The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
SEPTEMBER/OCTOBER 19 9 3 ECONOMIC PERSPECTIVES Contents Tracking Midwest manufacturing and productivity growth..............................................................................2 P h ilip R. Is ra ile vich , K enn eth N. K u ttn e r, and R obert H. Schnorbus During the 1980s, Midwest manufacturing experienced a “productivity takeoff.” The authors explore the scope and causes of that takeoff as well as its implications for the mixed-frequency Midwest Manufacturing Index. Why the life insurance industry did not face an "S&L-type" crisis............................................................... 12 Elijah B rew er III, T ho m as H. M o nd schean , and Ph ilip E. S trah an Declines in real estate values and junk bond prices in the late 1980s adversely affected both life insurance companies and savings and loan associations. Yet important differences between the two industries, especially in the way they are regulated, prevented a crisis from occurring in the life insurance industry. R <( )i\( )\11( i PER SPL( j1 IV ES i( Sep tem ber/O ctob er 1993 V o lum e X V II, Issue 5 Karl A. Scheld, Senior Vice President and Director o f Research ECONOMIC PERSPECTIVES is published by the Research Department of the Federal Reserve Bank of Chicago. The views expressed are the authors’ and do not necessarily reflect the views of the management of the Federal Reserve Bank. Single-copy subscriptions are available free of charge. Please send requests for single- and multiple-copy subscriptions, back issues, and address changes to the Public Information Center, Federal Reserve Bank of Chicago, P.O. Box 834, Chicago, Illinois 60690-0834, or telephone (312) 322-5111. Articles may be reprinted provided source is credited and the Public Information Center is sent a copy of the published material. Editorial direction Janice W eiss, editor David R. Allardice, regional studies Steven Strongin, economic policy and research Anne Weaver, administration Production Nancy Ahlstrom, typesetting coordinator Rita M olloy, Yvonne Peeples, typesetters Kathleen Solotroff, graphics coordinator Roger Thryselius, Thomas O ’Connell, Lynn Busby-Ward, John Dixon, graphics Kathryn Moran, assistant editor ISSN 0164-0682 Tracking Midwest manufacturing and productivity growth P h ilip R. Israilevich , Kenneth N. K u ttn er, and R obert H. Schnorbus higher in the region than in the rest of the nation. After years of lagging econom We then use the mixed-frequency MMI to assess ic performance that led to the region’s characterization as a the quantitative significance of the increased “rust belt,” Midwest manufac productivity growth for current estimates of turers have exhibited increas Midwest output. ing competitiveness in the last several years, The productivity takeoff identified in this compared with the rest of the nation andanalysis has important implications for the MMI with model, which relies on historical rates of produc their own earlier performance.* Evidence of this tivity growth to account for the divergence of strong performance is the fact that the Midwest’s output and output grew faster on average than the nation’s, employment data, and to compute current given observed rates of capital and labor usage. estimates of the index. In light of the evidence, we reconsider the assumption of a A common explanation for this resurgence constant rate of productivity growth for each has been that in the 1980s, Midwest manufactur industry, and suggest a modification to the MMI ers undertook aggressive modernization pro grams in an attempt to reverse their fortunes. model that will allow it to capture the technical This explanation, however, rests largely on progress that resulted from modernization of core Midwest industries. anecdotal evidence; data have been hard to come by. With the help of annual production models, Investm ent and th e p ro d u c tiv ity ta k e o ff the mixed-frequency Midwest Manufacturing in th e M id w e s t Index (MMI) developed by Israilevich and Kutt The key to regional growth is improving ner (1993), and annual capital expenditure data competitiveness, and the key to increasing a from the U.S. Commerce Department, we are region’s competitiveness is productivity growth beginning to reach a clearer understanding of the relative to other regions. Such productivity region’s improvements in productivity and com gains can be achieved in at least two ways: fast petitiveness as Midwest manufacturers move er withdrawal of the least productive capital into the 1990s. stock (downsizing) than elsewhere, and faster In this article, we explore the reasons for the introduction of new, more technologically ad so-called takeoff in Midwest manufacturing vanced plant and equipment than elsewhere. productivity, tracing its growth to significant While both measures can yield increased output modernization efforts in several key industries. per worker, they have different implications for Investment data and estimates of production models both suggest that productivity gains were ■ *The Midwest is defined here to include the five states in the Seventh Federal Reserve District: Illinois, Indiana, Iowa, Michigan, and Wisconsin. Philip R. Israilevich is Senior Economist, Kenneth N. Kuttner is Senior Research Economist and Research Officer, and Robert H. Schnorbus is Senior Business Economist and Research Officer, all at the Federal Reserve Bank of Chicago. 2 ECONOMIC PERSPECTIVES future growth. If the recent produc FIGURE 1 tivity gains in the Midwest were Capital expenditures per worker achieved only by shrinking the man thousands of current dollars per worker ufacturing base without moderniz ing, the region would be vulnerable to further declines as other regions improve their competitiveness and increase their market share at the Midwest’s expense. However, if Midwest manufacturers were mod ernizing while they were closing antiquated facilities, they might offset any net reductions in capital stock with productivity gains suffi cient to allow output growth relative to the rest of the nation. In the early 1980s, manufactur Source: U.S. Department of Commerce, Bureau of the Census, Survey of ers were under severe financial Manufacturers and Census of Manufacturers, various issues. stress, particularly in the Midwest. A relatively deep recession in 1980-82 was followed by an intensification of worker in the Midwest and in the rest of nation global competition caused in part by a strong does not appear to be due to differences in in dollar. Many well-known companies such as dustrial mix. Indeed, both the auto and steel Caterpillar, USG, and Chrysler were pushed industries show higher investment per worker in the region than in the rest of the nation. For dangerously close to bankruptcy; virtually all manufacturers in the Midwest scrambled to cut example, between 1986 and 1990, investment costs in order to be as competitive as possible in per worker in the transportation industry was 16 an increasingly tough global market. As part of percent higher on average in the Midwest than in that effort, many old or marginally profitable the rest of the nation; in primary metals it was 22 plants were closed under the banner of percent higher on average. While these two “rationalization”—a term which in the 1990s industries show larger differentials than other would be dubbed “re-engineering.” industries, they demonstrate that the pattern Despite these financial problems, many observed at the aggregate manufacturing level Midwest manufacturers met the increasing com reflects a widespread commitment to moderniza petitive pressures of the early 1980s with aggres tion among Midwest manufacturers. sive capital spending programs. While with A closer look at the auto and steel industries drawing older capital stock, they also invested in reveals the dual nature of the adjustments that new plants and equipment. The only question manufacturers made in response to competitive was whether these adjustments were occurring at problems. During the 1980s, automakers closed a faster pace in the region than they were else seventeen car and truck assembly plants, of where in the nation. which six were in the Midwest. At the same Before 1985, the Midwest tended to invest time, they constructed seventeen plants, seven of at roughly the same rate as the rest of the nation. them in the Midwest. Some of the new plants Investment in the region picked up in the late were essentially replacements of existing Big 1970s but slowed again with the onset of the Three plants, for example, Chrysler’s Jefferson 1980-82 recession. Thereafter, Midwest invest Avenue plant in Detroit. But some were entirely ment lagged the rest of the nation until a push new plants built by foreign auto companies, resumed in 1985.1 As figure 1 shows, between often in conjunction with a Big Three producer. 1986 and 1990, average capital expenditure per Among the foreign-owned plants are the Dia worker in the Midwest was 9 percent above the mond Star Plant in Illinois (Chrysler and Mit amount for the rest of the nation. subishi) and the Flat Rock Plant in Michigan The Midwest contains a high proportion of (Ford and Mazda). capital-intensive industries, notably auto and A somewhat similar pattern of investment steel, yet the difference between investment per occurred in the Midwest steel industry, where FEDERAL RESERVE BANK OF CHICAGO 3 integrated mills were closed and the remaining mills modernized. Inland Steel, for example, has invested roughly $1 billion since 1985 to mod ernize its Indiana Harbor Works in East Chica go, Indiana (which included converting to con tinuous casting). The company spent another $1 billion on a new mill in Indiana, a joint venture with a Japanese producer. While integrated steel producers were modernizing, they were also opening mini-mills that brought a wholly differ ent production process to U.S. steelmaking. The result was that both the auto and steel industries saw more productivity gains in the Midwest than in the rest of the nation. These gains made Midwest producers more competi tive and allowed industry in the region to grow faster than elsewhere. capital inputs), 0 and < are the elasticities of \> output with respect to labor and capital, and r\ is a random error term. The Cobb-Douglas specification is also consistent with the mixedfrequency MMI introduced subsequently, as well as a variety of other indices discussed in Israilevich et al. (1989). We first estimated the production function over the sample running from 1973 through 1985, dates chosen on the basis of the Midwest investment patterns discussed above. Using this estimated function, we then projected output for the 1986-90 period on the basis of pre-1986 “old” technology, and compared the projection to the actual VA data, the result of production with “new” technology. P ro d u c tiv ity g ro w th in th e M id w e st: e v id e n ce fro m ann ual data Table 1 reports the difference between projected and observed output growth for 15 key manufacturing industries in the Midwest, aggre gated into five sectors: transportation, metal working, machinery, chemicals, and consumer products. Table 2 shows the composition of these sectors and a breakdown of Midwest out put by industry. For comparison purposes, simi lar calculations were done for the rest of the nation. According to these estimates, between 1986 and 1990, Midwest manufacturing sectors improved efficiency by 8 percent more than the corresponding sectors in the rest of the nation. Given the capital-intensive nature of most Mid west industries and their relative maturity, such a gain is substantial. It would also help explain why output has been growing faster in the region than in the