Some of the changes generated by life insurance�s increased investment orientation are revealed by changes in the industry�s aggregate assets. (See table 32-1.)

The first�and a crucial�observation is that despite economic instability and rapid change, the growth rate of industry assets over the 21-year period from 1975 to 1996, at over 8.9 percent annually, has been ahead of the inflation rate and comfortably in line with the average of other financial institutions. The industry has avoided such disasters as have befallen the savings and loan and mutual savings bank industries. (The number of these industries has shrunk by more than one-third over the 21 years covered. Widespread failures occurred as a consequence of excessive risk-taking in real estate and frequent fraudulent practices, associated partly with inadequate supervision.)

The life insurance industry has also done better than commercial banks because the industry has successfully adapted its products to the public�s greater financial orientation. On the other hand, insurance has not grown as rapidly as mutual funds and money market funds because liquidity and flexibility in response to changing consumer demand have only gradually become life insurance industry attributes.

Investments generally considered long term (corporate bonds and mortgages combined) accounted for over two-thirds of total investments in 1975 but were down to about 62 percent in 1996. The industry continues to be a leading supplier of corporate debt and commercial mortgage funds. Yet the shift toward annuities and guaranteed interest contracts (GICs) as sources of funds, together with the need for liquidity in a volatile external environment, has led to major changes in the characteristics of these investments.

First, the average maturity of the "nonmarketable" part of long-term investments has been cut sharply. Crudely and imprecisely, as average maturity was halved, the annual turnover rate of the portfolio was doubled in the 21-year period covered in table 32-1 to roughly 10 percent. This permits a more frequent "fresh look at your money," including its use to pay off liabilities instead of reinvesting.

Second, both bonds and mortgage investments have been heavily redirected toward public securities rather than toward directly negotiated deals with debtors. Thus marketable bonds (including privately issued mortgage-backed bonds) have gained heavily at the expense of direct placements, and securitized

mortgages (especially including pass-through securities and collateralized mortgage obligations [CMOs] of federal housing agencies) have virtually become the exclusive life insurance method of investing in residential mortgages. (Efforts to securitize commercial mortgages are hampered by the individualized characteristics of commercial structures and by high underwriting costs associated with parcel-by-parcel securitization.)

Third, the tripling of United States government securities in their share of life insurance investments over the period studied is confirmation of the liquidity drive of the industry. (The $289 billion figure for U.S. Treasury and agency securities at year-end 1996 included over $130 billion of federally backed mortgage securities, where good yield or "portfolio fit" with liabilities may have counted as heavily as liquidity per se.)

Common stocks in insurance companies� general accounts are limited to
10 percent of assets under New York insurance law, which tends to dominate investment regulations throughout the country. The desirable liquidity characteristic of equities is deemed to be largely offset by the problem of substantial price and total-return fluctuations of common stocks. The industry�s general account holdings have remained well below the legal limit, but total holdings have risen to above 10 percent because separate accounts have gained relative to the general account. In such accounts, the nonguaranteed return of best-effort stock market investing flows through directly to the client. Separate accounts originated with the competition for corporate pension funds. Common stocks account for as much as 50 percent of aggregate pension fund assets because it is hoped that long-term equity gains may reduce the cost of pensions to companies. Hence insurance companies had to be granted separate-account powers (in the 1960s) to compete effectively for pension fund management.

Since the mid-1980s, the growth of investment choices and client discretion in growing individual products, such as variable life and deferred annuities, has provided a new vehicle for life insurance investment in common stocks�pooled separate accounts of individual clients. Bond funds for such clients are also growing. In the mid-1980s separate accounts passed the 10 percent mark as a share of life insurance assets, and they have gained further in the 1990s�a clear manifestation of the rising importance of investment management in attracting funds to insurance companies. Separate accounts are advantageous to the industry because the impending risk-based capital standards treat such accounts lightly; the client, not the company, bears the risk.

The growth of policy loans during disintermediation periods was the original reason for turning toward enhanced liquidity in life insurance portfolios. When market interest rates were above those that insurance companies were able to charge contractually, as occurred during tight-money episodes, it paid the insured to take out a policy loan. (The proceeds could be invested at a rate above that paid to the insurance company, or borrowing at rates in excess of the cost of a policy loan could be avoided.) Legal and regulatory changes, combined with redesigned contracts, have enabled companies to whittle down the policy loan share of assets. Loan rates are now indexed to a bond market rate, or if they are fixed, tend to be at 8 percent. Yet concerns about liquidity and policy loan earnings remain.

