Wednesday, July 29, 2009

Knowing, Determinism, and Ball Four

I recently saw Knowing, a very entertaining if occasionally silly film starring Nicolas Cage. The film’s plot is bizarre. Somehow, a little girl living 50 years ago is able to predict every major disaster on the planet, leading up to a potential extinction event in the modern day. And, of course, it’s up to Nicolas Cage, finder of National Treasures, roller of Snake Eyes, and the lover of Carla (the prom queen), to save the world.

The central theme of the film is the issue of determinism – that is, the notion that everything in our lives, and in everyone else’s life, is essentially predetermined. You can call this fate, I suppose. This notion stands somewhat in opposition to free will, which is the idea that we “make our own destiny.” Of course, such a weighty topic deserves serious thought, and while the film doesn’t really offer a nuanced take on the issue, it does leave you thinking after you turn off the TV.

I got to thinking about what my view is on the issue. As you may know, I am someone who believes very much in probability and randomness, which might suggest that I fall on the side of free will over determinism (to create a somewhat oversimplified dichotomy). But in thinking about it a bit more, I’m not really sure where I lie.

As usual, my first thoughts on this topic were about sports. I thought about a professional baseball game. Surely, in advance of a game, one has a decent idea as to what the result will be. The rampaging Dodgers against the hopeless Nationals? Clayton Kershaw vs. Ross Detwiler? Yeah, I think I can predict that outcome pretty well. But no one on earth could possibly predict how every pitch in the game would unfold. Why is this? Well, because there are far too many things to consider in such a calculation. Practically speaking, it is absurd to think that you can predict the exact outcome of every single play in the game.

But, in theory, imagine that right before the game began, we could freeze time, and collect every conceivable piece of data that would affect the game: the condition of every cell in the players’ bodies, the thoughts in every players’ head, the weather at the moment of play and the expected weather later in the game, the effect of the breath and voice of every spectator, the bend of every blade of grass, the composition of the wood in every single bat, etc. In theory, all of this data would allow us to accurately model the outcome of the game, wouldn’t it?

If you are skeptical, think of this small example. It’s the bottom of the 9th inning of a tied game, and the bases are loaded with two outs. The tiring relief pitcher becomes erratic, and throws three balls to the batter (who is one of the opposing team’s weaker hitters), making the count 3-0. One more errant pitch and the game would be over.

Now, it’s hard to predict exactly what would happen on the next pitch, but I can give you a pretty accurate guess. The pitcher will throw a fastball, and the batter will not swing. Why? Any baseball fan will know why – a fastball is the easiest pitch to locate accurately, and the batter will be willing to take a strike right down the middle on the off chance that the erratic hurler will once again miss the mark, thus ending the game. In fact, I’d be willing to bet that the batter wouldn’t even take the bat off of his shoulder as the pitch sailed by.

So really, there are only two possible outcomes in this scenario: the pitcher will throw a fastball strike and the batter won’t swing, or the pitcher will throw a fastball that misses the plate and the game will be over. Of course, there are a very large number of ways that this could unfold, but we have narrowed it down a tremendous amount. And in this case, we didn’t need much information to narrow down the possibilities either. In theory, any other situation in the game is the same as this one, in that sufficient information would allow us to accurately predict the outcome. Of course, the amount of information needed in other cases would be far greater, but in theory the information is all there.

So this leads me to believe that determinism is not really antithetical to probabilistic thinking or randomness, as I had generally supposed. Perhaps my view, having now thought about the issue a bit more carefully, can be captured as follows.

When a pitcher gets the sign from a catcher to throw a curveball in the dirt, he registers this mentally, and begins his physical preparations. If you like, every cell in his body receives direction from his nervous system on how to proceed (I know, this is inaccurate biologically, but bear with me here). Then he begins his windup. Now freeze it, halfway through his delivery. Many of the cells, joints, muscles and nerves that were given instructions by his brain are responding as he hoped they would. But many are not. Why? I believe it is down to randomness, caused by unrecognizable fatigue, weather conditions, the friction from the uniform affecting the pitcher, etc. Most important among these reasons, however, is biological randomness. The human body is not a machine.

Does this mean that we cannot predict what happens to the pitch? That it could go anywhere, even backwards? Well, no; and I want to highlight two reasons why not. First, the randomness and variation occurs at such infinitely small levels that it does not actually affect the outcome in a dramatic way (he’s probably not going to throw it into the dugout, basically). Second, on aggregate, these random deviations might just cancel each other out.

