The development of actuarial-based pension accounting at the Bell System, 1913-40

1. Introduction

The initial development of pension funding and accounting practices at the Bell System from 1913,1 was an outgrowth of President Theodore N. Vail's "universal system" program that was designed to strengthen the firm's managerial and technological capacities (Garnet, 1985, Ch.9, pp. 128-154). Following the company's near financial collapse in 1907, the House of Morgan recruited Vail to revitalize the firm. The new president (whose incumbency lasted to 1919) successfully overcame long-standing weaknesses in operational control by promoting a greater standardization of network technology and management practice. Vail also modified the early strategy of horizontal growth dating back to the firm's founding in 1877, which involved the granting of exclusive franchise territories and the leasing of firm equipment. Instead, the parent company began to absorb semi-autonomous local entities and linked them together nationally through the creation of a comprehensive long-distance network (Garnet, 1985, Ch. 5, pp. 55-73). The organizational challenges had become more complex with the acquisition in 1882 of Western Electric, a major producer of telephone equipment (Smith, 1985, Ch.5; Adams & Butler, 1992, Ch.2, pp. 45-70). By 1915, Vail's rationalization drive had achieved great success (Garnet, 1985, Ch.9). The firm dominated its industry with total assets in excess of US$1bn, with 6.2 million telephones (plus another 3.0 million largely connecting from independent companies to its national network) and 156,000 employees (AT&T, 1916, pp.53-4).

Pensions became an attractive perquisite at AT&T and other giant enterprises early in the twentieth century because they seemed responsive to the new circumstances brought about by the rapid expansion of the nation's urban-industrial economy. When American society had been predominantly rural and agricultural, the satisfaction of the economic wants of the elderly frequently came from support rendered by extended family units or from the self-sufficiency inherent in farming life. With industrialization, however, urban workers could neither rely on such sources of assistance; nor did contemporary private and local governmental welfare agencies have the resources necessary to assist large numbers of unemployed elderly. Thus, a corporate promise of income to provide a buffer against the vicissitudes of old age, doubtless, seemed highly attractive to prospective employees (Sass, 1997, pp.4-13). AT&T was not the first to introduce an employee pension plan. The Canadian Grand Trunk Railroad's plan of 1874 was the first corporate pension scheme in North America (Sass, 1997, pp.16,19). The American Express Company formed the first corporate pension in the USA in 1875. By 1900, a few progressive businesses such as Sherwin-Williams, Proctor and Gamble, Solvay Process and John Wanamaker also instituted similar programs (Stone, 1993, pp.253-4; Sass, 1997, pp.16,19). The federal government maintained the largest pension plan with about one million claimants, primarily veterans of the Civil War (Sass, 1997, p.12). By the first decade of the twentieth century AT&T joined the ranks of large companies providing such benefits. In 1907, AT&T's new president, Theodore Vail set up a special committee to design the company's comprehensive welfare program. Walter S. Gifford, the company's statistician and future president, assisted the committee by conducting the first census of the company workforce (Miranti, 2002).

One purpose of pensions involved the stabilization of management and the reduction of the high costs of employee turnover. Labor historians have tended to attribute such forms of "corporate liberalism" as paternalism intended to dilute worker enthusiasm for forming assertive unions (Lustig, 1982; Ernst, 1995). What this historical school has overlooked is the importance of these benefits in the retention of a rising but crucial class of professional labor that rarely formed unions. AT&T relied heavily on a growing body of specialists for the successful implementation of its high technology operations. As Moses Abromowitz and Paul David have explained, the early twentieth century witnessed the beginning of a shift in the drivers of productivity in the US economy from investment in capital assets to the investment in skills and knowledge (Abromowitz & David, 2000, pp.40-60). The effective operation of a complex organization increasingly depended on the services of a workforce with high levels of literacy and numeracy. AT&T was a beneficiary of the upgrading of human capital that derived from the great expansion of public secondary education beginning in the 1880s. It also benefited from the growth of collegiate and graduate education especially in such intellectually challenging fields critical to telephone operations as electrical engineering, physics and mathematics (Abromowitz & David, 2000, pp.54-60).

