The Map and the Territory Read online

  The t-statistic is a measure of the “statistical significance” of an independent variable—that is, the probability that its coefficient is different from zero.21 The higher the t-statistic, the higher the probability that the relationship between an independent variable and the dependent variable is real, not merely the product of chance. Economists usually require a t-statistic, whether positive or negative, to be higher than 2.0 before we accord an independent variable credibility as a “cause” of variations of the dependent variable. The Newey-West estimator measures the extent of bias in the t-statistic owing to serial correlation and resets the t-values so that they more truly reflect the actual probabilities.

  Another prominent bias in many economic correlations results from two time series that are only slightly related, if at all, but nonetheless exhibit a high R2 when regressed against each other because both series reflect population growth. This bias can be largely eliminated by expressing the dependent and independent variables on a per capita basis.

  Exhibit 7.3 is a typical example of regression analysis. The dependent variable is capital investment as a share of cash flow for nonfinancial corporations. We collect quarterly observations not only for the dependent variable, but also for the three independent variables22 from 1970 to date. The dependent variable is regressed against the three independent variables and we create a fitted estimate of capital investment’s share of cash flow. With an R2 of .76, we have in effect “explained” three fourths of the variation in the share. As can be seen from the associated chart, the fitted series closely follows the actual share. The Newey-West–adjusted t-values are well in excess of 2.0 and hence the probability that the relationships are wholly the result of chance can be dismissed. The Durbin-Watson measure of serial correlation is only 0.57, suggesting there is still significant serial correlation. But as can be seen from the chart, it has not prevented the independent variables from following the dependent variable through boom and bust. Moreover, if we split the forty-three-year regression period into two equal halves, the results of the regressions for the two smaller periods are similar to those of the full regression. This is a useful test of whether the independent variables’ impact on the dependent variable has changed over the forty-three years. The results here suggest not.

  A POINT OF CAUTION

  We must be careful to distinguish correlation (which regression analysis can measure) and causation (which it cannot). A high R2 and high t-statistics are not necessarily, in themselves, a credible measure of causation. Regression analysis has turned out to be one of the most effective tools for divining economic cause and effect. But we must always keep in mind that correlation or association is not the same as causation. It must be backed up with a credible economic explanation of the association.

  Regression equations as well as economic identities (see Box 9.3) are the most prominent inputs of our macroeconomic models. Regression analysis became widespread only with advances in computational capability. Back in the 1950s, it took me hours, if not days to estimate a regression using the desk calculators of the time. With today’s computers and software, I need press only a small number of keys to produce an almost instantaneous final result.

  THREE

  THE ROOTS OF CRISIS

  The toxic securitized U.S. subprime mortgages were the immediate trigger of the financial crisis, but the origins of the crisis reach back to the aftermath of the Cold War.1 The fall of the Berlin Wall in 1989 exposed the economic ruin produced by the Soviet bloc’s economic system. East Germany, touted as a strong competitor to West Germany, turned out to have achieved little more than one third the productivity levels of West Germany after four decades of rivalry. Most Western analysts, including those at some of the most vaunted intelligence agencies of the Western democracies, had estimated levels of productivity in East Germany that were 75 to 80 percent of the levels reached in West Germany.2 They were shocked by the reality of the situation.

  Competitive markets quietly but rapidly displaced much of the discredited central planning so prevalent in the Soviet bloc and in much of the Third World. India, a longtime bastion of Fabian socialism, initiated significant reform in 1991. Under Finance Minister Manmohan Singh (later India’s prime minister), it opened markets and reduced the heavy burden of extensive regulation (though admittedly much of the bureaucratic apparatus remains intact in India). China, with its highly collectivized economy and with the Cultural Revolution only fifteen years in its past, embarked on what turned out to be a major embrace of free markets, though the Chinese economy continues to be plagued by too-frequent instances of crony capitalism that make its system a far cry from textbook capitalism.

  China, and a large segment of the erstwhile Third World nations that followed suit, replicated the successful export-oriented economic model of the so-called Asian Tigers (Hong Kong, Singapore, South Korea, and Taiwan): Fairly well-educated, low-cost workforces, joined with developed-world technology, unleashed explosive economic growth.3 Between 2000 and 2007, the real GDP growth rate of the developing world was almost double that of the developed world. The International Monetary Fund (IMF) estimated that by 2005 more than eight hundred million members of the world’s labor force were engaged in export-oriented competitive market activities, an increase of five hundred million since the fall of the Berlin Wall.4 Additional hundreds of millions engaged in domestic-oriented competitive markets, especially in the former Soviet Union.

  Because consumption in the developing world was restrained, perhaps by culture, inadequate access to consumer finance, or in response to the aftermath of the Asian financial crisis, it did not keep up with the surge of income. As a consequence, the savings rate of the developing world soared from 23 percent of nominal GDP in 1999 to 33 percent by 2007, far outstripping its rate of investment. With investment elsewhere in the world slow to take up the slack, the result was a pronounced fall from 2000 to 2005 in global long-term interest rates, both nominal (Exhibit 3.1) and real.5 Of course, whether it was a glut of intended saving or a shortfall of investment intentions is of interest primarily to economists. The conclusion is the same either way: Real long-term interest rates had to fall, and they did.

