereading and revising my paper over the past few days, and following yesterday’s discussion on the blog, the main defect, that I found in the paper was the lack of any acknowledgment of Scott’s role in the paper. So this introduction is partly by way of making amends for that glaring oversight, which I hope eventually to fix. I will only respond to a few of yesterday’s comments and may make some further comments as the discussion continues on the blog, but if anyone would like to discuss the paper with me offline, I can be reached by email at dglasner@ftc.gov or glasner.david@gmail.com

    To Gregor, I wonder if you could explain to me how you arrived at your correlations between stock prices and inflation expectations. Where does the 0.7 correlation come from and what exact variables are being correlated?

    To Benjamin Cole: Obviously I agree with everything you say about Scott. There is a Talmudic saying to the effect that one should only offer partial praise to a person in his presence. Obviously we both agree with that precept.

    To Stats Guy, Could you explain to me how my (to be sure) oversimplified explanation of why expected inflation should theoretically have no effect on asset prices in any way depends on whether an economy is open or closed? Are you saying that if people expect relatively more inflation in terms of dollar than in terms of euros that stock prices quoted in dollars will go up to reflect the depreciation of the dollar? Does the fact that I included the dollar/euro exchange rate in my regression not address that issue? I also don’t understand what you mean by extending the data for 100 years. My whole point is that in “normal” circumstances, there is little or no correlation between inflation expectations and asset prices. 100 years of evidence that there is little correlation between inflation and asset prices would make a difference to how you view the strong correlation between inflation expectations and asset prices since 2008?

    To Mark Sadowski: Thanks for quoting your favorite part of my paper. I am not sure that I agree with your choice, but it is not bad, and I have tried to improve it slightly in the latest draft.

  38. Gravatar of Gregor Bush Gregor Bush
    1. February 2011 at 11:05

    StatsGuy

    Yes, I’m aware that there is a distribution of expected outcomes around the mean. And I don’t think that changes my argument. Again, to modify my previous example, suppose that investor’s expectations have a mean of 2% inflation with a standard deviation of 1% and that inflation expectations are roughly normally distributed. According to your theory, the Fed could move the spread to 0% by buying Treasuries and selling TIPS in large quantities. And I’m saying that there is sufficient liquidity, and potential liquidity, in the market to make that impossible. If the Fed “loads one side of the bet”, these securities will be mispriced relative to investor’s expectations about the future and there will be a flood of hedge funds, pension funds, active mutual funds and SWFs rushing in to take the other side. I don’t invest in Treasuries or TIPS. But if the spread between them fell to zero I would put a long TIPS short Treasuries trade on in a second. And I would be joined by many others – with a lot more money than I have.

    And at any rate, I’m not sure why you’re making this argument. The Fed isn’t “loading one side of the bet”. It’s buying BOTH nominal and TIPS bonds. So it’s hard to see how the Fed would be compressing the spread via a liquidity effect to any meaningful extent. On the other hand, its widening of the spread since last August via the expectations effect has been very large indeed. And of course that was the whole point of QE in the first place – to raise expectations about future aggregate demand (growth and inflation). The moral of the story is that in liquid markets like US Treasuries, these liquidity concerns are fairly trivial – it’s expectations that matter

  39. Gravatar of scott sumner scott sumner
    1. February 2011 at 16:25

    Dirk, You asked:

    “So does the Fed know what they are doing or don’t they?”

    They know more than the Fed did in 1930, but less than they should know.

    Mark, I’m not sure why adding exchange rates makes a difference.

    Morgan, You said;

    “So Scott, you idea is that when people try to find safety in treasuries, that’s exactly when to try and make them seem less safe?”

    It’s not about making them seem less safe, it’s making monetary policy easier to offset the effect on the dollar from the hoarding of Treasuries.

    Oil prices were falling rapidly in August, September, October, 2008, that should have helped the stock market.

    Dtoh, The causal explanation is that both stocks and TIPS spreads responded positively to expectations of faster growth in AD, and vice versa.

