Thursday, July 12, 2012

Gaming The System Versus Designing It

The main point in this piece is that many people have become quite good at gaming the system but few understand what a good system design looks like.  Here I'm talking about social systems - whether driving on the highway, student and teacher behavior in school, citizen and legislator behavior in state or local government, or any other such system.  People have learned to play the game to their own personal advantage and do so without thought of whether their behavior is beneficial, benign, or deleterious to others.  Indeed, a further point of this piece is that quite often we don't know how to evaluate the outcome socially, so instead build a plausible (but possibly quite incorrect) narrative to give a thumbs up or thumbs down about the social outcome.  This second point I'm borrowing from Daniel Kahneman's Thinking Fast and Slow, where he argues that when we can't answer a difficult question because of the complexity involved we replace it with a simpler question, answer that, and then assume it is the answer to the original question as well, even when it is not. 

Innovation often aids the gaming behavior.  Consider radar detectors and driving, perhaps the least controversial example one might come up with.  Presumably a radar detector is purchased for one reason only, to lessen the likelihood of getting a ticket for speeding.  Armed with a detector, the motorist will drive faster, as long as the detector indicates there are no police in the vicinity.  If it is not the absolute speed of the vehicles that matters for safety (above a certain minimum speed, say 50 mph) but rather the variability in speeds that most impacts safety, then it seems clear that the detectors increase variability.  Those who have them will drive faster than the rest.  The impact on safety is unambiguous.  The detectors lower safety.  However, in a standard cost-benefit analysis that economists do, one must acknowledge the benefit to the drivers with the detectors.  They get to their destinations sooner.  So there is a time value to them.  What is the result in aggregate?  I really don't know.  What does seem clear, however, is that there is an equity issue with gaming.  Other motorists bear some of the safety risk but get none of the time value benefit.  So, conceptually, the analysis is straightforward but on the actual arithmetic, adding up the benefits and subtracting out the costs, who knows?   On a personal note, I'll make the following additional observation.  If the possessor of the radar detector is a well paid professional (hence someone with high time value), middle aged, and without serious mental or physical health issues, then I believe their own self-preservation instincts mitigate the safety risks.  On the other hand, if the possessor of the radar detector is the teenaged child of such a well paid professional, all bets are off.

Let's move on to a harder example, standardized test prep courses a la Stanley Kaplan.   When I took the SAT, in 1971, I believe the College Board's position was that the test prep was of no consequence with regard to the results.  My brother bombed the PSAT, took a test prep course thereafter, and his scores went up by over 300 points on the SAT.  That's only a single observation but based on that and what I have gleaned since, there is definitely real consequence from giving the student confidence about the strategy to use when taking the test, whether to guess or to not answer a question when the student is unsure, and then becoming proficient in implementing that strategy.  There can also be a real benefit from taking practice exams, doing diagnostics on those, and then cramming in areas of weakness where some concentrated study might help. Since each student should have the opportunity to put his best foot forward, some might not even consider private test prep as gaming the system.  Do note its only commercial test prep done outside of school that is being considered here.  Lower income students who can't afford commercial test prep therefore don't get its benefit.  So, again, there is an equity issue with regard to the behavior.  Leaving that aside, where is the social harm from the practice?  I'm not sure everyone would agree, but this makes sense to me.  If you treat the performance on the test as signal, an indicator of a hidden attribute of the individual which we might call "intellectual ability," then it might be that the test prep makes the signal less informative.  Students with good scores might have high ability, but perhaps their ability is closer to average and the score is more a consequence of the private coaching the students have received.  If that's true then schools in their admission practices either end up making more mistakes regarding whom to admit as a consequence of the test prep or the schools must incur substantial costs to consider other indicators of the students' ability, so as not to make those mistakes.  That's where the social harm is.

Before pushing on, I'd like to refine the above based on my perspective as an instructor.  The expression intellectual ability may not convey the right meaning.  What one really wants are certain habits of mind and that the student reads on a regular basis stimulating and challenging material of his own choosing.  Below is an excerpt from a post I wrote soon after starting this blog in 2005.  It indicates that frequently we fall far short of this requirement.   Further, I should observe that this requirement cannot be gamed the way a standardized test can.  It requires substantial deliberate practice

In particular I want to consider information literacy and its importance in the curriculum. As a teacher, I have to say that "old fashioned" literacy is more important to me. I'm of the mind that many of my students don't get the meaning from a New York Times story. I've tested that proposition on occasion with articles I've picked and assigned to the class, either from the Business section or the Magazine. I don't talk about this issue much if at all (except with a particular colleague who teaches Natural Resource Economics who agrees with me fully on this proposition). And I haven't seen it discussed, but it seems to me to be at the heart of the matter.