  Long-term interest rates in all developed economies and the major developing economies had by 2006 converged to single digits.6 Equity premiums (the excess rate of return required by investors in risky stocks over the return available on sovereign debt) and real estate capitalization rates were inevitably pressured lower by the fall in global long-term real interest rates. Asset prices, particularly home prices, accordingly moved dramatically higher.

  The Economist’s surveys of nearly twenty individual countries document the striking global nature of the rapid ascent of home prices during the decade.7 Japan, Germany, and Switzerland (for differing reasons) were the only important exceptions. At their peak increases, U.S. home price gains were no more rapid than the global average.8

  In short, geopolitical events ultimately led to a fall in long-term interest rates and the mortgage interest rates to which they were tied. That led, with a lag, to a global boom in home prices. In the United States the change in the thirty-year mortgage interest rate tends to lead the monthly change in home prices (in the opposite direction) by three months (Exhibit 3.3).9

  The subprime mortgage market that developed in the 1990s was a small but generally successful market of largely fixed-rate mortgages. It served mainly those potential homeowners who could not meet the down payment requirement of a prime loan but still had income adequate to handle the monthly payments of a fixed-rate mortgage. But when home prices accelerated after 1996 (Exhibit 3.4), subprime lending became increasingly attractive to investors. That said, subprime mortgages still constituted only 7 percent of total originations by 2002, and only a modest amount of those had been securitized.

  SECURITIZATION OF SUBPRIMES: THE CRISIS STORY UNFOLDS

  Belatedly drawn to the growing market for subprimes, many major financial firms, starting in late 2003, began
to accelerate the pooling and packaging of subprime mortgages into securities (Exhibit 3.5). Leading the pack of securitizers were Countrywide Financial and Lehman Brothers—firms that were ultimately brought down by their involvement with these instruments. Combined, they issued more than a fifth of securitized subprimes in 2004.10 The firms clearly had found receptive buyers. Heavy demand from Europe,11 in the form of subprime mortgage-backed collateralized debt obligations, was fostered by attractive yields and a foreclosure rate on the underlying mortgages that, since late 2000, had been in decline for almost three years. The securitizers pressed subprime mortgage lenders such as Ameriquest, New Century, and Countrywide to originate mortgages and sell them to securitizers. This reversed the conventional sequence of mortgage banks, assisted by brokers, originating mortgages and then deciding to sell them to securitizers.

  But a far heavier demand was driven by the need of Fannie Mae and Freddie Mac, the major U.S. government-sponsored enterprises (GSEs), to meet the expanded “affordable housing goal” requirements of the Department of Housing and Urban Development.12

  Given the GSEs’ expanded commitments, they had few alternatives but to invest, wholesale, in subprime securities.13 As a consequence, by 2004, the GSEs accounted for nearly half of all newly purchased subprime mortgage securities retained on investors’ balance sheets (Exhibit 3.6).14 That was more than five times their estimated share in 2002. To meet this demand, securitizers began to purchase and securitize much of the billions of dollars of subprime whole mortgages not already pooled to back outstanding mortgage securities. But that source was nowhere near enough to meet the GSEs’ needs, and at this point, securitizers unwisely prodded subprime mortgage originators to increase their scale of originations. But the number of potential homeowners seeking subprime fixed-rate mortgages was small. The originators therefore chose to reach out to a wholly new segment of potential homeowners who could meet neither the down payments of a prime loan nor the monthly debt-servicing requirements of fixed-rate subprime mortgages. The only available option to entice new, but risky, buyers was to offer adjustable-rate mortgages (ARMs) to subprime borrowers with initially lower monthly payments. As a result, the share of ARMs in the total value of first mortgages subprime originations soared to nearly 62 percent by the second quarter of 2007.15 The mortgage arrears of subprime ARMs almost immediately began to rise, and many borrowers failed to make even the first mortgage payment. Delinquencies of more than ninety days began to mount.16

  By the first quarter of 2007, owing to pressure on the securitizers to meet the demand for subprime securities from the GSEs, virtually all subprime mortgage originations (predominantly ARMs) were being securitized, compared with less than half in 2000.17 Also by the end of March 2007, more than $800 billion, or more than 80 percent of total outstanding subprime mortgages, was in pools supporting outstanding subprime mortgage securities.18 That $800 billion figure was almost seven times its level at the end of 2001.

  Subsequent to 2003, the securitizers, profitably packaging this new source of adjustable-rate mortgage paper into mortgage pools and armed with what turned out to be grossly inflated credit ratings, were able to sell seemingly unlimited amounts of these securities into what appeared to be a vast and receptive global market. But this proved to be a mirage.

