Thursday, February 27, 2014

Stock/Bond Asset Allocation - Don't bother with duration risk. (Pt. 2)

I want to look some more at asset allocations between stocks & bonds, because I think there might be some subtle but powerful issues to think about here regarding portfolio construction.  (Here's my previous post.) I find evidence that argues while a cash allocation can provide diversification and asset rebalancing benefits, bonds with duration risk have not provided useful downside portfolio risk mitigation, given historical maturity premiums.

More below the fold.

Tuesday, February 25, 2014

Stock/Bond Asset Allocation

A friend asked me about stock & bond allocations, and I decided to review the data.  Of course you're all doing it wrong.  I explain below.  Make sure you read to the end....You're welcome.  :-)


For bonds I have used the Moody's AAA bond rate series, from Fred, not adjusted for defaults, and adjust the value of the bond portfolio monthly, based on the monthly rate, assuming average bond duration of 30 years.  For stocks, I use Robert Shiller's historical S&P500 data.

Here is a graph of cumulative gains (rebalanced monthly) for different allocations:

The following two graphs are comparisons of the returns to these allocations over time, for a 10 year holding period and for a 20 year holding period.

There is a dispute in finance about the size and persistence of the equity risk premium (the additional return that an investor gets by taking on the extra risk of owning equities instead of bonds).  One issue is that these relationships just aren't stationary enough for a century's worth of data to give us an answer.

As can be seen in these graphs, almost all of the excess gains to equities came in the period from 1940 to 1965.  At shorter holding periods, there is a significant amount of variance in equity returns.  But, with a longer holding period, returns revert to long-term trends, so that with very long holding periods, there are two distinct eras:

(1) holding periods beginning from 1940 to 1965 where equities earned a tremendous premium with much less risk than bonds.

(2) all other periods, where equities earned a slightly higher return with slightly more volatility and the occasional crisis.

So the allocation decision boils down to two questions.

1)   Are we about to see another 1940-1965?  This is impossible to answer.  We are currently in a period of very low interest rates.  My intuition wants to say that in a low interest rate environment, bond allocations should be very low, because (a) in a low rate environment it would be very difficult for bond yields to match the combined growth and income value of stocks and (b) the risk profile would skew very negatively for bonds since bonds have defined income potential and a limit on valuation gains due to the zero lower bound on interest rates.

The one era of high interest rates that we have seen in the past century was associated with relatively high returns with relatively low variance for long holding periods, along with a modest equity premium.

Eras of low interest rates have had a more bifurcated set of outcomes, with periods of either very high equity premiums or periods of very negative returns on equities compared to bonds.  This is largely because the most damaging economic crises have come during deflationary liquidity crises associated with monetary policy.  These episodes have been associated with low interest rates.  (The graph at right shows the difference between stock and bond returns <blue line> overlaid on top of the 10 year treasury yield <red>.)

So, the current level of interest rates isn't as useful of an indicator as I would hope.  This is again due to the fact that the serial correlations we see in these broad market behaviors mean that over a century there have only been a few separate regimes of return behavior, with no clear forward indication of when the regimes might switch again.

2) What is your holding period?  This is a question that we can answer.  And, considering the lack of a definitive answer to number 1, the answer to this question should generally guide our decisions.  The decisions here are going to be imprecise, but the good news is that we can get a pretty good idea of the appropriate allocation within 10% to 20%, and over the holding period, the differences between the outcomes of allocations within that range are going to be pretty marginal.  As the two line charts to the right demonstrate, as the holding period is extended, the number of periods where stocks underperform bonds declines.  At a 20 year holding period, there are just a few times where stocks slightly underperformed bonds during the last 94 years.

This is the classic chart of returns for portfolios with different allocations.  These three return/risk curves assume annual rebalancing and reflect the average returns and variance of holding periods with start dates after 1919 and end dates prior to 2013.

Equity returns can have large annual variance, but they tend to revert to long term trends.  The effect that we can see from this is that, as the holding period is extended, the standard deviation of holding period returns of equity heavy portfolios is mitigated, and these return/risk curves rotate counterclockwise with longer holding periods.  The extra returns available from equities can be captured during these long holding periods without taking on so much extra risk.

In any holding period, there are very high expected costs from being too highly-weighted in bonds.  Even at short holding periods, holding more than 70% bonds is suboptimal.  But, once a portfolio reaches the efficient frontier where there is a reasonable trade-off between higher risk and higher return, there isn't that much difference between portfolios.  Even at 20 years, the difference in expected returns between a 70% equity allocation and a 100% equity allocation is about 1% annually.

So, here's where I would say that as long as you pick an allocation to stocks in the 40% to 90% range (lower for shorter holding periods and higher for longer holding periods), depending on your risk aversion, you'll be fine.  But, I would be wrong.

These models are based on normal distributions and linear correlations with homoscedastic errors.  Both of these assumptions are surprisingly wrong.

Buy Equities, but not for the reason you think you should.

Looking again at the cumulative holding period returns for different portfolios, you can see that bonds have a bifurcated return distribution.  There are long periods with low returns and long periods with high returns, and not much in the middle.  I thought maybe all of this would disappear with inflation adjustments, but the nominal and real data are surprisingly similar.

Here is a histogram of real bond returns over a 20 year time horizon.  Instead of providing a large basket of moderate return outcomes, bonds provide a bunch of relatively poor outcomes and a bunch of relatively good outcomes, without much in the middle.  The relatively low average returns from bonds are supposed to be the cost we incur for their moderating influence.  I would say that for the past century, bonds have not been doing that job well.