    I don’t follow your tax rates expectations argument–we are looking at daily movements in each variable.

    Morgan, You said;

    “Isn’t there a model where home vacancies are only 8%, and 3% of total houses were mercilessly torn from the dying clutches of the zombie banks – so that they might be sold to triumphant Austrians and rented cheap to the deserving foreclosed?”

    I hereby nominate Morgan to be poet laureate of the Austrian economists.

    nanute, I agree the Fed needs to do more. But I am confident monetary policy can more than offset any fiscal tightening. The question is: Will they be aggressive enough?

    Real rates are a tricky indicator. They fall with easy money (good) but they also fall with a weaker economy (bad.) Don’t jump to conclusions.

    bertusmaximus, You said;

    “like maybe you guys need to get out of your ivory towers.”

    I’ve just spent 3 hours out of my ivory tower shoveling snow. BTW, are you a Roman centurian?

    Marcus, Normally I suggest the person first try to replicate the results using the same technique, then make one change at a time. Did you use first differences? He also used exchange rates. Given two people found the same result independently, I’m going to tentatively assume it holds. But clearly “more research is needed.”

    Statsguy and Gregor, I lean toward Gregor’s view. FWIW, I think changes in the spread are normally a pretty good indicator of inflation expectations. I think liquidity considerations did distort the TIPS spread somewhat in late 2008 (making expected deflation seem bigger than it really was) but then I’ve argued that’s another good reason to ease–if people are hoarding nominal T-bonds that intensely, not even wanting perfectly good TIPS, then money is way too tight.

    Statsguy, You said;

    “[Also, by picking ’66 and ’82, you’re top-ticking and bottom-ticking the index. Alas, I know you believe in the EMH.]”

    Yes, but recall that 1966 was when the Great Inflation started, and 1982 was when it ended. I agree stocks did better than cash, but I also think they did worse than inflation hedges.

    You said;

    “That leaves points 1 and 2. HOWEVER, it’s wrong to assume the CPI futures market and the TIPS spread are independent, since it’s easy to lock up an arbitrage trade. If CPI futures were predicting 3% inflation, and the TIPS spread were predicting 1.5% due to Fed intervention, savvy investors would buy sell inflation protection in the futures market and buy TIPS instead of tbills, narrowing that gap.”

    In fact, they do differ, but it may be a timing lag issue. In principle I agree.

    I would add that the TIPS spreads are actually overestimating inflation right now, as commodity prices have recently risen, and there’s a lag in the TIPS adjustment. That distorts the rates slightly when there is a big up or down move in commodity prices.

    I feel very confident that we are looking at roughly 1% core inflation for several years. At Bentley we were just told to expect 2% annual raises over the next five years. Nationwide we are also seeing 2% wage inflation, which translates into about 1% core inflation.

    One other point. In my Depression manuscript written years ago I mentioned that stocks and inflation were positively correlated in the 1930s, but if anything negative correlated in recent decades (before this crisis.) I mention that to show that I didn’t just fit an ad hoc theory to the data David came up with, this was already my presumption, and I think for theoretically very sound reasons. But you are right that it’d be better to have 100 years of data. I’ve always thought that an under-researched area of macro was the relationship between macro variables and asset prices.

  40. Gravatar of StatsGuy StatsGuy
    1. February 2011 at 23:12

    @ David Glasner

    Post 2008 is not alone convincing because there were other contemporaneous events that might explain the pattern. 100 year data is more convincing if, as ssumner observes, the only other time period during which the correlation manifests is in the Great Depression. That would be an odd coincidence, especially since the other possible explanations didn’t occur in the 30s.