nation since the late 1980s. Figure 2 displays the efficiency gains graph ically, showing the Midwest’s lead as a function of time. While the gap between observed and The investment patterns noted above sug gest that Midwest manufacturers began to mod ernize aggressively around 1986. What is lack ing is some measure of how much efficiency increased as a result. How much has Midwest manufacturing output grown, compared with the growth that would have occurred using pre-1986 technology? One way to address this question is simply to compare the out-of-sample forecasts from a production function estimated on data from 1973 through 1985 with observed output from 1986 to the present. A natural and intuitive measure of the size of the takeoff is the difference between the Midwest’s observed output and the model’s prediction: that is, the amount by which actual output exceeds what would have been produced had pre-1986 technology been applied to the actual factor inputs. A convenient production function for this analysis is the Cobb-Douglas specification, The Midwest versus the nation TABLE 1 Efficiency gains, 1986-90 (percent) *,= Y-f + 0/,+ <K+Tl,, where * represents output of a given industry measured by the logarithm of real value added (VA), t indexes the year, l is the logarithm of payroll employment, and e is the logarithm of electricity consumption. For applications such as this, energy consumption is widely interpreted as a proxy for the utilized stock of capital.2 The y coefficient on the time trend represents the rate of Hicks-neutral technological change (i.e., productivity not embodied in either labor or 4 M id w e s t Rest of n a t io n Transportation 7.94 3.83 Metalworking 2.03 2.38 Machinery 2 .0 1 0.89 S e c to r Chemicals Consumer products T o ta l ECONOMIC PERSPECTIVES 0.76 3.10 -3.33 -0.58 1 .3 8 1 .2 8 TABLE 2 Composition of Midwest manufacturing output, 1990 Sector Share (%) Industry Share (%) Transportation 14 Transportation (SIC 37) 14 M etalw orking 14 Primary metals (SIC 33) 5 Fabricated metals (SIC 34) 8 M achinery 15 Nonelectrical (SIC 35) 26 Electrical (SIC 36 and 38) Chemicals 36 10 Chemicals (SIC 28) 8 Petroleum (SIC 29) 22 5 Clay, glass and stone (SIC 32) Consumer products 1 Rubber and plastic (SIC 30) 1 Food (20) Lumber and w ood (SIC 24) 11 1 Furniture and fixtures (SIC 25) 2 Paper products (SIC 26) Printing and publishing (SIC 27) 3 4 Miscellaneous (SIC 39) 1 Note: Industry subtotals may not equal sector totals because of rounding. predicted output remained positive, it flattened out in 1989 and declined in 1990. Only in 1990 did the rate of improvement in efficiency seem to subside, both in the Midwest and elsewhere. This pattern suggests that the shift in the national economy from a mini-boom in 1988-89 (rough ly 4 percent real GDP growth) to virtual stagna tion (roughly 1-2 percent real GDP growth) had an impact on efficiency gains. Perhaps the cy clical drop in output growth prior to the 1990-91 recession led to underutilization of labor and capital, which reduced the measure of efficiency gains over the second half of the 1980s. More over, the commitment to efficiency gains even in a sluggish economy may help explain why man ufacturers have been able to expand output since the 1990-91 recession even though employment growth has been virtually nonexistent. in the region’s core manufacturing sectors, trans portation and machinery. The Midwest’s transportation sector scored the most impressive gains in efficiency. Output in this sector was 7.9 percent higher than fore cast on the basis of pre-1986 technology, com pared to 3.8 percent higher in the rest of the Comparisons between Midwest industries The efficiency gains identified in table 1 were clearly not uniform across the Midwest’s industries. How widespread were they, and how much of the total gain was due to the industrial structure of the region relative to the rest of the nation? The gains did seem to be concentrated FEDERAL RESERVE BANK OF CHICAGO 5 nation. In the Midwest, the transportation sector is dominated by automobile manufacturers and parts suppliers, both of which were troubled industries throughout the 1980s. Japanese im ports and nameplates produced in the U.S. had been gaining market share for many years, leav ing the domestic industry with tremendous over capacity. The first wave of restructuring took place in the early 1980s when Ford and Chrysler began closing plants. GM began closing assem bly plants in the late 1980s and is currently in a second wave of closings that will extend into 1995. At the same time that Big Three automak ers were closing plants, both they and the Japa nese were opening state-of-the-art assembly plants in the Midwest as well as elsewhere. Over the 1980s, the region’s share of total car production actually rose from 39 to 44 percent, although its share of truck production declined from 40 to 28 percent. The Midwest’s machinery sector also out paced the rest of the nation. While in the rest of the nation machinery was on average 0.9 percent above its predicted level of output over the 1986-90 period, in the Midwest it was 2 percent higher than predicted on the basis of the pre1986 technology. The region’s machinery sector is largely focused on the auto industry and ex ports. As suppliers of the new capital, this sector has been in the forefront of the recent wave of investment targeted to meet global competition. Machinery producers themselves have faced stiff competition from foreign competitors, particu larly the Japanese. Moreover, some machinery producers have been bought out by foreign com panies, a change that often brings an infusion of fresh capital that improves productivity. It is encouraging to see that machinery producers, especially in the Midwest, have accepted the challenge of heightened global competition by increasing capital expenditures rather than by closing or shifting to other markets. In the aggregate, the Midwest’s metalwork ing sector displayed efficiency gains roughly in line with the rest of the nation. However, disag gregating the sector into its two constituent industries, fabricated metals and primary metals (the steel industry in the Midwest) reveals an interesting contrast. While the pace of technical change lagged the nation in fabricated metals, productivity growth in primary metals exceeded the nation’s—a divergence that also appears in the MMI results presented later in this article. Interestingly, the major downsizing in the steel 6 industry was over by the mid-1980s, leaving the Midwest as the dominant integrated steel-pro ducing region. Midwest firms continued invest ing in modernization, and even mini-mills were expanding in the region. It is the Midwest’s continued modernization, and perhaps its domi nance in the high-quality steel produced by integrated mills, that allowed the region to out pace the rest of the nation in productivity. In contrast to primary metals, the metal fabrication industry, which produces finished parts from raw steel, never experienced any significant consolidation. The small size of producers in this fragmented industry may have limited the adoption of technical advances. While efficiency gains were clearly wide spread in the Midwest, not all the region’s indus tries outpaced their counterparts elsewhere in the nation. The Midwest’s chemical and consumer products sectors actually lagged the rest of the nation in efficiency gains over the 1986-90 period. In fact, efficiency in the latter sector was lower during the period than in previous years. While these industries are important to the Mid west, it is interesting that they are generally outside the auto-steel-machinery complex that comprises the heart of the region’s manufactur ing. It is perhaps unfortunate that strength in this “heart” seems not to spill over into other industries, yet by the same token, it seems that weakness in some sectors does not retard effi ciency gains in other sectors. T h e p ro d u c tiv ity ta k e o ff and th e MMI The preceding section discussed measuring Midwest productivity gains by comparing annu al V A data with predictions from estimated production models. An alternative method is to apply a similar analysis to predictions generated by the mixed-frequency MMI, as described in the appendix. The main advantage of the mixedfrequency MMI is that it tracks actual VA more precisely than other purely annual indices, such as the annual Cobb-Douglas or Atlanta methods, when projected out of sample.3 Hence, the MMI should yield a more accurate assessment of Midwest efficiency gains than the annual model. A second reason to use the MMI in this context is to examine any implications the hy pothesized productivity takeoff might have for current estimates of Midwest output. Although the production model underlying the mixedfrequency MMI is re-estimated as new annual VA data become available, an increase in the ECONOMIC PERSPECTIVES rate of technical progress may require structural modifications to the model to enable it to track manufacturing output more accurately in the future. Out-of-sample comparisons To construct a quantitative measure of Mid west efficiency growth, we first estimated the mixed-frequency MMI using annual data from 1973 through 1985. We then used monthly energy, labor, and nationwide Industrial Produc tion (IP) data to project the MMI forward over the 1986-90 period, in which annual real VA data for the Midwest are available. Comparing the projected series with the actual VA data yields an index of efficiency gains that is compa rable to the measures reported earlier. As be fore, an increase in the rate of productivity growth would imply that the projected MMI would underpredict output growth. This short fall, therefore, represents the region’s gains expressed in terms of the additional output pro duced as a result of increased manufacturing productivity. Table 3 reports these gains, classified by industries and sectors. The results are expressed as the average percentage deviation between observed real VA growth and the annualized growth rate of the projected MMI. In metal working, for example, the reported 0.6 percent figure signifies that on average, the MMI under predicted VA growth by 0.6 percent for each year in the 1986-90 period. The results are broadly similar to those based on the annual estimates reported above. Most striking is the spectacular productivity growth in the transportation sector, which con sists entirely of SIC 37. Here, annual productiv ity growth over 1986-90 was roughly 9 percent higher than in the preceding 13 years. To restate this in cumulative terms, by the end of 1990, output in the transportation sector was about 40 percent higher than it would have been had firms applied pre-1986 technology to the same labor and energy factor inputs. Such are the quantitative effects of the investment flows and modernization efforts identified earlier. Although there are a few bright spots, none of the other sectors showed the kind of spectacu lar growth detected in transportation. Echoing the earlier annual results, within metalworking, primary metals (SIC 33) did well, turning in a robust average 2.8 percent per year increase in T A B LE 3 Efficiency gains based on the MMI, 1986-90 S e c to r G ain (%) In d u s try Transportation 9.0 Transportation (SIC 37) M etalw orking 0 .6 Primary metals (SIC 33) G a in (%) 9.0 2 .8 Fabricated metals (SIC 34) M achinery - 0 .6 - 0 .8 Nonelectrical (SIC 35) -0.9 Electrical (SIC 36 and 38) Chemicals -0.9 0 .2 Chemicals (SIC 28) -0.3 Petroleum (SIC 29) -5.7 Rubber and plastic (SIC 30) 0.4 Clay, glass and stone (SIC 32) Consumer products FEDERAL RESERVE - 1 .6 -4.0 Food (SIC 20) - 0 .1 Lumber and w ood (SIC 24) 5.5 Furniture and fixtures (SIC 25) Paper products (SIC 26) - 6 .0 Printing and publishing (SIC 27) - 0 .6 Miscellaneous (SIC 39) BANK OF CHICAGO -3.7 -9.6 7 its rate of productivity growth. The TABLE 4 slight deterioration in fabricated Estimated shift in Midwest rate of productivity growth metals (SIC 34) partly offset this (annualized percentage) gain, however, resulting in a modest 1973-85 1986-90 D iffe ren c e overall gain for metalworking of only 0.6 percent. Transportation (SIC 37) 0.1 1 0 .0 9.8* Neither machinery nor chemi 4.1 3.7 Primary metals (SIC 33) 0.4 cals displayed any significant evi dence of a productivity acceleration. * significant at the .05 level. The small improvement in machin ery sector productivity evident in the to the 1973-85 period. If this more rapid annual results is not apparent in the MMI. The growth were extrapolated into 1993, then with rates of technical change in both nonelectrical the same inputs, output (measured by VA) (SIC 35) and electrical machinery (SICs 36 and would be roughly 70 percent higher than it 38) remained close to pre-1986 levels. The rate would have been using 1973-85 technology. of technical change also appeared stable in the The results for primary metals also provide chemical sector, with chemicals (SIC 28) and some evidence for a higher productivity growth rubber and plastic (SIC 30) indices tracking VA rate, although the statistical significance is weak quite closely. The exceptions were petroleum er. While the estimated shift coefficient implies (SIC 29) and clay, glass and stone (SIC 32), an increase in annual productivity growth of whose performance appeared to deteriorate sig 4 percent, it is not statistically significant at the nificantly. However, given the poor quality of traditional .05 level. the data and the very small size of these indus tries in the Midwest (each only about 1 percent of Extending the MMI 1990 VA), little weight should be given to these These findings have potentially important results. implications for current appraisals of Midwest Performance within the consumer products output. One of the purposes of the MMI is to sector was rather disappointing overall. All indus assess the level of manufacturing activity prior tries showed some diminution in their rate of to the release of VA data, which become avail technical change, with the exception of lumber able after a two- to three-year lag. Contempora and wood (SIC 24). Since that industry currently neous estimates of the growth of industry output accounts for only 1 percent of the Midwest’s incorporate a weighted average of energy and output, its impact on the region is small. labor inputs, plus the rate of productivity growth relevant for that industry. Updates of the MMI, Modeling the productivity takeoff therefore, depend critically on whether this rate How important are these results statistically? of productivity growth is stable. Projections that How might the mixed-frequency MMI model be did not take into account any productivity accel extended to allow a changing rate of productivity eration might as a result seriously understate growth? What is the impact of more rapid techni current output levels. cal change on current estimates of the MMI? To To assess the consequences on the MMI, we address these three issues, we re-estimate the perform one final exercise, comparing post-1990 MMI for the transportation and primary metals MMI projections with and without a shift in industries—the two industries that show signifi productivity growth in 1986. Rather than cant acceleration in the region—allowing a shift re-estimate the model for every industry, we in the productivity growth rate in 1986. The again concentrate on the two showing some significance of this shift can then be evaluated evidence of a productivity takeoff: primary statistically. metals (SIC 33) and transportation (SIC 37). The results of this exercise, as reported in The results appear in figure 3. table 4, generally support the out-of-sample The top panel shows the impact of this findings. Again, the evidence for a productivity change on the aggregate MMI. The effect is takeoff is strongest for transportation, which small but perceptible. The cumulative discrep experienced a statistically significant increase in ancy relative to the unadjusted index was annual productivity growth of 10 percent relative 8 ECONOMIC PERSPECTIVES These results demonstrate that if the productivity acceleration had continued from 1990 to the present, it may have had a noticeable impact on the MMI; accordingly, the exist ing MMI would have understated the Midwest’s actual output from 1991 to 1993. Should the index then be modified to incorporate higher rates of productivity growth in cer tain industries? Clearly, the answer depends on recent productivity developments. For example, if we assumed that the 1986-90 rate of change had continued into 1993 but it had actually levelled off, then modifying the MMI would intro duce an upward bias into it. For this reason, the appropriate incorpo ration of changes to the MMI model requires an ongoing, disaggregated examination of the structure of the economy. C o n c lu s io n 2 percent as of April 1993. Naturally, the effects on the individual industries, depicted in the middle and bottom panels, are larger. As ex pected in light of the earlier results, the most pronounced effect is in transportation, where the adjusted MMI is 15 percent higher than the unadjusted by April 1993. The cumulative im pact on primary metals is a smaller but still substantial 3 percent. FEDERAL RESERVE RANK OF CHICAGO Despite falling levels of em ployment, Midwest manufacturing output expanded rapidly during the 1980s. This growth, which sur passed national output growth over the period, suggests improved com petitiveness among the region’s manufacturers. The evidence con firms this impression. Comparing the predictions of production mod els applied to annual Midwest data with similar predictions for the rest of the nation showed that the re gion’s brisk expansion was due in large part to strong productivity growth. The main cause of this growth appears to have been the aggressive modernization efforts of Midwest manufacturers, as reflected in the region’s higher rate of investment per worker relative to the national average. Using the MMI to evaluate the size and scope of the productivity gains, we found that they were largely confined to a few key industries, particu larly transportation and primary metals. Howev er, given the prominence of these industries in the Midwest, their impact on overall manufacturing output is substantial, possibly raising current 9 estimates in excess of 2 percent if the productivi ty growth observed from 1986 through 1990 continued into 1993. This finding underlines the importance of incorporating higher rates of technical change for certain industries into future updates of the MMI to reflect the continuing modernization of Midwest manufacturing. APPENDIX T ra c k in g M id w e s t m a n u fa c tu rin g w ith th e m ixe d -freq u e n cy M M I A useful tool for analyzing Midwest manufactur ing is the mixed-frequency Midwest Manufacturing Index (MMI) developed by Israilevich and Kuttner (1993). While this technique uses the CobbDouglas production function employed in the annual results, it differs from this specification in its use of a monthly production model. At the same time, it constrains the estimated monthly production series in such a way as to be consistent with the observed annual value added (VA) data; hence the “mixedfrequency” designation. Incorporating monthly data yields two signifi cant advantages over annual models. First, it makes it possible to track high-frequency fluctuations in Midwest output. Second, the mixed-frequency MMI has been shown to provide more accurate out-ofsample projections of manufacturing activity than pure annual models. Since annual VA data are not yet available for the Midwest, this benefit is particu larly useful for assessing the effects of accelerated technical change on the current output of the region’s manufacturing sector. The foundation of the mixed-frequency MMI is a Cobb-Douglas production equation applied to monthly data. Expressed as first differences of natu ral logarithms, the monthly change in the real output of any Midwest industry, Ax7s, is the weighted sum t of the change in employment hours, Al7 and energy ts, usage, Ae^: Axls = y+QAll + ^Aels+ i\ts. As in the annual model, y is the (constant) rate of Hicks-Neutral technical change, 0 and < represent the |> elasticity of output with respect to labor and capital (energy), and T is a stochastic error term. The super ) script 7 is used to denote Midwest data. Note that with the shift to monthly data, each variable now receives two subscripts. The first, t, denotes the year, while the second, s, represents the month within that year. Thus the change in output between the second and third months of the 13th year of the sample would be denoted Ax7 3. I3 A difficulty with this approach is that while monthly energy and labor data are available for the Midwest, no monthly output measure exists. The only available measure of region’s production is the 10 real value added (VA) data used in the annual results. In light of this data limitation, estimating the monthly model might appear to be a lost cause, since tradi tional regression techniques require the observations on the left-hand variable to be available at the same frequency as those for the right-hand variables. Using regression methods, therefore, requires that energy and labor be aggregated to an annual frequen cy. This is the approach used earlier to compare productivity growth in the Midwest and in the rest of the nation. Fortunately, there are ways around this obstacle. Techniques exist to combine data of differing fre quencies into a single model. For the mixed-frequen cy MMI, we use a state-space econometric model that treats Midwest output growth as a latent vari able. Given some additional relationships between the unobserved Ax7 and other data series, the month s ly model can be estimated even in the absence of direct information on Midwest output. One key link between Ax7 and something s observable is the “adding up” relationship between the monthly growth of output and the annual growth of the real VA data. Because the annual VA obser vations correspond to the sum of the output produced in each month, the year-to-year change in real VA is actually a weighted average of the monthly output growth in the current and preceding 23 months. Thus, constraining the monthly growth rates to pro duce an annual pattern consistent with the VA data implies that In(VAj) - In(VAj j) = -±- • I E Ax1 ., Imposing this equation enforces consistency between the estimated MMI and the annual VA data.This relationship alone is not enough for the monthly approach to yield any dividends, since all the available information is still coming at an annual frequency. In order to make inferences about fluctu ations within the year, we need an additional source of monthly information. One source of such informa tion is the monthly index of industrial production (IP) prepared by the Federal Reserve Board. Besides the energy and labor inputs used as inputs to the MMI, the IP typically incorporates some information on actual output, such as the dollar value or physical quantity of goods shipped. Thus the IP index con- ECONOMIC PERSPECTIVES tains information on industry output not captured by energy and labor inputs alone. However, the infor mation in the IP index pertains to the nation, not to the Midwest. Therefore we cannot simply use IP to compute Ax7s. Instead, we relate national to regional t fluctuations by using an equation to describe the co movement of the two series: Mixed-frequency MMI model estimates for primary metals (SIC 33) As before, Ax7 represents the growth in Mid ts west output; Ax"s is the growth of national output in the same industry as measured by industrial produc tion. The coefficient 5 relates the magnitude of the national fluctuations to those of the region, and v is random “noise” in the relationship. Unlike the production model introduced earlier, this equation does not describe any fundamental economic or structural relationship between the region and the nation. Neither is the national IP in any way a determinant of regional output in the same way that regional labor and energy inputs are. Rath er, this equation describes how Midwest economic fluctuations have historically been paralleled by movements on a national scale. Clearly, the fact that Midwest industry compris es a portion of the national total, implies a positive correlation between the region and the nation, repre sented by a positive value of 8. But to the extent that industries within and outside the region are subject to similar demand conditions, one might expect the correlation to be even greater than suggested by the industry’s share in total output. It is unlikely, how ever, that 8 would exceed 1, since many regional fluctuations will be damped by offsetting fluctuations IP indicator equation < = 0.33* f> p = 0.00 0 = 1. 