Perhaps a machine can help me make my point here. Open Excel and type in “=RAND()” in a cell. You will get a random number between 0 and 1. Now highlight the cell and drag this command down to multiple cells. On the bottom right of your screen, you will see the cell average, count, and sum. And you’ll see that the further down you drag, the closer you get to an average of 0.5, the expected value of a uniform distribution between 0 and 1. In a way, a muscle in a pitchers body is like this – a collection of RAND commands, converging on the intended goal as ordered by the human nervous system.

So my thoughts have led me to a question, more than an answer. If we froze the earth now, and took every piece of data possible – every thought, every cell, every molecule – could we accurately predict the rest of human history? Or would the randomness inside everything, the standard deviation of every sinew, leave us hopeless in our desire to know how all things will end?

Wednesday, July 15, 2009

A Paper of Mine

I wrote this paper for a class this last Spring, and I wanted to share it. And as a dorky Ph.D. student, I would love feedback of any and every kind on it!

---

The Perfect Behavioral Storm:
Hyperbolic Discounting, Herding, and Anchoring in Mortgages


“Listen, you, when you gave me that money you said I wouldn’t have to repay it ‘till the future - this isn’t the future! It’s the lousy, stinkin’ now!”
- Homer Simpson, to his mortgage broker


As the ongoing financial crisis rocks the global economy, everyone wants someone to blame. While some have heaped scorn on Bernie Madoff or the executives at AIG, others have pointed an accusatory finger at the ever-growing number of American homeowners who are unable to make their mortgage payments. Clearly, the bursting of the real estate bubble has played a central role in precipitating this crisis, and irresponsible borrowing has been a key factor in the mess. However, those who demonize “greedy” mortgagers as the primary drivers of the crisis miss the mark for a number of reasons.

In this paper, we will examine one of those reasons, namely the role that the interaction between various behavioral biases and a broken lending system has played in the downfall of the American financial system. Specifically, we will examine three behavioral phenomena - hyperbolic discounting, herding, and anchoring – and the impact they have had on the mortgage crisis. In doing so, we will see that human nature is as responsible for the mortgage crisis as human beings, if not more so.

Hyperbolic Discounting and the Curse of American Optimism


Americans are by nature an optimistic people. With this optimism comes an enduring belief that no matter how difficult the times may be, the future is bright. Though this is generally regarded as a positive mindset, it can result in irrational beliefs on the aggregate. For example, a 2001 study found that the majority of Americans believe that they will be earning above the mean income in the future. This is mathematically impossible, and hugely optimistic, given that the mean income in the US exceeds the median income by nearly $20,000.

This wishful thinking about personal finances may seem harmless, but its impact on household decision-making regarding mortgages is very dangerous. Why? Because mortgages are essentially a gamble on future income – a prospective homeowner considers his expected future income and decides on a house to purchase on the basis of this expectation. Thus, if individuals consistently overestimate their future incomes, they are more likely to occupy homes (and take on mortgages) that are outside of their price range. In theory, loan officers ought to ensure that this does not happen – but as the recent crisis has shown, perverse incentives at banks and lending institutions have regularly resulted in lax loan screening. As a result, bad mortgages have quietly piled up in banks, relatively unnoticed until now.

This problem is compounded by the behavioral phenomenon of hyperbolic discounting, whereby individuals discount the distant future at irrationally high rates compared to the near future. Hyperbolic discounting has become an especially relevant issue for mortgages in recent years because of “innovations” in the mortgage market, such as “balloon mortgages” and other back-loaded mortgage options. These mortgages, which offer lower monthly payments in the short-run in exchange for dramatically higher payments in the future, play right into the hands of hyperbolic discounting. By allowing individuals to defer payment for homes until the distant future, these back-loaded mortgages attract borrowers who do not fully realize, because of hyperbolic discounting, that they will be unable to make the large future payments. It is no surprise, then, that many lenders have found that when the time comes for the “balloon payments” to be made, a large fraction of borrowers default on their mortgages.

Herding and the House Price Fantasy

While over-optimism and hyperbolic discounting were influential factors in encouraging poor individual decisions about mortgages at a micro-level, macro-level forces also contributed to poor borrowing practices. Perhaps the most destructive of these forces was the view - held by essentially every household in the US - that housing appreciates in value over time. This view is evident in a 2005 study, which estimated that the average homeowner in San Francisco and Anaheim (where real estate was booming at the time) believed that their home was going to appreciate in value by 14% per year. Of course, we now know that this confidence was not warranted. So where did it come from?