    Regarding the issue of open/closed economies – yes, if the dollar suddenly lost reserve currency status in 2008, we would observe the US behaving less like a closed economy (wherein the Fisher effect would be more pronounced). All dollar denominated asset prices would become more correlated. I noted the inclusion of the EURUSD as a control variable, and was not convinced. This faces two challenges. First, this alone does not control for other currencies or for non-currency measures of wealth which were being used as stores of value. Indeed, around 2008, we saw a breakdown in currency basket correlations. Commodity-linked currencies behaved differently than the USD, EUR, Yen and even GBP. Consider the Aussie dollar. Observe the close correlation through mid 2008, then serious divergence.

    http://finance.yahoo.com/echarts?s=EURUSD%3DX#chart1:symbol=eurusd=x;range=5y;compare=audusd=x
    I don’t mean to be combative, but thinking back to the financial press in 2009, the dollar carry trades were the singular most important trading phenomenon from late 2008 through most of 2009. If you were to put the price of gold in as a control, and find the same result, I’d be impressed. But that’s not really fair to your hypothesis, since gold is a direct inflation (or deflation?) hedge.

    That’s not to say I disagree with your conclusions, but the inflation expectation/stock price correlation was neither surprising nor (alone) the “strongest evidence” yet of an NGDP (or AD) problem.

    So why bother with this argument? Sounds pretty nitpicky, right? The issue is this: your main opponents in this argument are the Austrians, and their argument (even right now) is that the “recovery” is illusory, that it’s being bought with debt, that the stock market ramp is the result of Fed debasement of the dollar but does not in fact represent real wealth increases, and that all of this is obvious if you look at the value of stock indices vs. the price of gold. Controlling for the EUR means little since it too was being debased, often with US dollar swaps that were funded directly by the Fed.

    [cont. below]

  41. Gravatar of StatsGuy StatsGuy
    1. February 2011 at 23:32

    Now if you look a the 5 year plot of SPY vs. GLD, well, they have a point.

    http://finance.yahoo.com/echarts?s=SPY+Interactive#chart2:symbol=spy;range=5y;compare=gld

    Generally correlated, but divergent with gold leading. Strongest break in the correlation was around March 09, which is interesting, and perhaps the last few months.

    In any case, does the data you represent in your paper constitute the strongest refutation of the Austrian argument we can muster? Unfortunately, it might fit easily into Austrian story at face value. The final test of the argument is just arriving on scene – the question is whether the price of stocks (denominated in gold) begins to increase (which happened immediately post QE in 2009, and again very recently). And ultimately, this comes down to the observation that QE will be deemed successful if unemployment drops.

    I suppose if you really wanted to prove the Austrians wrong, you could select (as your dependent variable) the period change in the price of stocks as denominated in gold, and remove the EUR covariate from the regression. The question then becomes whether stocks (a measure of future output/profit) outperform an inflation hedge when inflation expectations increase during a major recession. I dunno.

  42. Gravatar of StatsGuy StatsGuy
    1. February 2011 at 23:52

    @ssumner

    I recall a discussion (this blog?) about divergence between inflation swaps and TIPS. I think someone suggested it might be because of a discontinuity in TIPS at around zero when inflation expectations were low. Of course, that means expectations were low.

    @ssumner and Gregor – I think the question comes down to whether or not the Tbill market is so deep and liquid that the Fed can’t have a direct effect (but only the indirect effect of setting expectations). I would observe this, however… In some periods, treasury rates declined even as expectations of economic growth increased. Also, recent press suggests that major banks were anticipating Fed monetizations of specific issues and front-running the Fed (buying issues they expected the Fed to target). This actually accelerated the liquidity effect, rather than arbitraging it away.

    … But I can’t easily find a summary list of the specific issues the Fed has targeted in the past several months, or in round 1 of QE. The question of relative purchases of TIPS vs. non-inflation adjusted securities is a good one.

  43. Gravatar of nanute nanute
    2. February 2011 at 02:54

    Scott: You said in reply to my comment: Real rates are a tricky indicator. They fall with easy money (good) but they also fall with a weaker economy (bad.) Don’t jump to conclusions. Point taken. On the other hand, if deflation sets in, isn’t it the case that real rates of present value bonds will in fact rise?