11 * Ax"s = •u + SAxl, s + v,, s . t, t t Production m odel 8 7 = 0.001 Standard deviation of v = 0.025 Standard deviation of r| = 0.038 = 0.54* Note: Based on the 1973-90 sam ple. * significant at the .01 level. in the rest of the nation. While the 8 parameter picks up the relative magnitudes of industrial fluctuations, the standard deviation of v captures the amount of “noise,” or unpredictable variation, in the link be tween regional and national output. Table 5 shows the results from estimating the mixed-frequency MMI model for one representative industry: primary metals (SIC 33). The estimates of the production function’s parameters all fall within the range of economically reasonable values, al though the sum of < and 0 imply increasing returns to \> scale. The estimate of y (which is constant through out the sample) suggests only very modest productiv ity growth of 1.4 percent per year. The very small estimate of p. indicates that output has grown at roughly the same rate in the nation as in the Midwest. The estimated 8 of 0.54, however, suggests that IP fluctuations in the nation are approximately half the magnitude of fluctuations in the Midwest. FOOTNOTES 'Estimates o f Midwest capital expenditures for the years 1979-81 are not available in the Commerce Department’s Annual Survey o f Manufacturing (ASM). Values were calculated by first comparing a sample of 480 Midwest firms with 100 or more employees, taken from the Longitu dinal Research Data (LRD) base for the years 1985-88, with the reported ASM data for those years and, second, applying the average proportions to the LRD base to gener ate ASM-equivalent data. 2Moody (1974) discusses the use of energy as a proxy for capital services. 3A description o f the Atlanta method appears in Israilevich and Kuttner (1993). REFERENCES Israilevich, Philip R., Robert H. Schnorbus, and Peter R. Schneider, “Reconsidering the regional manufacturing indexes,” Federal Reserve Bank of Chicago Economic Perspectives, Vol. 13, No. 4, July/August 1989, pp. 13-21. Israilevich, Philip R., and Kenneth N. Kuttner, “A mixed-frequency model of regional output,” FEDERAL RESERVE RANK OF CHICAGO Journal of Regional Science, Vol. 33, No. 3, forth coming 1993, pp. 321— 43. Moody, Carlisle E., “The measurement of capital services by electrical energy,” Oxford Bulletin of Economics and Statistics, Vol. 36, No. 1, February 1974, pp. 45-52. 11 Why the life insurance industry did not face an "S&L-type" crisis Elijah B rew er III, Thom as H. M ondschean, and P h ilip E. S trahan Since August 1989, the Resolu tion Trust Corporation has l spent $84.4 billion of taxpayers’ money to close 653 savJSllfiB ings and loan associations (S&Ls).1 In addition, between 1986 and 1990, over 900 commercial banks were closed with assets totaling over $100 billion. On July 16, 1991, in response to policyholder runs during the previous three months totaling approximately $500 million, New Jersey regulators seized the Mutual Benefit Life Insurance Company. The asset quality problems that led to this and other runs on life insurance companies in the early 1990s have led some to wonder whether yet another category of financial intermediaries might suffer widespread failures requiring gov ernment intervention at taxpayer expense. Gov ernment closings of financial institutions can be extremely costly to taxpayers, and the safety of life insurance policies and annuity contracts is of concern to millions of policyholders. For these reasons alone, it is important to assess the risk exposure and regulatory structure of the U.S. life insurance industry.2 But there are other reasons as well. First, according to the Federal Reserve Flow o f Funds, the industry held approximately $1.2 trillion in assets at the end of 1991, accounting for 11.4 percent of total financial assets. Capital adequa cy or asset quality problems in this industry could lead to disintermediation, or the transfer of saving and borrowing activities from life insur ance companies to other financial institutions. This in turn would result in less efficient alloca tion of capital. Second, most state governments bear part of the cost of an insurance failure by 12 providing tax credits to life insurance companies (LICs) that pay guaranty fund assessments. Third, losses from failures are partially borne by insurance and pension policyholders, reducing potential income to retirees. Finally, the experi ences of the life insurance industry can provide some lessons for bank regulators. The 1980s witnessed two important changes in the mix of LIC business: continued growth in pension and annuity business relative to life insurance, and a shift toward interest-rate-sensitive products. Competitive pressures led some LICs to shift their asset portfolios from low- to high-risk investments in order to cover the high er rates on these new liabilities. By the end of the decade, this strategy had begun to unravel. The sudden but short-lived collapse of the junk bond market and the fall in the value of commer cial real estate reduced LIC profitability. In reaction, LICs pulled back from the commercial real estate market and certain segments of the corporate bond market. At first glance, there are many similarities between the savings and loan and the life insur ance industries. Both S&Ls and LICs act as financial intermediaries and face substantial government regulation. Life insurance policy holders, like S&L depositors, are protected by government-administered guaranty funds. Elijah Brewer III is a senior economist with the Federal Reserve Bank of Chicago. Thomas H. Mondschean is assistant professor of economics at DePaul University and consultant to the Economic Research Department of the Federal Reserve Bank of Chicago. Philip E. Strahan is an economist with the Federal Reserve Bank of New York. ECONOMIC PERSPECTIVES Because of the partial guarantee of their liabili ties, firms in both industries have incentives to take risk. Many have argued that regulators exacerbated the S&L crisis by allowing thrifts to invest heavily in high-risk loans and securities and by not closing insolvent firms promptly, while private creditors did not impose market discipline on S&Ls because their deposits were guaranteed. Yet despite the similarities between S&Ls and LICs, the life insurance industry has not suffered widespread failures. In this article we explore possible explana tions for the divergence in behavior and perfor mance between these two classes of financial institutions. First, we argue that in contrast to commercial banks and LICs, S&Ls were danger ously exposed to interest rate risk. As a result, when nominal interest rates rose sharply in the late 1970s, S&Ls experienced a larger decline in the market value of their portfolios than did LICs or banks. Then we suggest five key differ ences that reduced the moral hazard problem for LICs relative to S&Ls: 1) LICs possessed a larger capital cushion than S&Ls; 2) S&L creditors had more confidence in their government guarantees than did LIC creditors; 3) a smaller proportion of LIC liabilities were subject to a government guarantee; 4) LICs were subject to greater market disci pline from uninsured creditors; and 5) LICs were subject to greater monitoring by other LICs. The article is organized into six sections. First, we present financial information about the life insurance industry both to document the importance of LICs as financial intermediaries and to describe the environment in which they operate. Second, we describe the recent finan cial problems of the industry. Third, we sketch the regulatory framework that protects policy holders and manages insolvencies. Fourth, we discuss how interest rate risk differs across fi nancial institutions. Fifth, we examine key differences that reduced the moral hazard prob lem for LICs compared to S&Ls. Finally, we discuss the implications of these findings for regulatory policy. B a ckg ro u n d Traditionally, life insurance companies offer customers risk protection by agreeing to FEDERAL RESERVE BANK OF CHICAGO indemnify them against losses specified in a policy. Insurance guards against economic loss by compensating those policyholders suffering losses from a pool of funds paid by all policy holders who are exposed to similar risks. At the end of 1991, the most recent year for which data are readily available, over 375 million policies were in force in the United States, with coverage totaling approximately $10 trillion. LICs’ total 1991 revenues from premium and investment income were $411 billion. LICs raise funds primarily from the sale of life insurance policies, annuities, and pension plans that have a savings feature as part of their contract. LICs must set up reserve accounts for the excess of the value of benefits payable in future years over the value of the premiums to be collected for each contract. The reserve ac counts are divided into two types of liabilities: (1) life insurance reserves, which cover LIC obligations to policyholders and beneficiaries; and (2) pension reserves, which cover expected payments to retirees and other annuitants. These liabilities of LICs are savings instruments by which households can accumulate wealth for retirement and bequests. In turn, LICs use the premiums paid for these products to invest in debt and equity securities. In doing so, they help transform a large portion of the financial assets of households into real capital investment by businesses and governments. Premium income from life insurance prod ucts represented 44 percent of total gross income of LICs in 1970 but fell to 19 percent by yearend 1991 (see table 1). Much of this decrease occurred because traditional life insurance con tracts with savings components offered policy holders a substantially lower return after taxes than did alternative investments. During the 1970s and early 1980s, rising inflation rates and high yields on alternative investments created greater competition for household savings. Re turns on traditional life insurance contracts were tied to the average rate of return on the insurer’s portfolio. However, because LICs held a large share of fixed-rate bonds purchased previously at lower interest rates, the average rate of return on their portfolio did not increase as rapidly as market rates of interest. As a result, a large gap emerged between prevailing interest rates and the return on traditional LIC contracts. In addi tion, many policyholders exercised their right to borrow against their policies or cashed them in for their surrender value in order to invest the 13 customers, including liberal surren der provisions that allow withdraw Gross income of life insurance companies als without penalty when promised (billions of dollars) yields fall below benchmark rates Source (Cabanilla 1992). Because GICs are 1991 1980 1985 1990 of income 1970 relatively short-term liabilities, these contracts tend to reduce the average Life insurance 76.7 21.7 40.8 60.1 79.3 premiums duration of insurance companies’ (44.3)a (31.2) (25.7) (19.1) (19.3) liabilities. Table 2 reports that the 3.7 22.4 53.9 129.1 123.6 Annuities 6 share of life insurance industry (32.1) (17.1) (23.0) (30.1) (7.5) general account assets financed by Health insurance GICs rose from 8.1 percent in 1986 29.4 11.4 41.8 58.2 60.9 premiums to 10.8 percent in 1990. By year(14.8) (23.3) (22.5) (17.9) (14.5) end 1991, however, this share had 67.9 1 1 1 .8 119.0 Investments 1 0 .1 33.9 fallen to about 8 percent, primarily (28.9) (2 0 .6 ) (25.9) (29.0) (27.8) because some highly publicized 1 0 .2 28.2 Other 2 .1 4.3 26.3 failures caused GIC holders to shift (4.4) (6.5) (6.9) (4.3) (3.3) funds to alternative investments. 