Students need a well trained "voice in their head" which argues propositions, including what they read. They need to disagree with things when they don't add up, but they need to be able to "get it" without undo difficulty when the meaning is straightforward. It is a reasonable expectation (in the normative sense) that students have these abilities when they enter college. But, I fear, all too many of the students falter here. Because these kids are bright, I'm going to say the culprit is they don't read enough and so this habit of arguing with the voice in their head is not well cultivated. This is a real problem. I don't have a great solution for it, other than that the kids need to develop the habit of reading and to think of reading as internal argument.

* * * * *

The innovations discussed above happened some time ago.   By focusing on the past where we are familiar with how the innovations have been subsequently utilized we can consider the impact of these innovations over the fullness of time.  I don't know the extent of the radar detection business since the 1970s (during the second OPEC price shock many states lowered their speed limits and that gave a boost to radar detection but when the price of gasoline came back down the speed limits went back up), but clearly test prep has been a growth industry, at least till the recent downturn in the economy.   Let's now focus our attention on past innovation in financial engineering.  In particular, I want to look at junk bonds and the subsequent behavior that engendered as well as the securitization of residential mortgages and the consequences from that, from the lens of whether they enabled gaming the system, and if so how, or if alternatively they promoted economic efficiency and then describe the mechanism by which that occurred.

There are a several reasons for doing so.  First and foremost, these innovations in financial engineering are what many people associate with the economic consequences of the Reagan Revolution.  Ironically, both came into being during the Carter Presidency, though they clearly were popularized under Reagan.  Michael Milken, who worked at Drexel Burnham Lambert, invented the junk bond.  Since about 10 years later he was convicted of securities fraud, in the minds of many the junk bond concept is highly suspect.  I hope that readers can suspend judgment in reading my arguments below.  In contrast, Lewis Ranieri, the inventor of securitization, had a largely intact reputation, at least until the subprime crisis unfolded.  A second reason is that in preparing for my class this fall, which will use the text Economics, Organization, and Management by Milgrom and Roberts, in chapter six that introduces the concept "moral hazard" there is a rather extensive discussion of the economics behind the Savings and Loan Crisis of the 1980s.  I thought their straightforward analysis was a good way to consider both the gaming the system and efficiency questions. The third reason is to consider the issue of resolution of underwater mortgages and the impact securitization has had on such resolution, especially in the case where an entire community is beset by such mortgages and hence where there isn't a market for repurchase of the homes at discounted prices.  I doubt the issue was considered at all in the 1970s, when Ranieri invented these instruments.  Yet it is something to think through now, since its resolution clearly matters a lot moving forward.  Joe Nocera had an interesting column about this yesterday.  He takes the point of view of the homeowners, which is natural.  One should be sympathetic to them.  But what of the system overall.  That system includes those who hold the mortgage securities. What is a good outcome for the system taking into account all interests?  The last reason is that by thinking through the gaming the system versus efficiency arguments, one can get a better feel for how much regulation in this sector is appropriate. 

Before getting to this, let's take a brief look at the macroeconomics by looking at the following two graphs.  The reason for doing this is to ask whether it is possible to "see" the efficiency consequences of a change in the economic environment by looking at GDP growth.  One might conjecture that if an important practice makes industry more efficient, that in turn would make the economy grow faster.  Alternatively, one might expect that improved efficiency would reduce the volatility in growth rates.  Then too, by looking at growth rates, one entirely abstracts from the distributional consequences of the change in environment.  So much has been written about income (and wealth) inequality as of late that I don't want to take it on here.  This is not to say those issues are unimportant.  It is only to say that there is enough on our plate to look at these other issues here.