Next is the histogram of real stock returns over a 20 year time horizon.  There is a similar bifurcation here, but less pronounced than with bonds.  Out of 74 holding periods, a full 32 bond return outcomes are worse than any stock outcomes.  Bonds aren't just failing at mitigating failures at the left end of the outcome distribution - they are creating them by the bushel.

And, it's even worse than that.  Over 20 year holding periods, there is little correlation between stocks and bonds, so you would think that at least bonds would provide diversification benefits.  But this is not the case either.  Holding periods with low bond returns have no correlation with stock returns.  But, as bond returns increase, they tend to correlate relatively strongly with stocks.  In other words, bonds only perform well when you don't need them to perform well, and when you do need them to perform well, they never perform well.

On these scatterplots, the diagonal line represents equivalent outcomes between the two portfolios.  Outcomes above the line are outcomes where 100% stock allocations outperformed the other portfolios.

I have included two additional graphs comparing the fully invested stock portfolio with a 80/20 mix and a 60/40 mix.  In both cases, we can see again that the allocation to bonds has not provided any systematic protection against relative losses over the 20 year time frame.  As the bond allocation is increased, the average return decreases substantially, with no improvement in the number of poor outcomes.  There is one holding period that begins in 1930 where bonds provide a 70% return compared to 25% for stocks.  One holding period out of 74.

Even for a 10 year holding period, the past 94 years (84 holding periods) provides little reason to utilize bonds for stability related to the holding period.

For 10 year holding periods, I show the comparison between stocks & bonds, and between stocks and a 60/40 and 40/60 portfolio.  There are two holding periods (the two starting in 2000 and 2001) where bonds provide significantly positive returns while stocks provide significantly negative returns.  But, outside of these two holding periods, for every period where a bond allocation would have provided a boost, there is at least one other holding period where a bond allocation created a similarly negative outcome.

The final graph below shows the cumulative distribution of returns for a portfolio of 100% stocks and a portfolio of 60/40 stocks and bonds.  There is very little gain at the bottom of the outcome distribution from the 60/40 portfolio with which to justify so many holding periods in the heart of the distribution where significant losses are taken, relative to a 100% stock portfolio.

Some analysis of portfolio allocations might suggest a relatively large bond allocation, because on an annual basis, bonds and stocks do provide significant diversification.  Portfolios with a higher bond allocation will produce a higher annual Sharpe Ratio.  So, the paths these 10 and 20 year stock portfolios take en route to their final values will probably tend to be more volatile than the diversified portfolios.  And that is probably an important factor to consider if you have to deal with agency issues that come along with managing other people's money.  But, the non-normal distribution of outcomes, the heteroskedastic behavior of the correlations, and the mean reverting nature of these asset classes are strong enough influences on the performance of long term portfolio behaviors, that if you can commit to these longer holding periods without risking capital flight during potentially volatile return paths, the past century's experience gives absolutely no reason to allocate to bonds on a 20 year time scale, and scant reason to do it even at a 10 year horizon. (follow up post).

community animated gif on Giphy

Sunday, February 23, 2014

What is the expected date of the first interest rate increase?

Median exit time from the zero interest rate policyI was reading this econbrowser post , and I was surprised to see this graph, from the San Francisco Fed, of the median expected exit from the zero lower bound.

Here is how it compares to my very basic estimate, using treasury yields:

Now, the estimate I show in this graph is from a very broad analysis of treasury yields, with a blunt adjustment made to account for the asymmetry of the yield curve near the zero lower bound (ZLB).  I have a more rigorous model that starts with Eurodollar futures data in September 2012, but the blunt estimate from treasury data fits the estimate from that model surprisingly well.  (Below is a comparison of my two estimates since September 2012.)  The graph above is similar to this previous graph that I posted.  Keep in mind that the two graphs above measure the distance to the ZLB exit at any given time, whereas the graph in my previous post, and this little graph comparing my two models are showing the expected exit date on a fixed calendar.

And, the pattern fits the pattern of the estimate from the Fed pretty well, too, up to QE3.  We both see a spike before QE2, which is reversed during QE2, before spiking again at the end of 2011 and moving sideways into 2012.

But, after that, they seem to move in opposite directions.  My model shows the exit moving back, from 2.5 years to 1.5 years, from the beginning of QE3 until September 2013, when it levels off at about 1.5 years.  The Fed model shows the exit date remaining fairly level, around 2 years, until September 2013, after which it declines steeply toward 1 year.

This is an important difference.  My data would support what I would call the market monetarist version of events.  The expected exit date had been moving ahead over time, so that we were not making progress on escaping ZLB.  But, QE3 improved economic and inflation expectations, which stabilized the expected exit date until economic improvements in early 2013 caused the exit date to move toward us.  Taper talk in June 2013, followed by the establishment of a tapering schedule later in the year, reversed some of these expansionary expectations, which moved the expected exit date back into late 2015.  Continued improvements in the economic outlook, in spite of the taper, have continued to have a positive influence on the expected exit date.  This is my interpretation, and I expect the exit date to remain fairly stable.  I expect continued economic progress pulling the exit date back a little more, and my main fear is that the economy's inherent growth won't maintain enough momentum to overcome the slightly disinflationary effect of the taper, and the expected ZLB exit date will start to recede again.  (Here is a more detailed review of summer 2013 rate changes.)