The investment in knowledge necessary to develop a comprehensive pension plan also helped to reduce the risk of the firm by providing more precise and reliable information about the magnitude and timing of liabilities and future payouts. This was vital for effective financial planning. It also involved the need for organizational modification to assure the effective integration of special knowledge in such diverse fields as accounting, actuarial science, law, taxation, finance and management.

Pensions helped to create a positive image of large business enterprises that frequently had come under attack by contemporary Progressive reformers. For example, each of the candidates for president in 1912 - Theodore Roosevelt, William Howard Taft and Woodrow Wilson - incorporated strong planks in their platforms for stronger government action against the rising power of big business (Chandler & Tedlow, 1985, Ch.21, pp. 551-580). The contemporaneous corporate sponsorship of pension programs, however, tended to offset this negative perception. It seemed to reflect the ability of giant enterprise to contribute to progress by materially improving worker welfare. It also augmented the public image of big business because such programs promised to resolve a serious social problem that government had hitherto not been able to address adequately (Marchand, 1998).

Pension development modified the relationship between major enterprise stakeholder groups by establishing new contracts and channels of communication. First, authoritative guidance for reporting pension liabilities remained obscure during this era. The chief regulatory agency, the Interstate Commerce Commission (ICC), which had responsibility for prescribing uniform accounting reports for the telephone industry under the Mann Elkins Act of 1911, encouraged disclosure but did not detail measurement methodologies. Most state oversight authorities followed the federal example. Moreover, the problem of pension accounting was not addressed by any professional accounting body until the Committee on Accounting Procedure at the American Institute of Accountants (forerunner of the American Institute of Certified Public Accountants) issued Accounting Research Bulletin 47 in 1956. Second, the firm was compelled to draw on the expertise of external actuarial consultants to define the provisions of the pension plan and to implement actuarial methods. Third, the evolution of the plan affected the relationship between the firm and governmental officials who decided which costs could be deducted for tax purposes and which could be included in the rate base. Fourth it led to the development of new relationships with financial institutions, who took on the responsibility of maintaining custody over pension resources. Fifth, the new contracts modified the claims on the division of corporate cash flows between investors, management and employees.

In evaluating the Bell system's experience with pension accounting we draw upon an analytical construct, the "integrated learning base", put forth by Alfred D. Chandler in The Electronic Century. This line of inquiry focuses on the ways that firm-specific learning for such new activities as pension management functioned as a driver of organizational change. This knowledge building becomes embedded within businesses by shaping the development of managerial procedures and organization. In the Chandlerian model, a firm's "integrated learning base" may be broken down into three major components. The first, "technical capacities", relates to the ability of the firm to foster growth through the exploitation of technology and the creation of new products, processes or services. The second, "functional knowledge", involves a broad array of supporting activities necessary to commercialize innovation such as manufacturing, marketing, distribution. The third, "organizational capacities", include such high level staff activities as defining strategies, allocating resources, coordinating operations and evaluating performance (Chandler, 2001, Ch.1, pp. 1-12). The drive to extend learning about pension liabilities in the Bell System primarily affected the development of functional knowledge and organizational capabilities while having virtually no impact on technical capacities. Moreover, the principal factors that limited the scope of this initiative were, as we shall see, the sufficiency of internal funds to finance change and the policies defined by both tax and regulatory authorities for the treatment of such expense items.

The following sections of this study evaluate these and other questions about the Bell System pension experience. The second section analyzes the cash-basis system that was launched in 1913. The third considers the factors that impelled the firm to establish an actuarial-based system of pension accounting in 1927. The fourth section reviews the actuarial methods used by AT&T to compute its pension liability. The fifth section discusses some of the problems encountered in liability estimation and how this influenced managerial effectiveness and regulatory compliance. The last section discusses what this experience tells us about the nature of the learning at the Bell System and its role in shaping the development of organizational capacities.