To generate this first graph I took historical data from the Bureau of Economic Analysis available as an Excel workbook, with a series that measures real GDP in constant 2005 dollars.   For every pair of consecutive years, Y0 and Y1, I computed GDP(Y1)/GDP(Y0) - 1, expressed that as a percentage and then called that the growth rate in Y1.  Though Edward Tufte probably wouldn't like what I did next, I had Excel connect consecutive plotted points with straight line segments.  I found this easier to read than looking directly at the scatter without the line segments included.  Let me make a few observations, simply from reading the graph.  There are multiple periods where the growth rate is negative.  These are recessions.  (There were 2 in the 1950s, none in the 1960s.  Note that this chart uses annual data.  Typically when we speak about recession we use quarterly data and it requires two consecutive quarters of negative growth.)  Usually when the recession ends, there is a brief period of very high (over 5% real) growth.  Much of that is catch up for the lost time during the recession.  Then growth calms down to a more modest rate, until the next trough begins.  Also, note that external "shocks" can trigger a recession rather than simply the business cycle doing its thing.  In the 1970s, there were two OPEC Oil Price shocks.  The first in 1973-74.  The second in 1979.  This was the period known as stagflation.  Paul Volker pursued a tight money policy to wrest the inflation out of the economy.  The created another recession in the early 1980s.  The next recession, mild as indicated by this graph, happened during the 1992 election season.  You'll recall the the expression coined at that time, "It's the economy, stupid."  I trust that people know the rest of the history well enough to explain the remaining portions of the graph. 

GDP growth can be broken up into two components - population growth and productivity growth.  To abstract from the former, one also looks at per capita GDP growth.   This next table was generated from different data provide by the Census.  It is from the spreadsheet on Per Capita Income All Races.  The same sort of process was used to produce the growth rates.  It doesn't go back quite so far in time, but it covers the years that are relevant for this discussion.  The periods of negative growth in this table are what most of us think of as a recession, though that is not the official definition.  Using nothing more than an eyeball test, the cycle under Reagan looks quite similar to the cycle under Clinton.  The prior cycle, roughly from the end of Nixon to the end of Carter, has the same amplitude as the other two, but had shorter duration.  The cycle under Bush II had less amplitude, even before the financial crisis.  And the recession that started just before Obama took office had a lower trough and has not yet concluded. 

Let me make one more point before getting to the financial engineering.  These data only show the actual history.  They say nothing of the road not taken.  One might argue that without the move to deregulation, that Carter started but that was accentuated under Reagan, the economy would have performed much more sluggishly, especially in light of the threat from international competition, particularly from the Japanese.  That's a conjecture.  It might be right, but we really can't say.  A different conjecture is that it would have performed pretty much the same as the PC revolution would have happened anyway and that was a big driver of the economy under Reagan.  Likewise the rise of the commercial Internet gave a big boost to the economy under Clinton, irrespective of the policies his administration pursued.  This alternative conjecture suggests that much of the economic growth would have happened regardless of of which party controlled the White House.   I don't know if either of these arguments is correct.   All that the data allow you to compare is different historical periods.   The data also don't allow you to conclude whether the policy consequences are fairly immediate or, in contrast, only result after some very long lags.  It may be that the slower growth under Bush II is attributable to the deregulation under Reagan.  It's possible.  The data don't let you conclude that.

Let's get back to financial engineering and begin with the Milgrom and Robert's analysis.  Thee are three players - depositors, the S&L itself, and the government which provides deposit insurance.  The analysis shows the following, intuitive result.  When taking on an investment project that has both upside and downside risk, when the S&L faces limited downside risk (with the government bearing the bulk of that) then the S&L has a preference to invest in projects with a high upside regardless of the expected return on the project.   The moral hazard that Milgrom and Roberts discuss occurs in an example where the safer project also has higher expected return, but the S&L opts for the riskier project, thereby creating an expected social loss, which is born by the taxpayers. 

Let's next note that for any investor, financing a project with junk bonds in the presence of limited liability from the bankruptcy laws is very much like the situation for an S&L with deposit insurance.  There will be a preference for riskier projects, to take advantage of the big upside, since the investor is largely protected from the downside.  It's other creditors who will bear that risk.

Now the new wrinkle, an important one.  It may very well be that from the social perspective insufficient risk per project is taken.  Consider this piece about cancer research over the last forty years, which makes exactly this point.  Individual researchers, who have their own reputations to care about, take on safe projects so that they have a very good chance to succeed, thereby enhancing their own reputation.  But the field as a whole advances little as a consequence.  Let's bring this thought back to the investment context.  The efficiency argument about the hostile takeovers that the junk bonds enabled is that entrenched senior management, which received substantial perqs from the status quo, worked to preserve that rather than to invent the next great thing and engage in creative destruction of the old.  Further, the traditional motive for innovation, from competition in the product market, was muted in the early 1980s because of the oligopolistic nature of industry and the tacit collusion between the players.  (Think of the Big Three auto makers.)  Holmstrom and Kaplan use this argument to explain the merger activity and other changes in corporate governance that were widespread in the 1980s.