402.2 411.0 49.0 130.9 234.0 Total Because the interest income (100.0) (100.0) (100.0) (100.0) (100.0) credited on universal life policies aN um bers in parentheses are the percent o f total incom e. and other liabilities affected the bln 1986, there w as a large increase in annuity prem ium receipts demand for these instruments, insur because of an N A IC -m andated change in statutory reporting. Note: N um bers m ay not add to totals because of rounding. ance companies have an incentive to Source: A m erican Council of Life Insurance. offer high rates during the early years of these policies to attract new customers and to forestall policy lapses and surrenders by existing customers. funds where they could earn higher rates. This Wright (1991) claims that in order to maintain created outflows of LIC funds. the high returns being paid on GICs and other To stem outflows and attract additional liabilities, many insurance companies sought to funds, LICs developed new products such as increase interest income either by taking on universal and variable life insurance policies. riskier real estate loans or by reducing the quali These differed from traditional whole life poli ty of their corporate bond portfolios. cies in that the size of the death benefit and/or Historically, life insurance companies have the annual premium could change to reflect played an important role in the bond and mort investment performance over the duration of the gage markets. In 1960, they held about 50 per policy. Such interest-rate-sensitive products cent of all outstanding corporate bonds. While offered new options, including the ability to move the investment portion of the policy among alternative assets to reflect policyholders’ TABLE 2 current preference between risk and return. As Guaranteed investment contracts table 1 shows, premium income from annuity (billions of dollars) business accounted for 30 percent of gross in TABLE 1 come at the end of 1991, compared with only 7 percent at year-end 1970. In addition to standard annuity products, some life insurance companies have sold guaran teed investment contracts (GICs). Widely used as funding instruments for defined contribution pension plans, GICs typically obligate an insur ance company to repay principal and interest accruing at a predetermined rate in a single payment at maturity. Thus GICs have no insur ance element. Competition for this business has resulted in very favorable contract terms for 14 Percent Total o f assets 1986 67.1 8 .1 1987 74.8 8 .0 1988 105.1 1 0 .1 1989 1 2 1 .6 10.5 1990 134.6 1 0 .8 1991 130.0 8.4 Source: Am erican Council o f Life Insurance. ECONOMIC PERSPECTIVES this share has fallen with the growth of mutual funds and pension plans, LICs still hold about one-third of all corporate bonds. Within the bond market, they are major buyers of private place ment debt, which are securities issued in the U.S. but not registered with the Securities and Ex change Commission. LICs are also very active in the commercial mortgage market, which provides a market for loans on nonresidential properties such as office buildings and manufacturing plants. Together, LICs, commercial banks, and S&Ls supply about 80 percent of the credit for all commercial real estate loans. During the 1980s, LICs held about 30 percent of all commercial mortgage loans (Cabanilla 1992). Lending in the private placement and com mercial real estate markets requires substantial amounts of information gathering in the form of evaluating credit and monitoring of borrowers’ management through covenant enforcement. Recent studies of the private placement and com mercial real estate markets have indicated that the loans made by LICs in these markets generally have less uniform terms than do other invest ments such as publicly traded corporate bonds. As a result, private placements and mortgage loans are less liquid. Yields are higher to reflect information gathering costs and greater default risk. According to data from the American Council of Life Insurance, private placements and mortgage loans represented about 86 percent of new life insurance investments in 1980. At the end of 1991, they accounted for only about 29 percent. Conversely, the share of new funds that LICs invested in publicly traded corporate bonds and mortgage-backed securities has been increas ing during the 1980s and early 1990s. In 1980, these assets accounted for about 13 percent of all new investments of LICs. By year-end 1991, that figure had risen to 70 percent. The shift towards marketable and more liquid securities stemmed from the increased securitization of debt as well as from changes in liability structure and from the asset quality problems of life insurers. Life insurers' em erging fin a n cia l p ro blem s Table 3 examines the financial characteris tics of LICs classified by their 1986 book value statutory capital-asset ratios. More than threequarters of the industry’s assets were held by LICs with capital and surplus less than 9 percent of general account assets (low-capital LICs)3. Low-capital LICs held greater proportions of FEDERAL RESERVE BANK OF CHICAGO mortgage loans and junk bonds than did compa nies with capital ratios above 9 percent (highcapital LICs). Guaranteed investment contracts are a relatively more important funding source for low-capital LICs than for high-capital companies. Figure 1 presents the market capitalization-asset ratios for a sample of 44 publicly traded life insur ance companies classified as “high” junk bond holders (9), “high” commercial mortgage loan holders (11), and “others” (24).4 All three groups of LICs experienced a deterioration in market capitalization over the 1986-1990 sample period. However, the deterioration was the greatest for the high junk bond holders. Other things held constant, lower market capitalization-asset ratios at high junk bond LICs indicate a greater expo sure to the risk of failure. During the late 1980s and early 1990s, the increased emphasis on nontraditional insurance products along with shifts towards ex ante riskier assets took its toll. Declines in the market values of below-investment-grade bonds and commercial real estate reduced the market value of capital of many LICs; a few have been rendered insolvent. Two announcements in 1990 highlighted the industry’s emerging financial difficulties. In January, First Executive Corporation, a large holder of below-investment-grade bonds, an nounced that it would take a charge of $515 mil lion in the fourth quarter for junk bond losses. Then in October, Travelers Corporation, one of the largest holders of commercial real estate loans, announced it was setting aside $650 million in reserves for anticipated losses on its commer cial real estate portfolio. These and similar prob lems at other LICs led to policyholder liquidity runs and the collapse of several large companies such as First Executive Corporation in mid-1991. Liquidity runs could occur because many of the new products sold by LICs provide policyholders with liberal withdrawal provisions in which the holder may demand immediate payment of principal and accrued interest. According to Fenn and Cole (1992), holders of GICs and other inter est-rate-sensitive products are more likely than traditional policyholders to exercise withdrawal options on annuity products and to borrow against insurance products when the issuing firm appears troubled. Surviving LICs have responded to these financial problems by reducing their holdings of risky assets and improving capital ratios. The weakened condition of LICs reduced the supply of credit in both the commercial mortgage market and the below-investment-grade segment 15 TABLE 3 Financial characteristics of life insurance companies (billions of dollars) High-capital com panies3 1986 1987 1988 1989 1990 (----------------------- billions of dollars------------------------) 22.3 24.4 26.7 30.1 32.2 Junk bonds 4.7 6.7 5.6 GICs 2.3 3.4 5.3 6.8 10.2 13.8 179.7 201.2 229.7 259.0 290.5 M ortgage loans Total general account assets / 7.7 \ Book value o f net w o rth / m ortgage loans 163.8 157.4 153.9 148.8 144.4 Book value o f net w o rth / junk bonds 783.8 572.6 739.0 659.0 606.3 Book value o f net w o rth / total assets 20.3 19.1 17.9 17.3 16.0 Low -capital com panies3 1986 1987 1988 1989 1990 (----------------------- billions of dollars----------------------- ) 173.1 193.5 211.2 229.5 Junk bonds 28.9 40.3 38.9 44.7 43.3 GICs 67.6 82.4 95.7 110.0 117.7 683.6 757.1 842.1 918.2 979.1 M ortgage loans Total general account assets 242.6 t1 / ( Book value of net w o rth / m ortgage loans 16.7 16.5 17.2 18.1 19.3 Book value of net w o rth / junk bonds 100.8 79.0 93.2 93.3 108.5 Book value o f net w o rth / total assets 4.2 4.2 4.3 4.5 4.8 aL o w -c a p ita l life in s u ra n c e c o m p a n ie s are th o s e w ith b o o k ca p ita l-a sse t ra tio s less th a n o r e q u a l to 9 p e rce n t at th e end o f 1986. Th e re m a in in g c o m p a n ie s are c la s s ifie d as h ig h -c a p ita l. S ource: N a tio n a l A s s o c ia tio n o f In su ra n ce C o m m is s io n e rs (N AIC ), D atabase o f A n n u a l S ta te m e n ts. of the private placement market. Carey, et al. (1992) show that in the below-investment-grade segment of the private placement market, loan volume was down and loan rates were up. The rise in rates was not caused by a general increase in loan risk, but rather by LICs’ flight to quality. Corcoran (1992) also concludes that the reduced willingness of insurance companies to make new loans exacerbated the credit problems of the recent recession. The deterioration of commer cial real estate values and an increase in mort gage delinquency rates, as illustrated in figure 2, led LICs to reduce their exposure to both com mercial real estate as well as the private place ment market. 