2. The pay-as-you-go system, 1913-27

On 1 January 1913, AT&T established its first system-wide pension plan. It was financed through an employee benefit fund, which also provided compensation for sickness, disability and death liabilities. The new arrangement sought to promote the broader goal of the universal system program for enhancing managerial effectiveness by making the economic interests of employees and the corporation more concentric (Marchand, 1998, p.117). President Theodore Vail stressed these mutual benefits in announcing the approval of the plan:

The intent and purpose of the employer in establishing a plan of benefits is to give tangible expression to the reciprocity which means faithful and loyal service on the part of the employee, with protection from all the ordinary misfortunes to which he is liable; reciprocity which means mutual regard for another's interest and welfare. (House of Representatives, 1939, p.455)

The firm's labor force was classified into three functional groupings. Group 1, spanning job classifications 100 to 600 that included the managerial, professional and technical labor, were the primary beneficiaries of the plan. This highly educated and skilled cadre consisted of "general officers and their assistants; operating officials and assistants; attorneys and right-of-way agents; engineers, draftsmen, surveyors and student engineers; and accountants". Category 2 consisted of clerical employees and Group 3 represented all other workers (Federal Communications Commission [FCC], Exhibit 136, 1937, p.5).

The pension program helped to satisfy regulatory imperatives for efficient service by addressing the problem of "superannuation". By 1915,40 states had established oversight commissions who in evaluating rate increase requests considered the firm's record in satisfying two mandates (AT&T, 1916, p.28). These were: economy, which related to the reasonableness of the cost of service; and efficiency, which related to the quality, reliability and promptness of service (Miranti, 2005, p.47). Pensions softened the economic adjustment caused by mandatory retirement at 70 for employees whose work efficiency may have declined because of the onset of old age. Such a policy was further deemed beneficial because it created opportunity for advancement for younger employees who, because of rising national educational standards, may have possessed a more desirable range of skills than typically found in the older generation of employees. Personnel shortages, however, during World War I led to the abandonment of this rule in 1917. It was not reinstituted until 1930 when it was set at 65 (FCC, Exhibit 581b, 1936, p.2). The revised age rule led to a sharp jump in retirements between 1930 and 1934. During that period, 4,332 employees retired (3,347 males and 985 females), representing 72 percent of the total pensioned retirees since the plan's inception (FCC, Exhibit 581b, 1936, p.l).

The pension plan employed three vesting classifications. The first, Class A, had a guarantee of a pension if they reached age 60 for men and 55 for women and had 20 years service. There was no vesting period specified, but effectively, the vesting period was 20 years. The pension benefit amounted to one percent for each year of service times the employee's last 10 years' average income. Although the other two alternatives used the same benefit base, they were not guaranteed but depended on the discretion of an employee benefits board made up of senior corporate executives. Class B pensions required 25 or more years of service for male employees aged 55-9 and for female employees aged 50-4. Class C pensions required 30 years of service and age of less than 55 for men or 50 for women (FCC, Exhibit 583,1937, pp.23-38).

It is clear that pension benefits vastly favored the high ranking employees, due to longer terms of service, steeper salary trends and a more generous pension plan that did not provide a ceiling on pension payments. For comparable years of service, higher salaried employees like Executives and Vice-Presidents received a pension that was a much larger percentage of average salary than lower salaried employees like janitors and clerks. In the New York Telephone Company, for instance, a select group of retired executives received pensions amounting to US$1,050 per month, which was about 15 times the average pension paid to other pensioners in the company. For the Bell Telephone System as a whole, as of 31 December 1934, 33 former executives (0.7% of the total pensioners) received average monthly pensions of US$903 each, while 2,300 low-salaried employees (46.4% of pensioners) received average monthly payments of only US$36 each. There was also a strong gender disparity in employment and pensioners. The number of male pensioners was more than three times the number of females in 1934. Figure 1 shows the trend in the number of pensioners by gender added to the service pension rolls in the period 1913-34. By and large the male-female ratio held steady during the entire period. Further, female pensioners received much lower pension amounts relative to male pensioners, as most of them were low-level employees. The pensioners at the lowest end of the benefit spectrum were predominantly female (FCC Exhibit 136, 1936, Special Docket No.1).