Let's stick with this framework as is for a bit before going outside it and critiquing it.   The incentive for hostile takeover or making other risky investments is driven by the potential upside, not the expected return.  When the expected return is high, the behavior can be socially advantageous, especially if the risks are independent going from one industry to the next.  When the expected return is low, the behavior will be socially deleterious and then amounts to gaming the system.  The same behavior can be one or the other.  There is no telltale sign up front which it is.   It may be that over the course of the business cycle early on there is a comparative abundance of risky but high expected return projects.  Later there may be comparatively few of those but more high risk and low expected reward projects.  Thus the gaming of the system is apt to speed up before the next trough.  It's what feeds the bubbles and makes the trough deeper.  Knowing this, a regulator would want to be pretty hands off coming out of a slump, but then start to apply the breaks as the economy improves and become heavy handed when the gaming behavior seems to be on the uptake.  If it is not possible to apply a time sensitive approach to regulation - the law is the law, the enforcement is the enforcement, some choice has to be made between more rapid growth out of a trough accompanied with more rapid plunging into the next trough, versus perhaps lower growth overall but less extreme in the business cycle.

Now some critique. I offer up three distinct criticisms.  Quite possibly more could be generated.
  • Reform the company or gut it?  Reform of a company does not occur with a flick of switch.  It is hard work to change the corporate culture.  It may take substantial new investment.  It likely requires an extended duration to put the new business practices into place and make them effective.  Perhaps in some cases the efficient solution is to gut the company - its current assets on hand exceed the expected value to be obtained from reform.  The issue here is whether the hostile takeover approach gets this determination right or if there is bias in it.  On this I think it is telling to read this George Gilder quote from Miliken's Wikipedia entry offered up to defend Miliken from his critics.  "Milken was a key source of the organizational changes that have impelled economic growth over the last twenty years. Most striking was the productivity surge in capital, as Milken … and others took the vast sums trapped in old-line businesses and put them back into the markets."  To me this reads - if a company has a lot of cash on hand, it is a good target for acquisition. That money should be put into circulation in new business.  Observe that many quite successful companies today, particularly the big technology firms, have a practice of keeping a lot of cash on hand.  If that really was the main criterion for takeover (and takeover threat) it readily could be biased toward too much gutting activity and insufficient reform of businesses that should have been sustained.
  • Reactions at other companies?  It is well understood that many other companies which ended up not being acquired adopted defensive practices to avoid being taken over.  Thus poison pills and golden parachutes became part of the lexicon.  I'd like to focus on a different adjustment - the movement away from defined benefit pension plans to 401K plans. I've written about this elsewhere in an essay entitled Rethinking The Social Contract.  There I argued that we should return to the defined benefit pension approach, but make the plans balanced in an actuarial sense.  People have confused the generosity of benefits from the issue of who should bear the risk in the returns to savings.  Retirement savings should be insured.  Defined benefit plans do that while 401K plans do not.   The move to 401K plans surely has been pernicious for society as a whole.  It has lowered the personal saving rate.  It has substantially weakened the bond between employer and employee.  And it has created a large population on the verge of retirement yet unprepared for it, because employees have been myopic with regard to their contributions.  (They don't contribute enough early in their careers and can't make it up by contributing more later.)
  • The wrong tool in a flat world?   What does the threat of takeover do in industries where there is vibrant competition between the players?   A good argument is that the threat is pernicious in that it forces incumbent firms to overly focus on near term profitability, especially when longer term projects require substantial lumpy investment and hence demand that they build up of piles of cash as discussed above.  So these firms eschew the longer term projects and invariably shorten their own half-life as successful companies as a result.  Perhaps there was a one time benefit in the Reagan years from the spate of takeovers that took place then.  But in the presence of tough external competition, under performing firms will fail of their own accord.  Yet these possibilities engendered by leveraged buyouts still exist.  Why?  Here it also should be noted that balance sheets are a very limited way to measure the likelihood of future success in an industry.  There has been a historical tension between MBA leadership in firms and engineer leadership. The financial engineering tools have tilted the balance in the wrong direction. 
Let's wind up this piece with a consideration of securitized mortgages versus mortgages held by local lenders as it pertains to the issue of what to do with underwater mortgages, where the value of the loan exceeds the market value of the house.  When this is true for an individual home in a community where otherwise the mortgages are healthy, the efficient solution is to have the current owner default on the loan, have the property assumed since it was collateral for the loan, and then have the property resold at the current market price.  In this solution most of the loss is borne by the lenders.  The current owner bears the loss to his own credit rating, including making it harder to obtain a subsequent mortgage.  But the financial loss is borne by the lender.  At the outset when the loan was originated, the lender indicated a willingness to assume the risk of such loss by making such a loan.  That's why in the old days mortgages were restricted to 80% of the selling price or, if a 90% loan was issued, then mortgage insurance was required by the lending bank.  Neither of these practices eliminate the possibility of underwater mortgages, but they do make them less likely.