But, the Fed data would tend to support what I call the "Wizard of Oz" view of the Fed, which ascribes a powerful ability to target interest rate levels over time to the Fed.  With this data, and in this view, QE3 signaled a plan from the Fed of holding rates lower for longer, so the expected exit date kept moving into the future, but with the talk of taper in June 2013 and the subsequent implementation after September 2013, the Fed has signaled that they will raise rates sooner than they had previously planned, so the expected date of the exit has moved back toward us.

What's Up?

My model is constructed by assuming that the range of possible dates for the ZLB exit is described by a normal curve.  My model fits the forward yield curve to a curve defined by that normal distribution's mean and standard deviation, combined with the expected slope of the yield curve from that date.  I am reporting the mean expected date of the ZLB exit, which, by assumption, is the same as the median expected date.

The Fed's model is much more sophisticated than mine, and they are measuring a set of inputs that, through Monte Carlo simulations, produce unique distributions of the expected exit date.  These distributions tend to have a positive skew, which should be expected.  The Fed is reporting the median expected date of the exit, which because of the skew, is sooner than the mean expected exit.  This is why the Fed's expected median dates tend to be earlier than my expected mean dates.

It looks to me like the mean exit date from my model is stable, regardless of the shape of the distribution, at least within the range of skew that we have experienced.  If the positive skew did become so excessive that a measurable percentage of potential ZLB exits would happen in the range of the yield curve where the slope levels off to the long term yield levels, it should cause my model to report a future yield curve slope that is slightly understated, and possibly slightly understate the amount of time to the mean.  And some of the skew would increase the measured standard deviation of the distribution of the expected exit date, which makes the yield curve less convex around the exit date.  I don't believe that these distortions were significant.

Now, while both the mean and the median would be useful if you are trying to assess the shape of expected outcomes and risks of taking positions on the yield curve, the mean date would be the more important measure for determining the intrinsic value of forward rate contracts.  So, I believe that my measure is useful.

But, looking at both the mean and the median might allow a more fine interpretation of market reactions to Fed policies since the beginning of QE3.

Here is the mean from my model, compared to the median (estimated from the Fed graph above).

For a period in 2013, the median date of the ZLB exit was farther away than the mean date.  That suggests a negatively skewed distribution, which seems implausible.  I don't have the Fed's data, but either I am misunderstanding something here, or there seems to be either an error in the published chart or something wrong with the Fed data.

y-axis is Eurodollar contract price, which is 100 minus the interest rate.
It is inverted, so that the graph represents the yield curve.
The next chart compares the forward Eurodollar yield curve for the three dates noted in the mean vs. median chart.  On May 1, 2013, forward rates were very low, signifying both a distant exit from the ZLB and a low slope rising from the ZLB.  By September 6, the date had moved much nearer and the subsequent slope had steepened.  By December 16, 2013, the expected date of the ZLB exit had reverted into the future, but the slope of the post ZLB curve had continued to steepen.

Following are three charts comparing each of these specific yield curves, showing the estimated locations of the mean date of the exit from ZLB (from my model) and the median date of the expected exit from ZLB (estimated from the Fed's graph).

(A brief note on these charts. 
The mean date appears to be shifted by 3 months, because these are 3 month Eurodollar contracts.  So, if the first rate increase is in September 2015, the June 2015 contract would be the last contract settled at the ZLB and the September contract would reflect the higher rate.)

On May 1, 2013, we can see the signatures of a positively skewed distribution of expected exits.  A median date that is sooner than the mean date, and a long stretch of curvature that reflects a wide range of expectations.  You can tell from the slope of the linear plot of my modeled forward curve that my model is able to pick up the expected slope of the post ZLB curve from the shape of the convexity, even though the slope of the actual yield curve never quite gets that high, because a measurable proportion of expected ZLB exits occur very far in the future.

By September 6, rates had risen substantially.  My model shows a quicker ZLB exit and a steeper subsequent slope.  Two issues are very clear in this graph.  Because at this point, the ZLB exit had become very near, the yield curve became very convex around the mean date of the exit, nearly mimicking my linear model of future rates.  The mean expected exit has to be at the center of that convex area.

I can't think of a plausible way for the mean and median dates to fall in these ranges in a way that leads to a yield curve with this shape.  I don't see how the Fed's stated median date at this point in time can be accurate.

By December 16, 2013, rates had fallen from the September levels, and the mean and median exit dates return to their expected positions.

Since December, the curve has flattened slightly, and the mean expected exit date has moved back to about August 2015.

Friday, February 21, 2014

CEO Candor

Rittenhouse Rankings is an attempt at quantifying the honesty and transparency of firm management.  Rittenhouse claims to be able to capture significant alpha by investing in firms with the highest scores.

I think this is an interesting idea, because even if these excess gains mainly come from avoiding unforeseen negative shocks to specific stock positions, you could say they come from a vulgar kind of market inefficiency.  If the market really is naïve and tends to buy into management obfuscations and overstatements about short term performance, then some of this alpha could come from mispricings among firms based on this inefficiency.  In other words, honest management causes their share price to decline.  This decline reflects a higher required return, and that return is showing up as alpha in the Rittenhouse portfolio.  Put yet another way, you could imagine this as a factor, with returns being a function of smallness, low market to book, momentum, and management honesty.

I would have hoped that markets would be functional enough that management honesty would create reputation, which would increase a firm's share price over time, manifest through a lower required return that would reflect a firm's earned reputational capital.

But, at least, if this is in an inefficiency, that means there are profits to be had for the discriminating speculator.