From 1913 to 1927 the fund had no specific assets and simply represented an appropriation of the earned surplus on the books of the parent and each of the participating operating subsidiaries. Although each of the 59 active corporations within the system had separate pension contracts, their details were in broad conformity with one another. AT&T the parent company served as the custodian. Payments of benefits reduced the appropriation and were charged off against operating expenses. The initial appropriations amounted to US$8.9m.The fund accrued interest at four percent on its outstanding balance. Payments from the fund, however, were restricted to a maximum of US$500,000 annually for the operating subsidiaries and up to 0.2 percent of payroll for AT&T, the parent. It increased from 1 January 1913 to 1 July 1928 by US$26.5m because of additional appropriations and by US$50.5m because of accrued benefit expenses. The fund paid out benefits amounting US$51.3m during this period of which only US$4.2m represented pension payments. By the end of the first stage the benefit appropriation amounted to US$34.8m (FCC, Exhibit 583, 1937, p.75).

With the inception of the federal income tax in 1913, pension plans provided tax deductions to the sponsoring company. They could deduct pension benefit payments and deferred annuity premiums as business expenses. In 1918, these deductions were expanded to include contributions to pension funds organized separately from the sponsoring company. In 1921, pension trusts were also exempt from paying taxes on investment income. Further, employees were taxed only when they received pension income.2 The 1928 Revenue Act provided tax deductions to pension plan sponsors who shifted assets to trusts or insurance companies to fund pension liabilities. For funding that represented previously accrued liabilities, the deduction had to be spread over a 10-year period (Sass, 1997, pp.102-9).

The pension fund's accounting structure also benefited a capital-hungry firm like AT&T with many internal investment opportunities to finance. The benefit appropriations reduced the amount of earned surplus that could qualify for payment as dividends to shareholders. This was advantageous as the firm became deeply committed to the development of automatic switching capabilities for a rapidly growing telecommunications market (Gherardi & Jewett, 1930, pp.21-2).

Unlike death, disability and sickness payments, which remained relatively constant as a percentage of payroll, pension payments increased at a faster pace. The firm paid out over US$47m in non-pension benefits in 546,112 cases during its first 15 years - costs that would have been incurred even if there had been no benefit plan. Pensions payments, on the other hand, grew from US$20,000 or 0.02 percent of payroll in 1913 to US$633,701 or 0.15 percent of payroll in 1927 (FCC Exhibit 583b, 1937, Appendix 15, Sheet 1). The accelerating trend worried management especially in light of the failure of the Morris Meatpacking Company's pension plan in 1923 (Sass, 1997, pp.56-7). Moreover, the annual cost at some subsidiaries began to exceed the upper boundaries specified earlier with respect to absolute dollars and percentages of overall payroll. As we shall see in the next section, these new pressures relating to pension finance soon motivated the firm to explore the utilization of an actuarially-based system in order to better control costs.