  44. Gravatar of StatsGuy StatsGuy
    2. February 2011 at 07:07

    @glasner

    Note that I never mentioned the second issue – it has to do with functional specification. If the idea is to extract out the exchange rate issue, I was wondering why one would use the Euro (or gold, or some weighted basket of currencies) as a liner control? That model is

    d(Stocks) = B0 + B1*d(Eur) + B2*d(InfExp)

    an alternative model is

    d(Stocks/Eur) = B0 + B1*d(InfExp)

    Why does this matter? Imagine that the EUR is an incomplete measure of some underlying variable called ‘currency depreciation’ (since EUR is only one currency). As such, any measure of the relationship would be attenuated (the ‘iron law’), and the value as a covariate diminishes – residual covariation is captured by the other variables in the model (InfExp). The latter model seems more intuitively descriptive of what the model is, and less vulnerable to the iron law. But the error structure is possibly weird.

  45. Gravatar of Ottovbvs Ottovbvs
    2. February 2011 at 15:18

    Benjamin Cole
    31. January 2011 at 09:50

    With all due respect there have been a number of leading economists, mentioning no names, who have been saying “It’s the Demand, stupid” for two years. In fact it’s been completely obvious to me and I’m not an economist, just a widget maker.

  46. Gravatar of scott sumner scott sumner
    2. February 2011 at 19:32

    Statsguy, So the Austrians think high inflation is pushing up stock prices. How do they explain 1966-82?

    You said;

    “I recall a discussion (this blog?) about divergence between inflation swaps and TIPS. I think someone suggested it might be because of a discontinuity in TIPS at around zero when inflation expectations were low. Of course, that means expectations were low.”

    It occurred on December 1, 2008, when the US agency that keeps data on TIPS switched from a new to an old, or an old to a new bond. The difference reflected the fact that new bonds can’t fall in price during deflation, but old ones can. That’s all I remember.

    Reagrding T-bills, I don’t think anyone claims they are useful guides to inflation expectations, except in a very crude way. There is a liquidity effect, so I agree with you that the Fed can influence short T-bill rates (real and nominal.) I don’t know as much about short term TIPS spreads, but I can imagine they might be distorted a bit on occasion. I usually look at 2 year or 5 year TIPS spreads.

    I also don’t know what they have purchased, although I was told it was going to be mostly medium term securities (say 3 to 10 years)

    nanute, This question is harder than it seems. I happen to think that deflations are usually unanticipated. But it partly depends which prices should be in the Fisher equation; flexible prices (which tend to be unforecastable) or all prices, which show some momentum. If it is all prices then real rates probably would rise during deflation. If only flexible prices belong in the Fisher equation (as I believe) then ex ante inflation is almost never strongly negative, so (ex ante) real rates may not rise. Of course it’s a given that ex post rates rise during deflation.

    Ottovbvs, Yes, but how many were also saying in late 2008 that monetary policy is the best way of addressing the issue?

  47. Gravatar of StatsGuy StatsGuy
    2. February 2011 at 21:11

    ssumner:

    “Statsguy, So the Austrians think high inflation is pushing up stock prices. How do they explain 1966-82?”

    I’m not an Austrian, but I think an Austrian would say that was supply driven inflation (union wages + oil prices), which caused margin compression for corporations. By comparison, they see current monetization causing dollar depreciation and lowering domestic production costs for exports (aka, debasing real wages). But the Austrians also claim that the debt binge will rebound when commodity price increases erode margins and will eventually trim earnings, thus wiping out gains, collapsing asset prices, and triggering an even worse economic implosion. Keynesians claim inflation will only happen if capacity utilization increases, and then QE will have been a success. So the acid test is coming – do we see capacity utilization (employment) recover before real inflation kicks in? Personally, I think we’re going to need to see some _real_ inflation to thin down the debt load and get a decent recovery, but without a credible commitment to fiscal discipline this is dangerous.