My experience has been that this is frequently an issue that can be very lucrative.  Micro caps that are coming off of extended periods of poor operating and share price performance, with managed turnarounds and conservative, honest management, can be very satisfying positions to own.  Frequently, these stocks will languish with very little trading volume, even after the turnaround becomes quite clear.  I used to own shares in Onyx Acceptance Corp, which has since been bought by Capital One.  They were a sub-prime auto loan firm, and they had really taken a hit in the 2000-2002 time period.  They made very extensive public reports of their monthly loan performance, so with just a little work, an investor could get a good idea of the differences between the expected values of their loan portfolios and the book values.  For some time, the market was discounting their book value, even though their new loans were clearly setting up to exceed book value.  Management would report perfectly honest, transparent results, with the tone of a palpable yawn, and take a part of their compensation in options.  Those were fun quarterly reports to get.  It was like knowing a secret hand-shake.  Management was putting it all out there, but they weren't cheerleading.  It just took a little patience to wait for the market to catch up.  It surprises me how those type of opportunities seem to be available to individual investors.

While management candor is good for share buyers, I don't know that we can always say it is good for existing shareholders or management.  There are a lot of agency issues here.  On this kind of topic, in addition to conflicting goals between shareholders, management, and other stakeholders, there are conflicting goals between long term shareholders, short term shareholders, and potential new shareholders.  And, while stock options might be associated in the popular mind with short term hucksterism among management, stock options may sometimes encourage management candor, at least for management that expects to have some permanence.

Thursday, February 20, 2014

You can't analyze market interventions without considering sunk costs. (Yet another minimum wage post.)

Let's consider a new batch of capital investments in a competitive market.  And, let's make some basic assumptions, that while returns are unknown for any individual investment, the required return for this capital is 10%, and the aggregate return over the lives of the investments will be 10%, with a normal distribution around that outcome.

So, here is a picture of the ROIs of these investments:

This is the basic picture of finance.  There is some normal rate of return that emerges from the supply and demand of capital, and nobody knows exactly which operations will produce it, so we diversify investment capital among projects that meet our risk profile.

But, continuing to imagine this basket of investments, the returns don't come from a static set of operations.  Over time, losses gradually are represented by discontinued operations, so that, eventually, this set of assets evolves into something like this:

If the original set of investments represented $1 billion from a single firm, under our original assumption that on average these investments provided the required return, then the remaining operations would represent, say, 2/3 of the original investment.  These assets would show a very high ROI - 20% in the hypothetical numbers that come out of my example.  But, the ROI for the entire original set of investments remains 10%.  The market value of productive assets remains $1 billion.  The original investments in the failed operations and losses incurred during their operating lives are sunk costs now.

In the context of allocating capital within these projects, those sunk costs should be discarded now.  But, in the context of understanding capital and investment more broadly, those sunk costs are still important.

Several characteristics to note:
  • The ROI of individual active projects in a competitive market will tend to be higher than the ROI for the market as a whole.
  • So, seemingly monopolistic competition can arise from competitive markets because of deviations in ex post returns.
  • The excess returns of successful operations may include excess returns from real options embedded in the operations that include follow-up investments.  But, additional other investments may continue to have an expected return of 10%, even though the existing and future pools of active investments will have returns on initial investments higher than 10%.
  • This diversity of outcomes exists throughout a market, within and among firms.  In some ways, the categorization of projects among firms when analyzing the entire market is arbitrary.  Some failed projects will have been a part of failed firms, some will have been part of active operations within firms, and some will have been discontinued within active firms.  The complete ROI on all invested capital in a given market would be difficult to ascertain, and would tend to appear higher than the aggregate experienced return, because of extensive survivorship bias, whether at the project level or the firm level.

So, why is this a post about the minimum wage?

Much of the technical discussion about minimum wage effects appears to revolve around the idea that low wage employers reflect monopolistic competition.

Using the model I outlined above, I can graphically describe four effects of minimum wage increases on a low-wage employer market.  A minimum wage hike would increase costs, creating a first-order effect of lower ROI for all firms.

Immediate effects

1) Lost or reduced operations

Research that picks up employment losses is probably largely picking up this effect.  Added costs would lead more marginal firms and operations to be discontinued or reduced.  As the graph helps to visualize, even in a competitive market with relatively free entry of new capital, this probably tends to represent a small proportion of the set of affected firms at any given time.

2) Higher prices & lower profits

Firms that weren't marginal would initially move to a new equilibrium which might mostly involve continued operations at lower ROI's.  This probably includes most existing employers.  So, research following specific firms would probably pick up this effect, which would appear to point to minimum employment effects.

Long Term effects

The initial effect on operations would be an across the board reduction in ROI.  The combination of more failed operations and lower ROI among the successful operations drops the aggregate ROI.  Remaining operations would continue to have ROI's well above the required return for the market, but without fundamental adjustments in the entrance of new capital, the expected ROI of new capital would be less than the original 10%.

3) Expanded Operations

Existing employers who were not marginal might tend to experience some expansion, first due to the exit of marginal operators, and then due to the reduction of new entrants because of the lower ROI.  This could be the source of expanded employment that some research finds.  But this particular increase in employment would come about only as the result of shifting market share within a shrinking market.

4) Reduced New Investment

The original ROI would have been the result of an endlessly complex set of financial and cultural factors that would have settled on the original expected return.  Lower ROI's would lead to some new equilibrium that would involve some combination of changing capital mixes, prices, etc. which would settle at a new reduced level of investment that could again achieve 10% expected returns.