Life insurance investment has always been a spread business in the sense that management seeks a yield on investments in excess of the implicit rate credited to policies as their cash value grows over the years. Until universal life became a factor in the mid-1990s, however, this spread was known to very few and understood by even fewer. Business judgments and accounting factors relating to mortality, commissions, and operating expenses often helped to mask the investment yield spread and related issues. Once interest rates and/or equity performance became explicit, competition forced an all-around sharpening of pencils, and margins in general shrank. Meanwhile, the product portfolio was constantly expanding during the 1990s, and differential characteristics among products forced product-by-product asset management.

One obvious way to manage assets on a product-by-product basis is the individualization of product accounts. Thus whole life, universal life, single-premium life, variable life, variable universal life, single-premium annuities, equity indexed annuities, and GICs for corporate savings plans can each be considered a specialized asset/liability problem subject to individual solutions. This may involve establishing product-differentiated subsidiaries of a parent company, formal separate accounts, or internal segmentation of the general account. Creating product-differentiated subsidiaries or separate product accounts typically requires much legal work and regulatory approval. Segmentation, however, is almost entirely at management�s discretion, provided that the company makes no attempt to avoid its de facto responsibility for all liabilities, regardless of its internal accounting.

Investment management�s first task in segmentation is to develop criteria by which to distinguish investment policy by product. The scheme in table 32-2 helps to illustrate the criteria and possible outcomes of such deliberations.

This scheme is far from noncontroversial. Moreover, it is only a sample of products and criteria. Thus the intuitive results�such as life insurance requires less liquidity than annuities and GICs, and total yield requirements are highest in the most investment-oriented and interest-sensitive products�are subject to further evaluation. For example, the specific withdrawal provisions of the company�s single-premium deferred annuities (SPDAs) and the term distribution of the portfolio make a difference. Stiff back-end withdrawal penalties, a long stretch-out phase for the expiration of such penalties, and a large percentage of annuities to which the penalties still apply all work toward making the portfolio less susceptible to withdrawal and therefore to liquidity requirements, and vice versa. Because product and asset characteristics interact, they must be determined by simultaneous equations.

The recency of product differentiation and asset segmentation suggests that much in the type of scheme outlined here remains subject to review on the basis of future experience, especially for investment-sensitive products. The average duration of the huge amount of SPDAs put on the books in the late 1980s and 1990s is unknown, as is the percentage of outstanding SPDAs still subject to surrender charges. Nor have all of the new products yet undergone a full interest rate and economic cycle. Nevertheless, asset/liability management by product is a vital step forward in making (eventually profitable) business decisions, and enough time has passed to permit verification of the direct relevance of surrender terms to the likelihood of actual withdrawal.

A study of yield curves offers a crude, yet instructive, approximation of the practicalities of modern investment management. At any given moment, an almost infinite variety of interest rates of the market confronts the financial intermediary who seeks funds for investment. To protect against interest rate risk means matching assets and liabilities, at least approximately, along the yield curve (defined as an array of interest rates by maturity of the underlying debt).

The normal yield curve is sloped upward because price and credit risk�as well as uncertainty itself�tend to increase with maturity. The steepness of this upward slope, however, varies sharply with the business cycle phase. Generally the slope is steepest at or near the bottom of the cycle and flattens as full employment and inflation dangers approach. About one-fifth of the time from 1970 to 1990 the yield curve was inverted�that is, short rates exceeded long rates, typically under the influence of the extremely restrictive monetary policy during inflation and interest rate peaks. Normal yield curves have prevailed for most of the 1990s.

Figure 32-1 offers a sampling of yield curves for United States government and selected corporate securities at a reasonably "normal" time. It shows a total spread of about 9 percentage points between very short and highest-grade securities (90-day U.S. Treasury bills) and the longest low-grade corporate bonds (B-rated high-yield, or junk bonds of 10 years� maturity). Figure 32-1 also shows the relatively small spread between U.S. Treasuries and investment-grade

(A-rated) industrial bonds. Even a difference of one-quarter point in average yield on the portfolio, however, can be highly significant to an insurance company�s bottom line.