Robert Shiller, a Yale economist and co-author of the 2005 study, has written extensively about this problem. He argues that the misplaced faith in the rising price of housing is the result of a behavioral phenomenon called herding, whereby large groups of individuals tend to act in unison because of the individual propensity to look to others for guidance. Imagine, for example, an individual who is skeptical of the housing market in 2005 – that is, his private belief is that housing is a risky investment. This individual would likely look to the market for more information about the true value of housing. What would he find? In 2005, he would have seen scores of optimistic buyers, willingly paying high prices. As a result, this individual – whose initial skepticism was warranted – would be convinced that his private assessment was wrong, and would therefore invest in a home. The cycle would then repeat, with more and more individuals pulled into the market by the decisions of the masses.

In this way, herding encouraged the housing bubble and the individual acceptance of mortgages that were unaffordable. As we know from basic economics, individuals are more likely to consume when they believe their future income will be higher. In the case of the housing bubble, people factored the perceived rise in housing prices into their mortgage decision, which encouraged them to buy homes that they now realize they cannot afford.

Anchoring and Option Adjustable-Rate Mortgages (ARMs)

We just saw that many borrowers were driven to sub-optimal mortgage decisions by various behavioral biases; however, it must also be said that the actions of banks and other lending institutions did little to alleviate the problem. For example, in addition to screening loans poorly, these institutions regularly offered complex mortgages that made things worse. A classic example is the Option ARM mortgage, which generally gave mortgagers four monthly payment options. Table 1 below offers an example of an Option ARM monthly bill:

Table 1: Option ARM Mortgage Bill Example (click on table to enlarge)


The reason for offering such a flexible payment schedule is that it allows households to decide, on a month-to-month basis, how much they are able to put into their mortgage. If they want to pay it off quickly, they can (Options 3 or 4 in Table 1). If not, there are smaller payments available (Options 1 or 2). The troubling aspect of this scheme, however, is that the minimum payment (Option 1) amount does not cover the interest payment on the principal. So by opting to make the minimum payment in a given month, a household increases the size of the principal. This results in increases in monthly payments in the future, which can be sudden and significant depending on the terms of the mortgage. Of course, this would not be a problem if households regularly made greater monthly payments (Options 3 or 4), and only opted for the minimum payment in an emergency. Sadly, this is not the case – according to a mortgage market tracker, over 65% of option ARM borrowers make only the minimum payments each month.

The implications of this are clear, and alarming. Large numbers of mortgagers are borrowing too much money for homes they cannot afford, failing to pay even the interest amount for months, and then defaulting entirely on the inflated principals when they realize that they owe more money than the value of the home. "Your payment is going to go up exponentially and your ability to pay will not,” says Keith Gumbinger, a mortgage expert at HSH Associates. “You just might chuck your keys in an envelope and mail it to your bank." According to recent estimates, of the nearly one million option ARM mortgages currently outstanding, 10-24% are currently 90 days or more past due. Moreover, experts expect that number to double in the next two years. What is behind all of this?

To a large extent, the structure of option ARM loans is the culprit. Behavioral research tells us that anchoring – the emphasis on one salient characteristic or value and the effect of this on decision making – is a powerful force for human beings. In the case of option ARM mortgages, lenders are essentially anchoring the minimum payment as the “mortgage cost” in the minds of borrowers. Given the behavioral tendencies to discount excessively and take an overly optimistic view of future income, this anchoring mechanism was an even more powerful force in the run-up to the mortgage crisis, as it encouraged countless borrowers to opt for the minimum payment in order to increase consumption in the short run.

Conclusions

In this paper, we looked at three ways in which behavioral biases have exacerbated the mortgage crisis in the United States. Specifically, we discussed how over-optimism, hyperbolic discounting, and the herding phenomenon created an environment in which individuals willingly took on mortgages that they soon found to be unaffordable. We also saw that by designing payment options that were lenient in the short-run, lenders dramatically increased the long-run default risk associated with mortgages. We will undoubtedly be paying the price for these oversights for years to come.

So who is to blame? In truth, everyone is a little bit culpable, and anyone who concentrates all the blame on a handful of individuals or institutions is misguided. Instead, what we have today is a fundamental disconnect between government policies, lender incentives, and homeowner psychology. Together, the government, the lenders, and the borrowers have been playing a dangerous game without really realizing it, encouraging home ownership and high levels of financial leverage without appropriate safeguards in place. In order to ensure that this does not happen again, policymakers and the public must learn from the behavioral mistakes of the past, and take seriously Robert Shiller’s conclusion that, "the ultimate cause of the global financial crisis is the psychology of the real estate bubble.”