    Anyway, I think Mervyn King agrees with me… Really an excellent speech

    http://www.youtube.com/watch?v=xSvRONc2IOY

    We need to see a number of years at 4-5% inflation to really recover (commensurate with restoring the NGDP trajectory, as you’ve claimed). My concern is that 2% inflation will fail to achieve the necessary recovery, but will further deplete support for aggressive monetary policy if the recovery remains weak.

  48. Gravatar of Dtoh Dtoh
    2. February 2011 at 21:56

    Scott,
    It seems to me what you are saying is that stock prices and inflation are both positively correlated to expectations of looser monetary policy during a recession. I’m not sure why you consider this significant or non-obvious. All you need to do is listen to market commentary to understand this.

  49. Gravatar of Balloon Juice » Modest Proposals Balloon Juice » Modest Proposals
    3. February 2011 at 08:08

    […] Meanwhile Kevin Drum reads a new paper by David Glasner which shows that low inflation plus low interest rates lead to companies hoarding cash, thereby decreasing aggregate demand. Scott Sumner writes: […]

  50. Gravatar of Are Sumner and Krugman Incorrectly Claiming Victory? Are Sumner and Krugman Incorrectly Claiming Victory?
    3. February 2011 at 19:25

    […] As the reader may have guessed, the reason I went through all this is that Sumner and Krugman claim that the above chart shows they’re right. Go […]

  51. Gravatar of david glasner david glasner
    3. February 2011 at 19:47

    StatsGuy, you said:

    “Post 2008 is not alone convincing because there were other contemporaneous events that might explain the pattern.”

    There is always more than one possible explanation of any observation. My explanation has the modest virtue of explaining why we usually don’t observe a close correlation
    between inflation expectations and did observe one in the post-2008 period.

    “100 year data is more convincing if, as ssumner observes, the only other time period during which the correlation manifests is in the Great Depression.” We simply don’t have the data on inflation expectations to run 100 years of data, but casual empiricism suggests that we wouldn’t see a positive correlation in 1970s when inflation was rising and stocks were falling. Can you think of any other period besides the 1930s when the correlation would be observed?

    “That would be an odd coincidence, especially since the other possible explanations didn’t occur in the 30s.”

    Sorry, but I can’t figure out what you mean, but you seem to be saying that the 1930s and the post-2008 period are not similar. Scott and I think that they are remarkably similar. Why would anyone think otherwise?

    “Regarding the issue of open/closed economies – yes, if the dollar suddenly lost reserve currency status in 2008, we would observe the US behaving less like a closed economy (wherein the Fisher effect would be more pronounced).”

    Sorry, but again I am not understanding you. The dollar has not lost reserve currency status, it would take a huge shock for the dollar to lose reserve currency status, and I seriously doubt, for all the idle speculation about alternatives to the dollar, whether it will lose its status in my, or your, lifetime. I don’t believe that the US behaves like a closed economy and I don’t understand how the Fisher equation is implicated in whether it does or not in a flexible exchange rate regime.

    “I noted the inclusion of the EURUSD as a control variable, and was not convinced. This faces two challenges. First, this alone does not control for other currencies or for non-currency measures of wealth which were being used as stores of value.”

    I think that I mention in a footnote, but maybe, come to think of it, I did not that I included the trade-weighted value of the dollar as a variable along with and instead of the $/euro exchange rate. In no period did that variable have a significant coefficient.

    “If you were to put the price of gold in as a control, and find the same result, I’d be impressed. But that’s not really fair to your hypothesis, since gold is a direct inflation (or deflation?) hedge.”

    I haven’t included the price of gold in my regressions, but I will try. I suspect that it will have little correlation, but we’ll just have to see.

    “your main opponents in this argument are the Austrians, and their argument (even right now) is that the “recovery” is illusory, that it’s being bought with debt, that the stock market ramp is the result of Fed debasement of the dollar but does not in fact represent real wealth increases, and that all of this is obvious if you look at the value of stock indices vs. the price of gold.”