  By 2005 and 2006, subprime mortgage originations had swelled to 20 percent of all U.S. home mortgage originations, almost triple their share in 2002.19 We at the Federal Reserve were aware earlier in the decade of incidents of some highly irregular subprime mortgage underwriting practices. But, regrettably, we viewed it as a localized problem subject to standard prudential oversight, not the precursor of the securitized subprime mortgage bubble that was to arise several years later. On first being told in the early months of 2005 by Fed staff of the quarterly data for 2004, I expressed surprise given that our most recent official Fed data—those of the Home Mortgage Disclosure Act (HMDA) compilations for 2003—exhibited few signs of problems. I had never heard of the private source Inside Mortgage Finance before. But in retrospect those data turned out to be right.

  But even had we had the hard official data from 2005 at that time (it was eventually published in the 2006 HMDA report by the Federal Reserve in December 2006), there was little the Fed could have done to contain the rise in home prices.

  Some academics favored an incremental defusing of the bubble through a gradual tightening of monetary policy, but such incremental policies have never appeared to have worked in the real world. The Federal Reserve’s incremental tightening in 1994 in the face of the inchoate dot-com boom, in retrospect, may well have fostered the bubble rather than contained it.20 Policy makers confront nonmarket problems all the time, a large majority of which are readily resolved. We can spot bubbles as they inflate, but we are not able to forecast their complex resolution and collapse and, as I note later, perhaps never will. In responding to such issues, policy makers have to choose whether to clamp down on and prohibit a wide variety of market practices and accordingly accept the inevitable constraints on economic growth that that often implies.

  SIZE OF THE PROBLEM CLOAKED

  The true size of the American subprime problem was hidden for years by the defective bookkeeping of the GSEs. Fannie Mae was unable to get its books certified and had to stop reporting publicly between November 2004 and December 2006, pending an often delayed clarification of their accounts. Freddie Mac had had similar problems earlier. Not until the summer of 2007 did the full magnitude of the subprime problem begin to become apparent. Fannie’s third-quarter write-offs almost quadrupled relative to the second quarter. Moreover, the source of loss did not become fully evident until September 2009, when Fannie Mae very belatedly disclosed a significant reclassification of loans, from prime to subprime, dating back to loans that it had made and held in 2003 and 2004. More important, those revelations helped to explain how what was thought to be a relatively sound portfolio of prime conventional mortgages in mid-2007 could generate the huge losses Fannie and Freddie had been reporting.21 It is doubtful, however, even had such data been available in a timely manner, that regulators would have been able to head off the crisis, especially given our experiences with monetary tightening during the dot-com boom.

  The role of Fannie and Freddie in the American financial system has been highly controversial at least since 2003, when the Federal Reserve raised serious questions about the potential systemic risk that these institutions might pose down the road. As I put it in testimony before a Senate committee in early 2004, “we see nothing on the immediate horizon that is likely to create a systemic problem. But to fend off possible future systemic difficulties, which we assess as likely if GSE expansion continues unabated, preventive actions are required sooner rather than later.”22 Our problems with the GSEs had been simmering for years.

  I recollect getting a call in late October 2008 from Henry Waxman, a senior House Democrat, inquiring whether I agreed with the position that the housing bubble’s primary cause was the heavy demand by the GSEs for subprime mortgages and securities, as some House members had argued. (The housing market had peaked and fallen in 2006.) I responded that I thought the GSEs did contribute to the crisis, but that their holdings of subprime securities did not seem sufficiently large to account for the size of the bubble. With the new disclosures, it was obvious that Fannie and Freddie played a far more important role—perhaps even a key role—in the momentum that developed behind the housing bubble than many had theretofore recognized.

  Had Fannie and Freddie not existed, a housing bubble could still have taken hold. But had such a bubble developed, it is likely that in and of itself, it would not have wreaked such devastation in late 2008. The paths of Canada and Australia’s house price gains, for example, were in fact quite similar to those of the United States. Yet in retrospect, those gains did not have the destabilizing bubble characteristics experienced by Americans. Their financial systems were not breached and neither experienced a severe financial crisis.


  Analysis of such crises is subject to considerable uncertainties, and I have little doubt that longtime political supporters of Fannie and Freddie will continue to advance potential explanations other than that which I have just offered. Many still claim the buildup of mortgage assets during 2003 and 2004 was simply a business decision on the part of the GSEs to regain lost market share (they are, after all, private corporations) and had little to do with the issue of affordable housing. But as Fannie Mae in December 2006, in political-speak, acknowledged, “We have also relaxed some of our underwriting criteria to obtain goals-qualifying mortgage loans and increased our investments in higher-risk mortgage loan products that are more likely to serve the borrowers targeted by HUD’s goals and subgoals, which could increase our credit losses.”23 They sure did.

  A CLASSIC EUPHORIC BUBBLE TAKES HOLD

  The housing surge of the last decade had all the hallmarks of a classic euphoric bubble. Financial bubbles occur from time to time, and usually with little or no forewarning. The sources of bubbles and the markets in which they occur are quite varied, but patterns these bubbles trace have very common features. As the top of a speculative boom in stocks, homes, or commodities is approached, there obviously still must be a preponderance of buyers over sellers bidding up those already high prices. For were it otherwise, those prices could never have reached so high a tipping point. In fact, only if a vast majority of investors expect prices to move higher, and are fully committed, can prices break sharply. At that point, as buyers finally become sated, bids disappear, only sellers are left, and prices plunge.