When an entire community is beset by underwater mortgages, there is no market for resale of the home and the efficient solution probably entails the existing owners staying in their homes.  If the mortgages were held locally, this solution could be obtained by a negotiation between the lender and the homeowner to a new mortgage that makes sense in the current environment.  There might be some delay getting to that new mortgage as the result of the haggling between the lender and the homeowner over the terms of the loan.  But given those terms, each party has incentive to move to the new mortgage asap, for fear that the old mortgage will become under performing and then the lender will assume the property with no ability to resell it.  That outcome would be unfortunate and benefits nobody.  Thus, if the mortgages were locally held, one might envision that it would take a while to get the first few mortgages renegotiated but thereafter many of the mortgages would be renegotiated quickly as the new terms became more standardized.

The situation is quite different with securitized mortgages.  There is first the question of the extent of reinsurance (credit default swaps).  If a loan renegotiation triggers a payment on a credit default swap, then those obligated to make such payments have incentive to block the renegotiation if they can.  There is second the issue of whether the valuation technique in the securitization accurately reflects true loan value, inclusive of the likelihood of renegotiation, or if it is more an artistic rendering tied principally to the face value of the original loans.  If there is this artistic valuation aspect, the holders of the securities may have incentive to block the renegotiation, since that would force downward the securities asset values.  There is third the issue of whether the terms of renegotiation of loans in one community would be independent of the terms so negotiated elsewhere in the country.  If rather than independence early settlement in one community created a benchmark for settlement elsewhere then the security holders would have incentive to block negotiation in that initial community unless the terms were quite favorable to them.  Taken together, these reasons suggest that securitization is a significant force in blocking getting to a sensible solution and hence is a significant factor in prolonging the economic slump.

Now let's unfold this via a backwards induction.  Miliken, in the NY Times piece about him that is linked above, argues that the sub-prime crises was entirely attributable to a decline in underwriting standards (making 100% loans and not verifying the income of the new homeowner).  But he does not ask whether underwriting standards can be controlled by the market or if they will deteriorate naturally over the course of the business cycle and further whether securitization facilitates their decline, by masking the riskiness of the loans.  Even absent this masking from combining mortgages,  at the local level there is an incentive for appraisals to come in high and for the lenders to want to make the selling price high, especially for loans with points, because the originators of the loan profit more this way.  Full securitization of the mortgage undoes the incentive for the local lender to play the role of impartial monitor of the loan and thereby to not make loans that are likely to go bad.  This is a problem that can be anticipated.

Yet a system of partial securitization, where some of the loan is locally held and the remainder is resold, means there is a lower volume overall in mortgage securities.  Since these instruments seem so profitable in good times, the market wants 100% securitiztation then.  This myopia can then be seen as a reason to regulate and also provides a sense of the sort of regulations that are necessary.

The gamers of the system can move faster than the designers of it.  System design is always a compromise between competing ends.  Gaming can be pure towards a sole purpose.  That much should be understood about any system.  These thoughts should be with us and it is why a total embrace of Laissez-Faire is extremely dangerous.  It empowers gamers and allows them to do much damage.  Our debates on the matter should be about comparing different forms of imperfection rather than about a search for the elusive and the impossible.


Lanny Arvan said...

Some validation for my first example. Read the comment at the link that appears after the piece.

Lanny Arvan said...

Oops, second example.