16 As a result of these problems, the industry capital-asset ratio fell in 1990 to 8.5 percent. In 1991, the life insurance industry increased its capital-general account asset ratio to 9.3 percent, signalling an improved ability of firms to absorb losses without becoming insolvent. This cush ion should help reassure policyholders about the solvency of LICs. Regulation o f life in surance co m p a n ie s Just as a capital cushion protects policy holders and other creditors from losses at LICs, government regulation also safeguards their interests. Life insurance companies are regulat ed for many of the same reasons as are other ECONOMIC PERSPECTIVES F IG U R E 1 Market capitalization of some LICs p e rcent Despite the uniform standards proposed by the NAIC, life insurance companies are still subject to widely varying degrees of regulatory scrutiny. Examinations vary with the size and sophistication of state insurance departments or with the level of resources that states allocate to regulation. Further, LICs vary in their ability to lobby for less restrictive regulations or scrutiny, and states vary in their susceptibility to such pressures. To protect policyholders and to manage insolvencies, all fifty states and the District of Columbia have established guaranty funds. Prior to 1970, only one state had a guaranty system to cover the obligations of life and health insurance companies. Then in 1970, the NAIC adopted a “model” guaranty system for subse quent consideration by individual state legisla tures. In addition to provisions stating what the guaranty fund covered, the NAIC model also allowed insurance companies to credit guaranty fund assessment costs on their state premium taxes. Within a year, nine states adopted legisla tion based on or similar to the NAIC model. Guaranty systems satisfy benefit claims of poli cyholders and annuitants in the event that an insolvent company lacks sufficient assets after liquidation. Harrington (1991) claims that the growth of these guaranty funds has contributed to the increased number and magnitude of insol vencies in the insurance industry in recent years. Guaranty funds are financed by ex post assessments on surviving insurance firms operat ing in the particular state, with each company financial intermediaries: first, to offset the mor al hazard problems exacerbated by government guarantees of LICs’ liabilities; second, to de crease the probability that failure of one LIC may cause policyholders at other LICs to exer cise their surrender options after losing confi dence in their companies’ ability to meet obliga tions;5 and third, to protect taxpayers from losses resulting from LIC failures. State insurance departments are the agen cies charged with regulating LICs. State regula tors enforce rates, asset restrictions, and other policies established by state legislation. If a company wishes to write insurance in a particu lar state, it must first receive permis sion from the state insurance com F IG U R E 2 missioner. Thereafter, LICs must Delinquency rates of commercial real estate mortgages provide regulators with income statement and balance sheet infor p ercent mation annually. In addition, state insurance departments usually audit companies operating within their borders once every three years. Most states levy a tax on insurance premiums to finance part of the cost of regulation. The National Associ ation of Insurance Commissioners (NAIC) also monitors LICs by per forming annual computerized audits. Companies failing four or more of eleven NAIC audit ratio tests face increased monitoring from state regulators (see Cummins 1988 for more details). FEDERAL RESERVE BANK OF CHICAGO 17 paying an assessment based on its share of total premium income. As of December 31,1992, in 39 states, LICs may offset assessments against their state taxes, thereby shifting the cost of failure directly onto state taxpayers. In the re maining states, LICs may impose a premium surcharge to cover the cost of the assessment. In most states, coverage under guaranty funds is $300,000 in death benefits, $100,000 in cash or withdrawal value for life insurance, $100,000 in present value of annuity benefits, and $100,000 in health benefits. Some states cover all insurance policies written by an insol vent firm located in the state; others cover the policies of residents only. In the case of unallo cated annuities such as GICs purchased by com panies to fund pension plans, some states cover up to a certain amount, usually $5 million. Oth er states, such as California, Massachusetts, and Missouri, do not cover GICs. Because of variations in state guaranty funds and in the way insolvencies are handled, the parties bearing the costs of an insurance failure differ across states. Surviving insurance companies initially pay their assessments and claim them as an expense on their federal corpo rate income tax return, reducing their federal income taxes. As companies receive tax credits in subsequent years, these credits become tax able income. As a result, the federal government bears part of the cost of an insolvency since it does not fully recover the present value of the tax decrease granted in the assessment year. In states with premium tax offsets, however, the majority of the cost is paid by state taxpayers. A study of 1990 life/health guaranty fund assess ments found that 73.6 percent was paid by state taxpayers, 8.9 percent by federal taxpayers, and 17.5 percent by the equity holders of the surviv ing firms.6 The way in which state guaranty systems manage insolvencies raises several policy con cerns. First, LICs pay nothing ex ante to receive the guarantees. Assessments are based on the ex post cost of a given failure and bear no relation ship to current or future LIC risk exposure. Sec ond, companies in states with premium tax off sets have little incentive to monitor each other, since over 80 percent of the assessment will be recouped through lower taxes. Third, insurance guaranty funds reduce the incentive for policy holders to exercise market discipline. In the absence of guaranty funds, policyholders would have more incentive to buy from safe LICs or to 18 demand lower premiums from high-risk firms. As the S&L crisis demonstrated, government guarantees of firm liabilities could create a mor al hazard problem. If these guarantees are mis priced, institutions with low net worth may have strong incentives to gamble for resurrection by investing in riskier assets.7 Interest rate risk at fin a n cia l in stitu tio n s The value of LIC portfolios has traditionally been relatively insensitive to changes in interest rates.8 A large proportion of LICs’ liabilities consists of life insurance reserves, and most of the payments for these products occur in the distant future. Most LIC assets consist of long term corporate debt, mortgages, and long-term government securities. In the absence of credit risk, both the nominal death benefits and the payoff of these long-term assets are determined at the outset. As a result, the firm is less ex posed to unanticipated changes in interest rates. If the firm decides to hold short-term assets such as Treasury bills or commercial paper against life insurance policies, it would have no guaran tee that its portfolio could support future claims. Declines in interest rates would reduce the firm’s earnings and its ability to meet future obligations. Regulation of savings institutions, on the other hand, has encouraged these firms to hold long-term, fixed-rate mortgage loans financed with short-term deposits. This strategy worked well during the period of stable interest rates from the end of World War II to the 1960s. But S&Ls remained vulnerable to changes in the level of interest rates. Because of Regulation Q interest rate ceilings, S&Ls were prevented from offering depositors competitive rates when mar ket interest rates rose above the ceiling rate. When this occurred, many depositors withdrew their funds in order to invest them in higheryielding money market instruments, which caused outflows of S&L deposits. To stem the outflow, S&Ls were allowed to offer several deposit products not subject to Regulation Q ceilings. However, because over 80 percent of S&L assets were invested in long-term, fixedrate mortgage loans made previously at lower rates, their interest income did not increase as rapidly as their cost of funds. As a result, S&Ls suffered negative interest rate margins. This predicament—interest rate risk— is particularly characteristic of the S&L industry. Figure 3 ECONOMIC PERSPECTIVES compares the capital-asset ratios for the S&L and life insurance industries. Between 1978 and 1982, the S&L capital ratio fell from 5.6 to 0.6 percent but the LIC capital ratio actually rose from 8.3 to 9.1. Since there is a better corre spondence between the durations of assets and liabilities of LICs, these institutions were less exposed to interest rate risk; hence, they did not experience the large losses and subsequent de clines in capital as a result of high nominal inter est rates from 1978 to 1982. To judge a firm’s exposure to interest rate risk, we use stock market data. The stock re turns of financial institutions depend on many economic variables besides interest rates, such as expectations of future economic conditions, future investment opportunities, productivity, and tax policies. Using a two-factor market model from the finance literature, we relate the return on a portfolio of each type of institution to the return on an index of the overall stock mar ket and the return on a portfolio of long-term government securities. The following equation allows us to compare the relative exposure of the three types of financial institutions to interest rate risk: (') h , - aj + + E j,t where Rj t = return on financial institution j at t, R ,., = return on stock market, Rj t = return on portfolio of long-term government bonds. The variable R .., controls for all economic variables that would affect profits for all corpo rations. The value of the second variable, Rf depends solely on interest rates, so its coefficient provides an estimate of the interest rate sensitivi ty of each type of financial institution. We estimated equation 1 using monthly returns for two sample periods, 1972-1982 and 1983-1991. We split the sample at the end of 1982 for several reasons. During the first period, S&Ls and banks faced governmentmandated interest rate ceilings. After the pas sage of the Depository Institutions Deregulation and Monetary Control Act of 1980, these regula tions began to be phased out. Moreover, the Gam-St Germain Depository Institutions Act of 1982 substantially liberalized S&L asset-holding FEDERAL RESERVE BANK OF CHICAGO FIGURE 3 Capital-asset ratios p e rcent Note: The m easures of capital used are statutory capital for LICs and tangible capital for S&Ls. Sources: Am erican Council of Life Insurance and O ffice of Thrift Supervision. powers. Both of these laws allowed S&Ls to reduce interest rate risk. Also, the market value of S&L capital dropped sharply during the 19811982 period. Brickley and James (1986) show that stock returns for poorly capitalized firms may respond less to economic variables since the deposit insurer bears the brunt of all losses. The results of estimating equation 1 appear in table 4. They show that S&Ls were much more exposed to interest rate fluctuations than either banks or LICs. In the first sample period, for instance, interest rate changes did not signifi cantly influence the stock returns of LICs. By contrast, S&L stock returns were highly sensi tive to those changes. For example, the estimat ed coefficient shows that S&L stock returns exhibited 90 percent as much sensitivity to inter est rate changes as did a portfolio of twenty-year government bonds. In fact, one cannot reject the null hypothesis that during the 1972-1982 peri od, S&L stock prices were as sensitive to interest rates as were long-term government bond prices. Flannery and James (1984) show that the degree of sensitivity of bank stock returns to interest rates depends directly on the duration mismatch between its assets and liabilities. Since life insurance companies actively try to match the maturity of both sides of their balance sheet, it is not surprising that LIC stock returns exhibit little interest rate sensitivity. In the second sample period, the interest rate sensitivity of S&L stocks decreased from 19 TABLE 4 Estimates of interest rate sensitivity for portfolios of commercial bank, savings and loan, and life insurance stocks3 Industry Intercept Return on market portfolio Return on government bond portfolio R2 Durbin-Watson statistic 1972-1982 Savings and loans -0.003 (0.004)b 1.030* (0.066) 0.904* (0.128) 75.4% 2.185 Commercial banks 0.001 (0.002) 0.510* (0.029) 0.150* (0.056) 75.4% 1.866 Life insurance 0.001 (0.002) 0.707* (0.030) 0.074 (0.057) 84.0% 1.819 Savings and loans -0.010 (0.004) 0.996* (0.077) 0.484* (0.125) 65.6% 1.622 Commercial banks 0.003 (0.003) 0.662* (0.046) 0.154 (0.075) 67.8% 1.378 Life insurance 0.002 (0.002) 0.722* (0.038) 0.164* (0.062) 79.1% 1.618 1983-1991 aThe m o n th ly p o rtfo lio o f returns fo r each indu stry includes all pu blicly traded stocks on the New York and Am erican Stock Exchanges and the NASDAQ. The data are fro m the Center fo r Research in Securities Prices (CRSP). The market index is the m o n th ly return on an equally w eighted p o rtfo lio o f all stocks on the three exchanges, inclusive o f dividends. The interest rate index is the m o nthly return on a p o rtfo lio o f long-term go vernm ent bonds w ith m a tu rity o f ap proxim ately 20 years. These tw o indices are also fro m CRSP. bStandard errors appear in parentheses. •s ig n ific a n t at the .01 level. 