3. The transition to actuarial-based pension costing

AT&T switched to actuarial-based pension accounting in 1927 It was also possibly the first major uninsured corporate sponsor to employ an in-house actuary to determine its funding obligations based on future projections during the employees' active working years (Sass, 1997, p.83). AT&T's receptivity to the use of probability theory and actuarial methods for enhancing control over the uncertainties and risks in pension management was partially rooted in its pioneering use of this knowledge for solving engineering problems. Beginning in 1905 the firm began to apply probability theory to the problems of message queuing and later in the design of automatic switching stations (Miranti, 2002). In 1912, it had applied probability theory to plan the disposition of magnetic loading coils for enhancing the power and range of its telephone signals in building its first coast-to-coast long distance line (Crisson, 1925). By the 1920s, Bell engineers had created the intellectual foundations of statistical quality control, which used probability theory to bolster the effectiveness of product inspection activities (Miranti, 2005). During this same era the firm also applied this special knowledge for estimating the useful lives and depreciation rates for its enormous investment in fixed assets (FCC, 1937; AT&T, 1957, Chs.5-6).

As part of Theodore Vail's reforms, AT&T began to build up its accounting and statistical capacities by forming in 1909 a statistical division under the direction of Malcolm C. Rorty. This unit included the accounting department, which among other duties, subsequently had responsibility for pension liability reporting. It also included a special telephone statistics group, which monitored the performance of both Bell associated companies and competing independent telephone enterprises. In 1912, the division created a public utility section to analyze rate regulatory data nationwide and the efforts of some advocacy groups to promote telephone nationalization. After World War I, Walter S. Gifford became division chief and expanded its scope by launching departments for financial and economic statistics, special statistical analysis, statistical methods and mathematical studies (Miranti, 2002, pp.743-4).

The accounting department began studying the feasibility of shifting to an actuarially based system in August 1924. The comptroller, Charles A. Heiss, assigned his assistant RW. Saxton to determine what the firm's actual liability was assuming a six percent interest on fund balances. Saxton estimated that the balance would amount to US$94m at the end of 1925. Its funding would require a US$74m appropriation of surplus. This represented about a third of AT&T's surplus (less capital stock premiums). Moreover, the allocation of this burden to the operating subsidiaries would drastically reduce their combined equities from about US$82m to US$12m. Some subsidiaries such as Pacific Bell, Illinois Bell and New England Telephone did not have sufficient equity to offset their proportional burdens. Saxton also pointed out that the payments to cover the interest on fund balances might not qualify as operating expenses and, thus, might not be considered as part of the recoverable costs in rate cases before commissions or courts (FCC, Exhibit 583g, 1937, Appendix 21).

Based partly on this report, Heiss recommended on 10 August 1926 the measurement of pension liabilities on a full accrual basis and the use of actuarial-based estimation. In his view the prevailing pay-as-you-go system materially understated the firm's true pension liability and also the amounts that might ultimately be recovered through charges to the firm's rate base. He identified three alternatives for measuring these costs. The first involved spreading pension costs over the estimated lives of all retired employees. The second spread the costs over the lives of active employees after their fifteenth year of service. The last allocated such costs over an employee's entire working life (FCC, Exhibit 583a, 1937, Appendix 10).

In 1927, the firm initially embraced actuarially based pension accounting and selected the 15-year service option in calculating pension costs. Corporate attorney Roscoe T. Holt researched precedents in the law to better define the firm's commitments to employees; corporate general counsels, R.T. Guernsey and C.M. Bracelen, considered the organization of the new arrangements and their impacts on the regional subsidiaries (FCC, Exhibit 583d, 1937, Appendix 17; FCC, Exhibit 583e, 1937, Appendix 18). A key change involved the abandonment of the surplus appropriation in favor of the establishment of an irrevocable trust fund for maintaining pension assets at Bankers Trust Company, which served as custodian (FCC, Exhibit 583,1937, pp.36-40). AT&T, however, retained the power to manage the fund. Payments for sickness, disability and death continued on a pay-as-you-go basis. The new contract abandoned the old limits of two percent of payroll or US$500,000. Under the new plan and subsequent revisions, interest accrued on the total pension liability obligation at four percent per annum. The Interstate Commerce Commission and most state authorities accepted the arguments that pension payments were part of employee compensation and, thus, a valid operating expense that impliedly could be included in the rate base (FCC, Exhibit 583,1937, pp. 15-16). In 1928, the reserve in the corporation's equity held for the employee benefit fund was discontinued and its balance of US$34.8m was paid into the trust fund at Bankers Trust (House of Representatives, 1939, p.463).