    I don’t take most Austrian arguments about the financial crisis seriously, because there is nothing Austrian about them. To borrow an epithet from Joan Robinson, what goes under the name of Austrian economics nowadays is a bastardized Austrianism. It has nothing to do with the capital theory of Bohm-Bawerk and Hayek, and it’s just a simplistic and simple-minded repetition of the idea, which has nothing to with Austrian capital theory, that easy money causes asset bubbles. My paper is completely orthogonal to that proposition. To believe that the doubling of stock prices since March 2009 is illusory because of a trivial increase in prices and reduction in the exchange value of the dollar is so crazy that I can’t even imagine how a rational person could even entertain such a lunatic thought. So, for my money trying to prove the Austrians wrong is just a waste of time and effort and I have no interest in playing their game. You seem pretty sensible, so I guess I don’t understand why you seem to have an interest in taking them on.

    Finally about whether to use stock prices or to deflate stock prices by some measure of the change in the value of money, deflating by the euro or by the value of gold forces the regression to assume that value of stocks changes by the same amount as the change in the chosen measure of value. Doing it my way lets the data speak for themselves.

  52. Gravatar of StatsGuy StatsGuy
    4. February 2011 at 11:54

    I hate absorbing your time, since I’d rather you be fighting the good fight, so “briefly”:

    “you seem to be saying that the 1930s and the post-2008 period are not similar”

    I think I’m saying that some alternative explanations for the change in correlation in 2008 that come to mind (the domination of HFT correlation desks, for example) weren’t present in 1930s, but that your argument applies equally well to both. So, that’s a point in favor of your hypothesis, and adding the 1930s case makes the argument more compelling.

    “Can you think of any other period besides the 1930s when the correlation would be observed?”

    Maybe the post-war demobilization. Not sure. You certainly don’t have to worry about the “open economy” argument there – we blew up the rest of the world economy.

    Re: “Fisher effect” & open economies

    Let me clarify – an alternative to your argument that the correlation was an indicator of the lower bound nominal effect on real interest rates is to argue that in an open economy, during a period of aggressive easing it is possible that depreciation of the currency can have a real effect on _profitability_ of firms. Stock/asset prices are not a direct measure of the health of the economy, but profitability. In this sense, inflation (and accompanied dollar depreciation) would not induce “precisely offsetting effects on expected cash flows” for firms if firms as a whole were more exposed to foreign currencies as revenue sources and more exposed to domestic currency as costs (households pick up the offset, since they are paid in dollars and buy imports for consumption). This effect could occur without the nominal boundary effect you mention, and it would be quite large if currency changes were expected to be permanent due to discounting. If firm costs were $100 per widget and revenue was $101 worth of other currencies (profit of $1), then a 1% devaluation means revenue of $102, and a profit of $2. Big impact, particularly on companies operating with low gross margins.

    The open economy argument goes like this. Firms have debt and production costs in dollars, and sales in something else. If the dollar drops (due to inflation worries which were obsessing markets at the time), profits go up. There is evidence in favor of this – I believe company stocks with more exposure to international sales saw greater correlation with inflation, while stocks with domestic sales saw very low correlation with inflation. I haven’t run that model, but if you split stocks into two groups – let’s say domestic (power companies, telecoms, stable retail) and export companies (manufacturing exports, chiefly), which ones had higher correlation with inflation expectations? I’m pretty sure we know the answer.

    But if this were the case, why do we not observe that correlation in earlier time periods as well? That’s the discussion about “other factors” (most of which were not relevant to the 1930s).

    “I included the trade-weighted value of the dollar as a variable along with and instead of the $/euro exchange rate”

    Sorry, missed that.

    Separately, for the nominal boundary effect to be the right explanation, we need to see “expected rates of deflation above some threshold”. Do you have any thoughts about what that threshold would be in the relevant time periods? And, do you think the deflation rate really exceeded the real return generated by firm assets throughout the entire period in which you observe positive correlation? (And what % of the post2008 period do we actually have true deflation expectations instead of merely disinflation?) I recall true deflation expectations being only mildly negative and only briefly, but the positive correlation extends across a wider period.