The first and most important point is that life insurance companies now operate throughout the maturity spectrum as they seek outlets for withdrawable funds generated by investment-sensitive insurance products. As late as the early 1970s, the industry operated almost exclusively at the long end of the yield curve (with only a marginal involvement at the shortest end).

A second key point is that the industry has also diversified by type of investment. As mentioned earlier, there has been a rise in public corporate bonds (compared to private placements) securitized mortgages, and directly negotiated deals. One may add junk bonds for general account or separate account invest-ment, a greater variety of common stock and bond portfolios for pension or new- product investments (growth, income, or balanced accounts, for example), and the increased involvement of a few companies in equity real estate through such devices as joint ventures with developers (the insurance company receives a share of the equity in return for supplying a long-term mortgage). This diversity has added to the complexity of investment operations in a new-product environment for insurance. It has also created new wrinkles in investments that enhance the ability to ferret out additional niches of funds (with or without a major insurance component).

It stands to reason that increased complexity brings difficult problems. Among these, the actual process of matching maturities of assets and liabilities stands out. The first difficulty is that retail insurance sales cannot be matched with regularity and precision in an investment operation that has to be wholesale to be efficient. The very nature of an insurance company therefore creates a degree of mismatch. The second difficulty is that the actual maturity of liabilities may be misjudged or be in fact unpredictable. The disintermediation episodes of the 1970s and 1980s showed that even whole-life-based liabilities could suddenly "mature" by having contracts surrendered or cash reserves borrowed against if market rates exceeded inside rates by a large margin. The late 1980s and early 1990s may well prove that the prepayment risk on high-interest securities is a very real one when the general interest rate level turns down. A third (and fundamental) difficulty is that the cost of acquiring the liabilities may preclude precise maturity matching at orthodox credit risk levels.

Failure to resolve this last difficulty was at the bottom of most insurance company crash landings of the late 1980s and early 1990s. The "best" solution� foregoing the business�may be unacceptable to top management. Other noninvestment answers include reducing the expense margins necessary to acquire and administer the liabilities (possible but painful) and stiffening the backload penalties for withdrawals and thus reducing their probability (also possible but subject to the same competitive pressures that caused the problem in the first place).

The temptation therefore is to do one of the following: (1) risk a mismatch by taking in assets with longer maturities than the most probable maturity of liabilities, which improves the calculated financials in a normal yield-curve environment or (2) take additional credit risk, thus obtaining higher returns at any given maturity of the yield curve. A third risk is diversifying investments insufficiently, which may or may not be directly related to the yield curve. Thus in the failure of Executive Life (California and New York, 1991), First Capital Life of California (1991), and Fidelity Bankers Life (Virginia, 1991) extra yield was obtained through excessive credit risk and concentration in junk bonds. (At Executive Life, maturity matching appeared to be very good until suspicions led to mass withdrawals, especially of SPDA funds, regardless of back-end penalties.) In the case of Mutual Benefit Life (New Jersey, 1991) an unusual concentration in one asset class and location�Florida mortgages�constituted a maturity mismatch, as well as an excessive credit risk, during a major commercial real estate recession.

The keen competition among asset managers has forced life insurance portfolio management to adopt highly technical and sophisticated evaluation and simulation models. The rapid evolution of academic finance and computer calculating capacity has also been a major factor in making the rigorous analysis feasible.

In practice, asset/liability matching is approximated through duration balancing of blocks of business (general account segments or their subsets). Duration, a traditional concept of bond finance, measures the weighted average maturity to term of a series of cash flows arising from an obligation. (For example, the less frequent interest and principal repayments are, the longer the duration of a bond. Thus duration can differ among two bonds of identical average and final maturity.) The duration of the corresponding liabilities tends to be even more difficult to estimate than that of assets, but even assets can deviate from initially calculated duration if restructuring or defaults occur, or if they have prepayment or call options. Thus the reinvestment risk for the large blocks of mortgage-backed securities acquired in recent years is considerable. The underlying mortgages can typically be prepaid when interest rates decline, and new securities can be substituted only at lower interest rate earnings while rates payable on the corresponding liabilities may not be equally adjustable.

In addition, duration itself changes with interest rate changes so the value of any portfolio will tend to diverge from initial calculations. A concept known as convexity measures the variability of portfolio asset durations in response to interest rate changes. Convexity helps to define acceptable limits on the asset side to the almost inevitable mismatches of duration of liabilities.