This effect might get picked up in macro-data time series, but it is probably difficult to see clearly with any method, because it would be part of complex long term processes.

This paper from the Chicago Fed finds increased firm exits after minimum wage hikes, but it finds that, increased firm entries mostly counter the negative employment from the exits in the short run.  So, the first effect above may not lead to extensive loss of economic activity.  This result seems plausible - an initial rush of replacement capital that was previously kept out by marginally productive existing capital.  Keep in mind, however, that this result involves capital destruction from the higher level of firm failures.  The new capital that replaces the obsolete capital in the given market has had to come from somewhere.  So, while in the measured market, there is an exchange of higher wage employment for previously lower wage employment, the capital allocation required in order to create that exchange has some unmeasured cost elsewhere in the economy.  (edit:  Hmm.  Thinking about this some more, if this effect is dominant, it would mean that employment losses would largely come from outside the markets for low wage labor, because of this shift in future invested capital.  It might explain an ironic outcome where measures tracking specific minimum wage employment levels don't show diminished employment from MW hikes, but measures of the broader economy do.)

I have found employment losses associated with the period of time from slightly before an initial federal minimum wage hike to about two years after.  This seems like a quick reaction if only the 4th effect above is leading to significant net employment losses.  But, I suppose the interplay between exiting firms, entering firms, and long term changes in capital allocation could have a net effect that leads to a variety of time series patterns.

In terms of measuring long-term employment levels, a labor market with flexible wages would be expected to grow to utilize substantially all available supply.  This couldn't necessarily be said of a labor market with a price floor.  But, again, the financial and cultural processes would be so complex, that it would be difficult to measure.  I suppose that over the very long term it could be possible for a new equilibrium to lead to widespread employment opportunities for employees formerly working below the price floor.  For normal products, we would expect to see a reduction in the quantity supplied after the implementation of a price floor, but reduced supply wouldn't need to result in only an increase of unemployment.  Changes in schooling, pre-job training, and labor force participation, in general, could lead to fundamental changes in types of labor that people are capable of providing.  Opinions about results that complex are likely to fall back on priors.

Added:  Patrick Sullivan points to this research, which seems to find some of the employment declines from outcome #4, above.

Wednesday, February 19, 2014

2013 4Q Household Debt

Household Debt numbers are out for the 4th quarter, and they continue to look expansionary.  This is a necessary development if momentum is going to carry us beyond QE3.  Here are some graphs (I exclude student loans).  The bottom is clearly behind us now:

Here is the cumulative change, in dollars, since 2010, by category.  This makes it clear how much mortgage debt dwarfs all the other categories.  Also, all categories, except Home Equity, are growing now.  And, all of them are accelerating.

The acceleration is clear in the aggregate Q/Q growth rate, also.  Q/Q growth is now back to the range of a healthy economy.  The question now is whether it stays there or continues accelerating so that we get some bounce back growth that is higher than normal.

New mortgage originations have weakened the last couple of quarters, although they are still higher than 2009-2011.  But, this probably means that most of the acceleration in credit is coming from a slowdown in foreclosures.  Delinquency rates on new credit are generally back to normal, so it would be nice to see some more growth from new credit.  But, in either case, I consider it to be a good sign for 2014 that banks are willing to expand their balance sheets.

QE3 has been adding just under $1 trillion in deposits to bank balance sheets per year, and it looks like household credit is setting up to increase at at least that pace this year.  A question I have had for this year is whether banks are healthy enough to pick up the slack from the QE3 taper.  This report suggests that the answer is yes.

Tuesday, February 18, 2014

What We Know that Just Ain't So and EMH

Scott Sumner has a great piece on the persistence of sloppy ideas about a financial crisis being the cause of the Great Recession.  It seems obvious to Scott (and to me, but I mostly learned it from him) that the Fed caused the bulk of the disaster.  (Well, I would add that the scale of the labor market decline was exacerbated by pro-cyclical labor policies.)  Economists largely agree that monetary policy carries much of the blame for the Great Depression.  As Scott notes about the Great Recession:
After rising at roughly a 5% rate for many years, the Fed brought growth in the monetary base to a complete halt between August 2007 and May 2008. That triggered the onset of recession in December 2007. Velocity actually rose during that 9 month period, but not enough to offset the Fed's tight money policy
Home prices had been falling for nearly two years, relative to rents, when the Fed started pulling on the emergency brake, by the way.  Emergency Unemployment Insurance was first instituted at the end of June 2008, while the Fed was still dithering, worried about potential inflation.  (I don't mean to blame the Fed.  There is only one way to provide the optimal quantity of an important commodity with dynamic demand, and it doesn't involve the creation of a Central <fill in the blank>.  I have a very high opinion of my own market perspective, and if you put me and the 10 smartest people in the world in charge of allocating wheat production, good luck making it to next spring.)

Scott references an earlier piece he did that highlights some apparent contradictions within the standard stories. (here)

Scott frequently mentions his assent to the efficient market hypothesis, and I largely agree.  At least in its weak form, EMH generally holds.  But, this is where it gets muddy.  What is information?  Is it just bits and bytes, lines on a financial document, whispers about new contracts?  Or does it include interpretations?  Does it include conventional wisdom?

What does EMH mean, exactly?  Does it mean that you can't systematically make excess profits where others are clearly being contradictory or dense?  Why not?  There are a number of consensus beliefs about the last decade that seem strange to me, in ways that I can trade against.  Could there be a meta-weak form of EMH?  It would say that you can't systematically beat the market unless you rebel against the consensus, but rebelling against the consensus is like eating from the tree of the knowledge of good and evil - not a history of good outcomes there, for a number of reasons.