0.90 to 0.48, while neither the bank nor the LIC interest rate sensitivity changed significantly from the first sample period. Evidently, the deregulation the S&L industry may have had the intended effect of reducing but not eliminating interest rate risk. However, with S&L industry capital at historic lows during this period, the lack of responsiveness of stock returns to interest rate volatility may reflect the put protection afforded by deposit insurance. As a firm’s capi tal approaches zero, further declines will be reflected in increased deposit insurer liability rather than in stock returns. Since the capital of LICs and banks did not fall to the same degree in the 1980s, those institutions apparently did not experience a similar decline in interest rate sen sitivity. In fact, for LICs the point estimate actually increases from 0.07 to 0.16, although this difference is not statistically significant. These results indicate that S&Ls were uniquely vulnerable to interest rate movements in the 1970s. We attribute the weakness of this industry to regulations that encouraged savings institutions to hold an unbalanced book. In 20 contrast, both LICs and commercial banks have been permitted to hold a sufficiently broad array of assets to facilitate better diversification. M oral hazard at fin a n cia l in stitu tio n s Insurers have long dealt with moral hazard. By its very nature, insurance reduces the costs associated with a particular bad outcome and thus weakens the purchaser’s incentive to take costly self protective actions. For instance, holders of fire insurance have less incentive to buy fire extinguishers to protect their property than do uninsured individuals. In private markets, one way in which insurers mitigate this problem is by adding deductibles and copayments to policies. In the case of financial institutions, government liability guarantees weaken the incentive for creditors to discipline the propensity of firms to bear additional risk; fully insured depositors with confidence in the Federal Deposit Insurance Corporation (FDIC) will not waste time monitor ing their banks’ investment decisions. Effective monitoring by regulators and/or other firms can mitigate this moral hazard problem. ECONOMIC PERSPECTIVES Many analysts have argued that the S&L crisis occurred because government regulators did not control the moral hazard inherent in fixed-premium deposit insurance.9 Regulatory oversight declined during the 1980s. Insolvent S&Ls that were permitted to remain in operation were not monitored very closely. In addition, S&Ls were given new rights to invest in high-risk assets such as junk bonds and acquisition and development loans. In pursuit of high profits, many S&Ls responded by collecting federally insured deposits and investing them in high-risk, high-expected-retum assets. This action deep ened the insolvency problems. As a result, be tween 1987 and 1992 over 800 S&Ls were re solved by the Federal Savings and Loan Insur ance Corporation (FSLIC) and later the Resolu tion Trust Corporation. Brewer and Mondschean (1993b) show empirically that life insurance companies face similar moral hazard problems. They found that over the 1986-90 period, low-capital LICs experi enced one-time increases in market value capital following a shift from low-risk assets to high-risk assets such as real estate direct investment and equity issues. As expected, increases in risky assets did not have a statistically significant effect on the market value of high-capital LICs. Brewer and Mondschean (1993c) also show that the largest LICs that failed in 1991 had siz able exposures to junk bonds. In fact, their expo sure was so large that a decline of 12 to 14 per cent in the value of their junk bond portfolio was sufficient to wipe out their book capital complete ly. These findings are consistent with a moral hazard problem associated with government liability insurance. In response to declining asset values, both LICs and S&Ls were forced to set aside funds to reserve against losses on securities and loans. However, regulators anticipate spending over $200 billion of taxpayers’ money to resolve the S&L debacle, while the cost of managing insol vent LICs should be much less. We suggest that five key differences between the environment in which LICs operated relative to S&Ls reduced the moral hazard problem sufficiently to prevent a crisis in the life insurance industry. Vulnerability to capital shocks S&Ls faced a massive capital shock when interest rates skyrocketed in the early 1980s. In addition, regulators lowered the minimum capital requirements all S&Ls had to meet. FEDERAL RESERVE BANK OF CHICAGO Neither banks nor LICs faced a comparable decline in net worth. As capital declines or capital forbearance grows, a firm has an increasing incentive to pursue an aggressive strategy. This is because the firm’s capital acts as a deductible payment in a traditional insurance arrangement. In this context, the chance of losing the value of the owners’ stake in the firm reduces the incentive to hold risky assets.10 A firm with little or no capital, however, has little or nothing to lose by pursuing a gambling strategy. This explains why many insolvent S&Ls invested heavily in junk bonds during the 1980s. If the investments paid off, the institution’s owners reaped the rewards; if the returns were low, the losses were passed on to the deposit insurer. Figure 3 compares S&L and LIC book value capital ratios from 1975 to 1991. LIC capital ratios fluctuated between 8.0 and 9.3 percent over the period but exhibited little trend. By contrast, S&L capital ratios, computed using tangible accounting principles, fell sharply after the 1979-1982 recession. Since S&Ls are more exposed to interest rate changes than banks or LICs, they suffered massive losses when interest rates rose in the late 1970s and early 1980s. This capital shock exacerbated the moral hazard problem. Federal versus state guarantees S&Ls’ guarantees are administered by the federal government and carry the implicit back ing of the U.S. Treasury. This fact is widely known and inspires near-universal confidence. By contrast, LICs’ guarantees are administered by their respective states and carry no compara ble backing. These guarantees are not as well publicized as federal deposit insurance and seem to inspire less confidence in policyholders. As a result, insurance companies are more sensitive to the impact of poor financial health and asset risk on their ability to raise funds. Three cases from the life insurance industry support this interpretation. Mutual Benefit of New Jersey, like other LICs in that state, had no government guarantee on its liabilties. In early 1991, the company’s asset quality problems led its GIC holders to surrender their contracts. The asset writedowns at First Executive Corporation in early 1990 were followed by policyholder liquidity runs at its life insurance subsidiaries in New York and California. Apparently lacking faith in the guaranty fund system, policyholders 21 increased their surrender requests from the New York subsidiary after the regulatory seizure of First Executive Corporation’s California unit in April 1991. Another New York example is the case of Mutual Life Insurance Company of New York (MONY). Despite the existence of a guar anty fund, policy and contract holders withdrew more than $900 million during the third quarter of 1990, reflecting concern about MONY’s large real estate exposure. Similar liquidity runs oc curred at S&Ls in Ohio and Maryland that were covered by state deposit insurance funds. No such panic has occurred in federally insured S&Ls. Depositor confidence in the FSLIC, or at least in the implicit backing of the U.S. Treasury, has remained sufficiently high to prevent runs.1 1 Breadth o f coverage Because of the breadth of de facto coverage, S&Ls are able to use fully insured deposits as their primary source of funds. Congress in creased deposit insurance coverage in 1981 to $ 100,000 per depositor per institution. More over, all uninsured depositors have received full reimbursement in resolutions not culminating in liquidation. Some of the asset growth by S&Ls in the 1980s was financed by brokered deposits. These funds allowed S&Ls to draw deposits from the national market without giving up the benefit of federal deposit insurance coverage. By contrast, while some LICs used GICs and single premium deferred annuities (SPDAs) during the 1980s to facilitate growth, these instru ments have not received the same level of gov ernment backing as did brokered S&L deposits.1 2 In several cases, failure resolutions have imposed losses on LIC creditors in the form of delays in repayment and loss of interest. Unlike traditional life insurance products, GICs and SPDAs could be put back to the company at face value. This fact helps explain why the run on Mutual Benefit of New Jersey was started by GIC holders. Monitoring Financial institutions may face losses as a result of the failure of a competing institution. In the deposit insurance system, all banks and S&Ls pay upfront for deposit insurance. LIC state guaranty funds make these losses explicit in that surviving LICs pay the costs of a resolution. LICs can reduce these costs by pressuring regula tors to tighten enforcement of safety and sound ness regulations. In some states, LICs can also pass resolution costs on to taxpayers through 22 premium tax credits. Brewer, Mondschean, and Strahan (1992) found that in states where premi um tax credits do not exist, LICs hold safer port folios. This is strong evidence that when guaran ty systems provide incentives for self-monitoring, they reduce risk-taking and increase industry stability. Calomiris (1989) reached a similar conclusion in his study of antebellum deposit insurance systems. He found that self-regulating mutual liability systems achieved stability and survived financial panics. Free rider problems The size of a government insurance fund may also influence the behavior of its members. Larger systems will face greater free rider prob lems, which lead to less monitoring and weaker enforcement of regulations. As noted earlier, in state guaranty systems, surviving firms pay the costs in the event of failure. In the federal deposit insurance system, taxpayers provide financial backing, yet member institutions also bear some of the costs associated with widespread failures. In