Barbell strategy is a concept applied in both liquidity and risk management. A portfolio may be forced into sufficient liquidity by having one concentration of assets at the shortest end of the yield curve while another concentration at the long end provides excess return over the yield requirements of the corresponding liabilities and thus offers a profit potential. Precisely the same strategy may be applied in concentrating on low-risk and high-risk assets rather than accepting middle-range risk throughout the portfolio. The measured weighted risk can therefore remain acceptable even while investment managers hope to exploit the potential extra return of, say, junk bonds or equity warrants. (Barbell strategy bears a family resemblance to traditional life insurance investment at only the shortest and longest maturity ends. Naturally, a barbell-weighted portfolio will not perform as hoped when the yield curve steepens. The long securities may decline in value, while the short securities gain little, if any, in price.)

Derivative securities transactions have multiplied immensely in financial markets and have gradually been practiced in life insurance investment management. New York insurance law restricts such transactions by characteristics and amounts; only hedging (and not speculating) is permitted, and amounts outstanding may not exceed 15 percent of admitted assets. Financial futures, options, and interest rate swaps have come into use, with hedging through Treasury bond futures contracts developing into a very broad and liquid market�possibly constituting the most frequent operation.

Suppose an insurance company has negotiated for a block of GIC money to be acquired 3 months hence at a rate guaranteed to the client for the next 5 years. Since the company does not yet have the cash to invest, it hedges by purchasing a 5-year U.S. Treasury bond futures contract at a yield known today to approximate (not necessarily match exactly) the guaranteed interest rate, with a forward delivery date roughly matching that of the expected funds delivery. The price change of the futures contract will then tend to inversely match the difference between the rate the company guaranteed at contract time and the rate actually available for investment when the cash is received. The insurance company has thus hedged the open position it assumed at the time of acquiring future funds and guaranteeing the rate it will pay.

The example above oversimplifies the process and the calculation and overstates the certainty of the result, but it accurately describes a practical method of reducing interest rate risk. One of the major possible refinements is the use of a combination of hedges�synthetic hedges�to closely approximate the terms of an interest rate guarantee that may have been extended but for which a direct public-market equivalent may not be available.

The sale of covered call options against a portfolio of marketable bonds illustrates permitted use of options. In broad markets, as exist for United States Treasuries, there will almost invariably be buyers of the right to purchase bonds at a predetermined "strike" price. This price may slightly or substantially exceed the actual cash price. Selling the option enhances current income on the bonds beyond their stated yield, but it gives up potential future price gains to the purchaser of the option. The operation is virtually riskless if the company holds the exact optioned bonds in its portfolio, and it is appropriate provided the client�whether the insurance operators or an outside investment client�understands the trade-off to obtain the extra yield.

Modern portfolio theory can also be applied to common stock strategy. Offering investment management to pension clients through commingled or individual separate accounts requires insurance companies to handle "efficient frontier" calculations and to manage accordingly. It is assumed that risk and total return on equities will rise or fall jointly, and risk is typically measured by the price variability (the beta) of the stock relative to an average or index of stocks. (A beta larger than one means that the stock fluctuates more widely than the market; a beta smaller than one indicates that the stock fluctuates less widely.) Exposing the trade-off between risk and return determines the point on a constructed convex efficient frontier line that corresponds to the client�s preference and thus attracts his or her business. If the preference turns out to be a simple averaging of stocks in the market, the company can develop and offer an index fund with expected results similar to the market as a whole.

While analyses clarify risks and technical operations ameliorate these risks, neither analysis nor execution is cost free. Typically, the potential gain is reduced along with the risk, and although uncertainty is reduced, it is not eliminated. Thus investment management can avoid extremes, but the fact remains that obtaining satisfactory investment earnings involves risks. The risks increase sharply if a significant contribution to a company�s bottom line is expected from the investment function. Such expectations may lead to untoward investment risk. Once bottom-line pressures are caused by insurance product terms and/or expenses that cannot be satisfied with conservative investments, investment management may be tempted to go out on a limb. When good investment judgment is overridden, technical expertise becomes largely irrelevant. Regulators and rating agencies have become increasingly alert to this problem since the recent investment-related failures.