For disagreement to be more than a rare occurrence, the human mind really has to be an instrument finely tuned for self-deception.  The topics that people commonly disagree about aren't topics on which we tend to merge to a concensus of shared ambivalence and doubt.  They are generally the topics that we all feel the most certain and exorcised about.  And, we usually share these strongly held beliefs with some group of friendly believers.  Empirically it is clear, even though there is no coherent way to admit it, collectively, to ourselves:  The things we are most wrong about are - in fact, they have to be - the things we feel most strongly about.

We have a bias for the consensus.  We have a bias for the omniscient designer.  We have a bias for blaming powerful outsiders.  We have a bias for defending the meek.  To be otherwise is to be untrustworthy.  I think the consensus about the causes of the Great Recession is riddled with logical and empirical errors, but what it has going for it are all the right biases - careless greedy bankers, helpless workers and homeowners, a Central Bank heroically pulling us through the dark times, policies of one political party that would have solved the mess years ago if the other party wasn't diabolical, a put-upon Main Street and a rapacious Wall Street, etc.

Being human means happily accepting the consensus when it is wrong.  Beating the meta-weak efficient market means systematically seeing through the haze of contradictions of the consensus.  But, we are such imperfect vessels in a highly complex world.  As Walt Whitman said, "Do I contradict myself? Very well, then I contradict myself, I am large, I contain multitudes."  If this can be said of us, it can certainly be said of the consensus.  But, the consensus, imperfect as it may be, also serves as a disciplining force.  For every contradiction that it feeds us, it corrects a hundred others.  If we imagine ourselves above the consensus, judging its foibles, who do we allow to judge our own contradictions?  With even a subtle lack of self-awareness, this becomes sociopathic.  I submit that it is possible to walk this tightrope, but with many trips and falls along the way - too many pitfalls to disprove the null hypothesis.

We pooh-pooh the herd, even while we largely live inside one herd or another.  We call them sheep, lemmings.  But, herds survive.  Zebras aren't playing stupid just to be with the cool kids.  The herd serves us, even though there are large untrampled, bountiful meadows out there, just over the hill.  You can have them all to yourself, and there usually aren't any lions around, until you're full and sated, almost, and decide to waltz out and dip your head under the tassels for one more scrumptious bite.

The other zebras are EMH'ers.  They say, "There is no way that you can be sure to get to that meadow and get back here alive."  But, I say, "The meadow is right there.  It's gorgeous.  I can see it.  Surely I can go sneak a meal every now and then, if I'm careful."  It doesn't take a genius to see that we're all trampling around in our own shit, does it?

I don't know if efficient markets will eventually force me to succumb to a final blow of bad fortune, but if they do, I can say with certainty that it will be over a position that I will have been very confident about, highly exorcised over, and insistent that it was logically sound and safe as can be.  After the fact, I will say, "That doesn't prove the EMH.  I was just being an idiot on this position.  Man, I wish I had just managed to hold on to a few dollars so I could keep trading.  Did you hear how wrong everyone on CNBC was this morning?"

The EMH says all available information has already been priced into the market.  But, I think it's more a matter of having two choices, (1) believing all the insane, wrong stuff everyone else believes, or (2) believing in your own insane, wrong stuff.  How could we expect any other choices?  And, how do you plug that into a statistical model to confirm whether (2) can reliably succeed over (1)?

Thursday, February 13, 2014

Unemployment and Interest Rates

Unemployment claims seem to be reaching a trough level.  Here is a graph from Fred:
FRED Graph
Blue: Initial Unemployment Claims, scaled to Labor Force
Purple: Insured Unemployment Rate
Green: Unemployment Rate (scaled by 1/2)
Red: Slope of Yield Curve (5 Year - Fed Funds) - right scale

I thought there were several interesting relationships here.
1) The difference between the Insured Unemployment Rate and the level of Initial Claims is a kind of proxy for duration of unemployment in the regular UI program.  This relationship isn't affected by the unusual long-duration behavior we have seen this cycle.  Durations among this group appear to be entering the mature phase of the recovery.

2) Unemployment is very high compared to insured unemployment.  This gap will mostly close over the next 6 to 12 months, giving the real economy a boost.  (speculative)

3) The flattening of the trend in unemployment claims has recently (by recently, I mean the last 25 years) been associated with a flattening yield curve (a drop in the red line).  Most of this flattening usually would mostly be the result of rising short term rates.  But, the current rate environment is difficult to compare to other cycles.

Rates might be rising earlier than many suspect, with the Fed adjusting interest on excess reserves upward as a secondary tool along with selling treasuries.

Wednesday, February 12, 2014

Quick Follow-Up on the Beveridge Curve (updated)

I always realize that I left out some useful item after I make a post.

For point of reference, here is the Beveridge Curve for 35-44 Year Olds.

Compare that to the two age groups I posted yesterday:

I also wondered, after I posted the original material, what effects population changes or labor force participation changes might have on this output.  The Boston Fed paper, by Rand Ghayad, used the unemployment of each group, expressed as a percentage of the entire labor force.  This was useful in their analysis, but it does allow for changes in labor force levels to affect the appearance of these Beveridge Curve outputs.