fact, the FDIC tripled its fees in the aftermath of the FSLIC’s bankruptcy and the deterioration of the reserves in the Bank Insurance Fund. Thus in both systems, firms have an incentive to reduce the costs associated with these government guar antees. But individual firms have more at stake in smaller, state-administered life insurance guar anty funds. As a result, LICs have a greater in centive to pressure regulators to enforce con straints on high-risk behavior.1 3 C o n c lu sio n s and p o licy p re scrip tio n s The recent failures of several large insurance companies have raised concerns about the sound ness of the life insurance industry. The industry’s overall portfolio risk appears to have increased during the 1980s. Moreover, LICs with lower capital ratios have higher concentrations of junk bonds and commercial real estate than do wellcapitalized LICs. In response to the liquidity runs in the early 1990s, the life insurance industry has restored profitability and raised new capital. The experiences of the life insurance industry stand in stark contrast to the disastrous problems that S&Ls experienced and suggest some conclu sions about how to contain risk-taking of deposi tory institutions. Like S&Ls and banks, life insurance compa nies may succumb to moral hazard because gov ernment guarantees weaken the incentive for creditors to constrain firm risk-taking. Our re search indicates that the use of premium tax ECONOMIC PERSPECTIVES offsets for guaranty fund assessments encourages LICs to increase portfolio risk. In addition, con cerns about liquidity runs have caused LICs to reduce their holdings of risky assets and improve capital ratios. These findings suggest a number of policy prescriptions that could help improve the safety and soundness of the life insurance industry. First, since government backing makes life insurance policies more attractive, LICs should pay for access to the guarantees. Premium tax offsets for the costs of resolving failures tend to lead to less industry monitoring because surviv ing LICs can pass a larger portion of the costs of resolving failures onto taxpayers. These offsets should be eliminated. Finally, regulators could increase market discipline by encouraging LICs to finance a portion of their assets with puttable, uninsured liabilities such as guaranteed invest ment contracts. Despite these weaknesses in the regulatory structure of LICs, it also contains strengths that should be extended where possible to depository institutions. For instance, risk-taking may be contained by encouraging financial institutions to monitor each other and thus reduce the need for costly regulation. What is crucial is aligning the incentives of taxpayers and financial institutions to reduce the cost of government guarantees. We believe that state guaranty funds create fewer incentive problems than does deposit insurance because they encourage self-monitoring to mini mize the potential costs of LIC failures. The behavior of financial institutions also may be more effectively controlled by complementing regulatory oversight with market discipline. Discipline could be imposed by a specific class of creditors which is willing to monitor financial institution risk and bear the risk of loss. The FDIC Improvement Act of 1991 (FDICIA) extends some of the features that exist in the LIC industry to depository institutions. The act improves monitoring with the require ment that all depository institutions, regardless of size, that are determined to have insufficient capital must be closed, recapitalized, or other wise restructured. These provisions for prompt corrective action allow bank regulatory agencies to intervene early and thus reduce the exposure of the deposit insurance fund to losses. Other provisions of the act authorize the FDIC to im plement a system of risk-based deposit insurance with premiums related, in part, to the cost of future bank failures. Thus banks have greater incentives to monitor each other to keep deposit insurance assessments down. As the experience of the life insurance industry has indicated, private monitoring can reduce the cost of gov ernment guarantees. FOOTNOTES 'See Resolution Trust Corporation (1993). 2The term life insurance company refers throughout this article to firms classified as life and/or life-health insurance companies. 3General account assets equals total assets minus separate account assets. Separate accounts are defined as groups of assets designed as backing for specific obligations in which the investment risk is borne by the policyholder, and the insurer’s guarantee is limited to mortality and expense charges (see Saunders 1986). 4To be considered a “high” junk bondholder, an LIC in our sample must have a junk bond-asset ratio o f 6.6 percent, the industry average at year-end 1990. The remaining LICs were classified as “high” commercial mortgage loan hold ers if their commercial loan-asset ratio was greater than or equal to 21.6 percent, the industry-wide average at the end o f 1990. The rest were classified as “others.” 5Fenn and Cole (1992) analyze the impact o f policyholder behavior on the market value o f insurance companies in the event o f an insolvency. 6See Barrese and Nelson (1992). 7Harrington (1991) makes this point for property-casualty companies, which also benefit from state guaranty funds. FEDERAL RESERVE BANK OF CHICAGO 8LICs were not immune to the effects o f high interest rates. Because insurance policyholders had incentives to take out policy loans at below-market interest rates, LICs suffered from disintermediation. (Curry and Warshawsky 1986). 9See Kane (1989) for a discussion of the theory o f moral hazard as applied to S&Ls. For empirical evidence on the subject, see Brewer and Mondschean (1993a) and Barth, Bartholomew, and Labich (1989). l0See Furlong and Keeley (1989) for an analytical deriva tion of this result. "There is some evidence o f a loss of confidence in FSLIC insurance. Both Brewer and Mondschean (1992) and Strahan (1993) show that weak S&Ls paid higher rates for both wholesale and retail deposits than did well-capitalized institutions. Moreover, Strahan shows that weak S&Ls that did not raise their rates faced deposit outflows. 12Todd and Wallace (1992) detail the growth of GICs and SPDAs during the 1980s. l3These free rider problems may be contained by organiza tions such as the Community and Savings Banks o f Ameri ca and the American Bankers Association. 23 REFERENCES Barrese, James, and Jack M. Nelson, “Distributing Corcoran, Patrick J., “The credit slowdown of the cost of protecting life-health insurance consum ers,” The College of Insurance, unpublished paper, April 1992. c e e d in g s o f a C o n f e r e n c e o n B a n k S tr u c tu r e a n d Barth, James, Philip J. Bartholomew, and Carol Labich, “Moral hazard and the thrift crisis: an analy sis of the 1988 resolutions,” P r o c e e d in g s o f a C o n f e r e n c e o n B a n k S tr u c tu r e a n d C o m p e titio n , Federal 1989-1991: the role of supply and demand,” P ro Federal Reserve Bank of Chicago, 1992, pp. 445-462. C o m p e titio n , Cummins, J. David, “Risk-based premiums for insurance guaranty funds,” J o u r n a l o f F in a n c e , September 1988, pp. 823-839. Reserve Bank of Chicago, 1989, pp. 344-384. Brewer, Elijah, III, and Thomas H. Mondschean, “The impact of S&L failures and regulatory changes on the CD market, 1987-1991,” Federal Reserve Bank of Chicago, working paper, December 1992. ______________ , “An empirical test of the incen tive effects of deposit insurance: the case of junk bonds at savings and loan associations,” J o u r n a l o f M o n e y , C r e d i t a n d B a n k in g , forthcoming, 1993a. ______________ , “The impact of portfolio compo sition and leverage on the stock market valuation of life insurance companies,” working paper, Federal Reserve Bank of Chicago, 1993b. ______________ , “Risk exposure at life insurance companies: market evidence and policy implica tions,” C o n te m p o r a r y P o li c y I s s u e s , forthcoming, 1993c. Brewer, Elijah III, Thomas H. Mondschean, and Philip E. Strahan, “The effect of capital on portfolio Curry, Timothy, and Mark Warshawsky, “Life insurance companies in a changing environment,” F e d e r a l R e s e r v e B u lle tin , July 1986, pp. 449-460. Fenn, George, and Rebel Cole, “Announcements of asset-quality problems and contagion effects in the life insurance industry,” P r o c e e d in g s o f a C o n f e r e n c e o n B a n k S tr u c tu r e a n d C o m p e titio n , Federal Reserve Bank of Chicago, 1992, pp. 818-842. Flannery, Mark, and Christopher James, “The effect of interest rate changes on the common stock returns of financial institutions,” J o u r n a l o f F in a n c e , Vol. 39, No. 4, 1984, pp. 1141-1153. Furlong, Frederick, and Michael Keeley, “Capital regulation and bank risk taking: a note,” J o u r n a l o f B a n k in g a n d F in a n c e , 43, 1989, pp. 823-839. Harrington, Scott E., “Should the feds regulate insurance company insolvency?,” R e g u la tio n , Spring 1991, pp. 53-61. Kane, Edward J., risk at life insurance companies,” Federal Reserve Bank of Chicago, Working Paper Series, Issues in Financial Regulation (WP-92-29), December 1992. Brickley, James A., and Christopher M. James, April 1993. “Access to deposit insurance, insolvency rules, and the stock returns of financial institutions,” J o u r n a l o f F in a n c ia l E c o n o m ic s , July 1986, pp. 345-371. S ta te m e n ts : K e y s to S u c c e s s f u l R e p o r tin g , Wash ington D. C.: Urban Institute Press, 1989. T he S & L In su ra n ce M e ss, Resolution Trust Corporation, R T C Saunders, R. A., L if e R e v ie w , Vol. 4, I n s u r a n c e C o m p a n y F in a n c ia l Chicago: Teach’em, Inc., 1986. Cabanilla, Nathaniel, “Analyzing developments in the life insurance industry—commercial mortgage investments,” P r o c e e d in g s o f a C o n f e r e n c e o n B a n k S tr u c tu r e a n d C o m p e titio n , Federal Reserve Bank of Chicago, 1992, pp. 878-905. Strahan, Philip E., “The impact of the collapse of FSLIC on the market for insured certificates of de posit: theory and evidence,” University of Chicago, Ph.D. dissertation, June 1993. Calomiris, Charles W., “Deposit insurance: lessons from the record,” E c o n o m ic P e r s p e c t i v e s , Federal Reserve Bank of Chicago, May/June 1989, pp. 10-30. Todd, Richard M., and Neil Wallace, “SPDAs and GICs: like money in the bank?,” Q u a r te r ly R e v ie w , Federal Reserve Bank of Minneapolis, Spring 1992, pp. 2-17. Carey, Mark, S. Prowse, J. Rea, and G. G. Udell, “The private placement market: intermediation, life insurance companies, and a credit crunch,” Wright, Kenneth M., “The structure, conduct, and regulation of the life insurance industry,” in T h e P r o c e e d in g s o f a C o n f e r e n c e o n B a n k S tr u c tu r e a n d F in a n c ia l C o n d itio n a n d R e g u la tio n o f I n s u r a n c e C o m p e titio n , Federal Reserve Bank of Chicago, 1992, pp. 843-877. C o m p a n ie s , Federal Reserve Bank of Boston, Con ference Series No. 35, June 1991, pp. 73-96. 24 ECONOMIC PERSPECTIVES ECONOMIC PERSPECTIVES B U L K R ATE P u b lic I n f o r m a tio n C e n te r Federal Reserve Bank of Chicago P.O. Box 834 Chicago, Illinois 60690-0834 U .S . P O S T A G E P A ID C H IC A G O , IL L IN O IS P E R M IT NO. 1942 Do N o t F orw ard Address C o rre c tio n Requested R eturn P o stag e G uaranteed M a ilin g la b e l c o rre c tio n s o r d e le tio n s : Make changes or mark D elete F ro m M ailing List on the label and fax the label to: 312-322-2341, or mail to: ATTN: Mail Services, Federal Reserve Bank o f Chicago, P.O. Box 834, Chicago, Illinois 60690-0834 FEDERAL RESERVE BANK OF CF1ICAGO