    “I don’t take most Austrian arguments about the financial crisis seriously, because there is nothing Austrian about them.”

    Pop Austrians frighten me because of their political impact – my background is political econ. I take them seriously.

    Lastly, I’m not sure I understand your preference for the control/covariate model instead of dividing the dependent variable, but you’ve worked with the data and I haven’t. If the deflator is an imperfect (but unbiased) estimate of some underlying phenomenon, I’m not sure of the concern (other than odd error structure in the dependent variable – maybe do a log/log if that’s a big deal). Ideally, both models say the same thing.

    Sorry for occupying time.

  53. Gravatar of ssumner ssumner
    5. February 2011 at 09:09

    Statsguy, You said;

    “I’m not an Austrian, but I think an Austrian would say that was supply driven inflation (union wages + oil prices), which caused margin compression for corporations.”

    Yes, you are definitely not an Austrian. They’d certainly blame the Fed, and they’d be 100% correct.

    You said;

    “Personally, I think we’re going to need to see some _real_ inflation to thin down the debt load and get a decent recovery, but without a credible commitment to fiscal discipline this is dangerous.”

    More NGDP with thin the debt load–makes no difference if the NGDP growth is real or prices.

    Dtoh, It may be obvious to you, but lots of bloggers on the left and right don’t see it as obvious at all. (Kevin Drum, Bob Murphy, etc.)

  54. Gravatar of Food Prices: Blame the Weather or Blame Ben Bernanke? – Also sprach Analyst Food Prices: Blame the Weather or Blame Ben Bernanke? - Also sprach Analyst
    7. February 2011 at 01:13

    […] some unrelated but interesting discussions on a paper by David Glasner (see more discussions by Scott Sumner, Kevin Drum, Paul Krugman, Karl Smith, and Kash), which found that the correlation between S&P […]

  55. Gravatar of david glasner david glasner
    7. February 2011 at 14:08

    Statsguy:

    First of all, you needn’t be so concerned about absorbing my time, I’m not all that diligent.

    You said:

    Re: “Fisher effect” & open economies

    “an alternative to your argument that the correlation was an indicator of the lower bound nominal effect on real interest rates is to argue that in an open economy, during a period of aggressive easing it is possible that depreciation of the currency can have a real effect on _profitability_ of firms.”

    I don’t disagree with that

    “Stock/asset prices are not a direct measure of the health of the economy, but profitability.”

    Profitability and the health of the economy are highly, but imperfectly, correlated

    “In this sense, inflation (and accompanied dollar depreciation) would not induce “precisely offsetting effects on expected cash flows” for firms if firms as a whole were more exposed to foreign currencies as revenue sources and more exposed to domestic currency as costs (households pick up the offset, since they are paid in dollars and buy imports for consumption). This effect could occur without the nominal boundary effect you mention, and it would be quite large if currency changes were expected to be permanent due to discounting. If firm costs were $100 per widget and revenue was $101 worth of other currencies (profit of $1), then a 1% devaluation means revenue of $102, and a profit of $2. Big impact, particularly on companies operating with low gross margins.”

    Sorry, but this is getting too complicated for me. I agree that some firms are sensitive to exchange rate fluctuations, but some are more sensitive on the cost side than the revenue side, so on balance, without fully understanding your argument, I have difficulty assigning much empirical weight to it. But that’s about all I can say off the top of my head.

    “The open economy argument goes like this. Firms have debt and production costs in dollars, and sales in something else. If the dollar drops (due to inflation worries which were obsessing markets at the time), profits go up. There is evidence in favor of this – I believe company stocks with more exposure to international sales saw greater correlation with inflation, while stocks with domestic sales saw very low correlation with inflation. I haven’t run that model, but if you split stocks into two groups – let’s say domestic (power companies, telecoms, stable retail) and export companies (manufacturing exports, chiefly), which ones had higher correlation with inflation expectations? I’m pretty sure we know the answer.”