Population and labor force changes don't affect the 20-34 year old group.  But, they do have a small effect on the 35-44 and 45+ groups.  The 35-44 group Beveridge Curve would shift right by about 0.1% and the 45+ curve would shift left by about 0.15%.  So, the scale of the unusual shift among 45+ year olds is reduced by about a quarter. (See update below.) The general pattern remains, however, and this adjustment would seem to strengthen the bifurcation between job losers and job leavers, since it reduces the small Beveridge Curve shift that the Boston Fed did find among older job leavers.

The labor force shifts among 16-19 year olds has been large enough to change the character of the shift for them, however.  Adjusting for labor force changes makes the 16-19 year old pattern look more like the 20-34 year olds.

I know I'm a broken record on these issues, but I think this strengthens the argument that the effect among the 20-34 year olds (now among 16-34 year olds) is minimum wage related.

Employment loss and recovery showing up in 16-19 year olds earlier than in the other populations, and for much of the 16-19 year old employment losses to be reflected in changing labor force participation is typical pattern for the few federal minimum wage episodes that we can analyze.

The good news is that these groups are recovering as the minimum wage level is reduced in real terms over time.

PS. (Added)  Here are the Beveridge Curves for the 3 main separate groups:  The young group where the extra unemployment is from job leavers or entries, the middle aged group that didn't see much of a shift in the Beveridge Curve using the Boston Fed's measure, and the older group where the extra unemployment was from job losers (which have had especially long unemployment durations in this cycle, coincidentally with the very long Unemployment Insurance benefits).  In these graphs, I have used age-specific labor force levels in the denominators so that changes in unemployment would not reflect relative changes in age-group population levels.

y=Openings Rate, x=unemployment rate
The young group saw a shift right in the Beveridge Curve, which has been partially reversed, with a little more recovery to come.

y=Openings Rate, x=unemployment rate
The middle group had a shift in the Beveridge Curve, which was not apparent with the Boston Fed's methodology.  It has partially been reversed.  The effect still appears to be smaller here than in the other age groups.  Because the effect doesn't show up when the Fed separates the unemployed according to reasons for unemployment (job losers, job leavers, etc.), it is difficult to know the weight of the shift that can be attributed to each type.  The shift in this ratio from this age group accounts for less than a quarter of the total excess unemployment, so the difficulty of addressing this age group in the Fed paper would only make a small difference to the aggregate picture.
y=Openings Rate, x=unemployment rate

The older group saw a large shift right in the Beveridge Curve, which has recovered slightly.

I predict that the young group will continue to slowly merge toward the pre-2008 trend.  The older group will merge more quickly toward the pre-2008 trend, and will account for much of the lower unemployment rate to come over the next 6 months.

By around summer 2014, we'll see unemployment at 6.0% or less and the leading edge of the Openings/Unemployed ratio for the older age group in the same range as the pre-2008 ratio.

There, I've put the gauntlet down.  Either, come summer, I'll be proven right, or there will be some previously unknown development that will have invalidated the test of the forecast.  ;-)

Tuesday, February 11, 2014

December JOLTS, and Very Good News Regarding the Beveridge Curve (updated)

JOLTS continues to indicate a tepid but consistent labor recovery.  The strange drop in short-duration unemployment from the December Household Survey doesn't seem to show up in the JOLTS data at all. Here are updated graphs:

Below is the Beveridge Curve - the relationship between job openings and the number of unemployed workers.  After a decades-long shift to the left, the relationship began to shift to the right in 2008.

I have attributed this shift, partly, to the sizable minimum wage hikes in 2008 and 2009 and to extended unemployment insurance.  EUI was discontinued at the end of 2013, and the minimum wage level, compared to average wages, is trending back to 2008 levels.  So, especially because of EUI, I expect to see a shift back toward the 2008 curve position over the next 6 to 12 months.  It looks like we have already started to see this reshifting in December and January.

Added Postscript:
Here is a short paper from the Boston Fed, which disaggregates the Beveridge Curve by age and by reason for unemployment.  It's findings are very interesting.  There is a bifurcation of results.  Many of the subpopulations show essentially no shift in the curve.  As of January 2013, the additional unemployment, relative to job openings, was mostly the result of 2 groups, split roughly 50-50:
1) Job leavers, new entrants, and re-entrants, between 20 and 34 years old.
2) Job losers, 45 years and older.

The second group is the group where we would expect to see a shift, if extended unemployment insurance were the culprit.  In some of my past posts, I have noted how the inflated long term unemployment levels that might be attributed to EUI seemed to be heaviest among the older age groups.  The relationships shown in the paper suggest that EUI may be even more skewed toward older workers than my basic model estimated.

This also pretty closely corroborates my estimates for the excess unemployment attributable to EUI.  I had attributed excess unemployment topping out at 1.2% in mid-2011 and falling to 0.9% by January 2013.  The Boston Fed paper shows a shift in the Beveridge Curve attributable to job losers topping out at 1.9% in mid-2011 and falling to 1.1% by January 2013.

The puzzle for me is, why is the other half of the explanation for the shifted Beveridge Curve concentrated among 20-34 year old job leavers, new entrants, and re-entrants?  There could be a number of reasons, none of which seem like good news.  The most hopeful cause would be the minimum wage, because at least that would be an easily identified, self-inflicted wound.  But, the shift does not show up in the 16-19 year old data.  Now, some funny things appear to happen in labor force participation in these minimum wage episodes, so that doesn't count MW out as a culprit, but it makes it harder to blame MW with confidence.

The gap from the old Beveridge Curve in January 2013 was about 2%.  About 1% of this was from 45+ year old job losers and about 1% was from the 20-34 year olds.  The gap had maxed out at nearly 3% in late 2011, and nearly 2% of the gap was due to the job losers.  But, the gap coming from the younger workers had been persistently around 1% since mid-2011.

To update the data to December 2013, the closest data I could find to the authors' were age-based unemployment levels, without the distinction for reason for unemployment.  But, I can use this data to get an idea of what has happened since January.  What I find is very good news.  Here are the updated Beveridge Curves for 20-35 Year Olds and for 45+ Year Olds.

The good news is that both groups are making strong moves back to the earlier Beveridge Curve trendlines.  In my data, in January 2013, the 20-34 Year Olds accounted for 0.8% and 45+ Year Olds accounted for 1.0% of the excessive unemployment rate.  By December 2013, the excess rates had decreased to 0.5% and 0.8%, respectively.  Employment seems to be recovering in both groups.

In the most recent months, my independent estimate of the excess unemployment due to EUI, from comparing short duration and long duration unemployment, is 0.9% in December 2013 and 0.8% in January 2014.

Unemployment has been declining, parallel to the previous Beveridge Curve for the past two years, but I think we are going to see a reversion back toward the old curve this year.

(quick follow-up)

Review of Tricky Issues in Labor Force Participation

This post from the Mercatus Center offers a good overview of the complications that arise from adjustments that need to be made with analysis of Labor Force Participation.  They have adjusted for age, which is an important first step.  Here's a graph they use, with my notations added:

1:  This is the one group with a truly unusual drop in LFP.  Some of this is a part of a long term cultural shift toward longer time spent in education.  But, the shock specific to the current time period was the 3 hikes in the minimum wage from 2007 to 2009.  In 2006, about 4% of the 16-24 year old labor force worked at or below the federal minimum wage.  By 2010, that was up to 10%.

Here is the 16-24 year old LFP rate, from the BLS:

Note how LFP in this age group dropped precipitously from 2007-2009, and has been flat since then.  This is a distinctly different character from the other age groups.  Minimum wages affect a larger proportion of this age group compared to the other age groups.

The unusual decline in this age group can be attributed to the minimum wage, but that shock to labor demand ended 4 years ago.

2:  The movements in the 45+ age groups have, on net, been positive, and mostly reflect the undulations of boomer populations through these groups and the tendency of baby boomers to work later than earlier generations.  So, this group has seen an anomalous net increase in labor force participation during a cyclical downturn.

3:  This group represents the heart of the labor force, and the age group LFP declines appear to still be damning, even when looking at the age groups.  The average LFP drop here has been about 1.9% since 2007.  But, there are additional adjustments that need to be made.

A) There has been a long-standing secular decline in the LFP of each age group of more than 1% per decade, which was temporarily countered by the increase in female participation which lasted until the early 1990's.  It has now been 6 years since 2007, so even within age groups, we should expect trend LFP to have decreased by about 0.7% during this time for reasons unrelated to short term cyclical or structural issues.

B) The 2007 labor market was very strong and LFP at the time was significantly above trend.  This was not commonly recognized because the trend was declining due to the aging workforce.  But, since the 2005-2007 period is widely recognized as a boom-time, it should not be difficult to believe that LFP was at least 0.5% above trend.  We could have experienced a shallow recession with unemployment topping out around 6%, and an age-group specific decline of 0.5% back to trend would have been reasonable.  This is roughly what happened during 2001-2002.

These two adjustments leave us with about a 0.7% decline in LFP, below trend, which can be attributed to short term issues.

Long term secular declines as well as demographic issues will continue to weigh down on LFP at a rate of up to 0.3% or so, annually, for the time being.  So, if LFP simply stops declining, much as it did from 2003 to 2007, we will be back above trend by 2016.

Obamacare, or some other structural issue, could prevent that kind of recovery in LFP.  But, in assessing those effects, we need to be careful in adjusting LFP for all of these other issues.  The fully adjusted LFP rate is, at most, a few tenths of a percent lower than normal cyclical fluctuations would portend (plus including the effects on the prime age groups of the minimum wage hikes that ended in 2009 - which might account for those few tenths of a percent).  The status quo should lead to an annual decline of something near 0.3%, together with a cyclical recovery of at least 0.7% over the remaining recovery period of the cycle, plus any additional fluctuations above trend if the recovery remains healthy.  Deviations from that trajectory could be attributed to labor market problems.  But, we should be careful to measure appropriately.

Saturday, February 8, 2014

The End of EUI in North Carolina and Unemployment (Pt. 2)

Andrew Hofer linked to my last North Carolina post on the change in unemployment, and noted "These suggest a higher level of work discouragement than I would have expected."  That made me realize that I had been lazy with that post.  I should have also included labor force participation and employment-population ratios.

Here was the histogram of the change in the unemployment rate, by state, from July to December 2013.

Below, I have added the histograms for the labor force participation rate and for the employment to population ratio.

LFP has declined slightly more than average in North Carolina, but EPR has increased slightly more than average.  Both of these changes would create a lower unemployment rate.  Roughly 2/3 of the excess relative unemployment decline came from a relative increase in employment and roughly 1/3 came from a relative decline in labor force participation.

 I don't think it is appropriate to identify all of these labor force exits as "discouragement" without more information.  The comparison to the aggregate national numbers also suggests that more of the recent labor gains in North Carolina have been due to employment growth.

The end of EUI in North Carolina was coincident with a steeply falling unemployment rate.  It looks like most of that has been related to relative employment growth, but there could be a range of possible weights.