    I would simply repeat what I just said above. Additionally, insofar as our inflation puts pressure on other countries to inflate, either they have to raise their exchange rates making their exports to the US more expensive or the internal prices rise which would also tend to make their exports to the US more expensive.

    “Separately, for the nominal boundary effect to be the right explanation, we need to see “expected rates of deflation above some threshold”. Do you have any thoughts about what that threshold would be in the relevant time periods? And, do you think the deflation rate really exceeded the real return generated by firm assets throughout the entire period in which you observe positive correlation?”

    Even though I used a relatively long-term measure of expected inflation, theoretically I would attach greater importance to a relatively short-term expected inflation rate and real interest rate. Since I think that the short-term real interest rate probably went negative in 2008, even zero expected inflation would have been enough to cause a sell off of assets. For sure during the crisis and the immediate aftermath expected short-term inflation was negative. And I think expected inflation and the real rate have been pretty close to each other so that we have been in this funny range for some time at which increases in expected inflation have a big effect on asset prices, but we may soon come out of it if QE proceeds a while longer.

    “Pop Austrians frighten me because of their political impact – my background is political econ. I take them seriously.”

    At that level I take them seriously also. And Sarah Palin and Glenn Beck, Lord protect us.

    “Lastly, I’m not sure I understand your preference for the control/covariate model instead of dividing the dependent variable, but you’ve worked with the data and I haven’t. If the deflator is an imperfect (but unbiased) estimate of some underlying phenomenon, I’m not sure of the concern (other than odd error structure in the dependent variable – maybe do a log/log if that’s a big deal). Ideally, both models say the same thing.”

    I am not enough of an econometrician to have a well-worked out view, so what I said before was just something I tossed off. I will continue to think about it, but intuitively deflating by the exchange rate or the price of gold does not seem right to me.

    “Sorry for occupying time.”

    No apology necessary.
    (And what % of the post2008 period do we actually have true deflation expectations instead of merely disinflation?) I recall true deflation expectations being only mildly negative and only briefly, but the positive correlation extends across a wider period.

  56. Gravatar of StatsGuy StatsGuy
    7. February 2011 at 17:57

    “And Sarah Palin and Glenn Beck, Lord protect us.”

    Hallelujah!

    🙂

  57. Gravatar of How Obama can fix the economy for free | PARTISANS How Obama can fix the economy for free | PARTISANS
    18. February 2011 at 15:46

    […] so little attention. Contrary to the numerous misguided claims about “structural” unemployment, overwhelming evidence points to a cyclical problem, which by its very nature can be solved with adequate stimulus. With Congress abdicating its […]

  58. Gravatar of Inflation Expectation and Stock Prices Inflation Expectation and Stock Prices
    26. June 2011 at 11:10

    […] change in the relationship between inflation expectation and stock prices (see more discussions by Scott Sumner, Kevin Drum, Paul Krugman, Karl Smith, and Kash), now it is worth pointing out that as the economy […]

  59. Gravatar of Inflation Expectation and Stock Prices | Brian Brown's Official Website Inflation Expectation and Stock Prices | Brian Brown's Official Website
    26. June 2011 at 14:52

    […] change in the relationship between inflation expectation and stock prices (see more discussions by Scott Sumner, Kevin Drum, Paul Krugman, Karl Smith, and Kash), now it is worth pointing out that as the economy […]

  60. Gravatar of Dangers of detrending. « Probably Matters Dangers of detrending. « Probably Matters
    6. July 2011 at 13:37

    […] may be one or two other […]

  61. Gravatar of Inflation Expectations Dove Today « Uneasy Money Inflation Expectations Dove Today « Uneasy Money
    22. September 2011 at 12:55

    […] a paper available here, and discussed here, here, and here, I have shown that until early 2008, the there is no systematic correlation to be […]

Leave a Reply