The move to accrual accounting resulted in a sharp increase in the amount of pension expense recognized. As seen in Figure 2, pension expense as a percentage of net income shows a steep jump in 1927. The year 1927 also saw a steep increase in revenues of the parent company, largely as a result of increased dividend revenues from the subsidiaries. The increased revenues were able to more than absorb the increased pension accruals, allowing for a sharp increase in net income as well (Figure 3).

Another committee composed of Heiss, corporate attorney, John H. Ray and statisticians, Harry J. Brandt and S.L. Andrew, recognized that the transition to an actuarial based pension system would place stress on the finances of some associated companies and required a significant investment in new corporate learning about how best to compile, analyze and report data for about a quarter of a million employees. They believed that it would take several years for the firm to develop such internal capabilities. For these and other reasons, they believed that a 10-year transition period was necessary to place the firm's system on a full actuarial basis (FCC, Exhibit 583d, 1937, Appendix 16). While the record is not clear, it seems that such considerations may have influenced their recommendation that employee pension costs be calculated beginning with the fifteenth year of service. The size of the workforce had increased almost tenfold from 37,067 in 1900 to 364,045 in 1929. Consequently, the selection of the 15-year measurement alternative significantly reduced the size of the employee population and the corporate resources necessary to support the analysis of pension costs. Moreover, the total payment of pension benefits (as opposed to accrued expenses) projected for 1927 was 0.17 percent of payroll, an amount that was within the acceptable range under the old contract (FCC, Exhibit 581,1936, pp.11-14).

The new arrangement lasted only one year before the firm decided to modify the plan by spreading pension costs over an employees' entire working life. In January 1928 a committee composed of President Walter P. Gifford, Vice President David F. Houston and two board members, Charles Francis Adams and J.S. Alexander, guided by the findings of two consulting actuaries, recommended a change to full service costing, that is, spreading pension costs over the all of the years of an employee's service. The committee had engaged Joseph H. Woodward of the firm of Woodward, Fondiller and Ryan of New York to test the validity of the original actuarial-based pension costs. They also engaged consulting actuary George H. Buck, who had helped to design the pension plan for the government of New York City, to evaluate the current system and to propose changes for more effective operation. Buck's most influential suggestion involved beginning the calculation of pension costs at the time of hiring rather than after the fifteenth year of service. The main advantage of full service costing resided in the fact that it reduced the variability in annual pension service costs. Under Buck's full service estimate the normal pension cost would stabilize each year at about 2.7 percent of payroll (not including any previous deficits for past service dating back to the plan's inception in 1913). This was at a variance from the alternative 15-year calculation, which was estimated to drive the annual pension cost percentage from 1.9 percent in 1929 to 5 percent during the 1930s (FCC Exhibit 583c, 1937, Appendix 15).

The choice of full service costing offered several advantages for both rate hearings and corporate finance. First, in rate cases it seemed less likely that the steady accrual rate of 2.7 percent under full service costing would be challenged as unreasonable as compared to the rising rates under the 15-year alternative. Second, the methodologies applied and the average cost percentages projected using the full service method were expected to be acceptable to rate boards because they closely approximated the experience and practice of several state pension funds. Third, several of the financially weaker associated companies would find it less onerous to bear a constant 2.7 percent projected annual normal pension charge rather than the rising rates implicit under the 15-year rule alternative (FCC, Exhibit 581, 1936, pp.124-9). Further, having a steady rate would make pension accruals less uncertain and easier to account for.

More problematic, however, was the treatment of what AT&T accountants called the "accrued liability", or what modern accountants term "past service cost", for employee pension credits earned from the initiation of the original plan in 1913 to the adoption of the accrual system in 1927. At the end of 1926 this liability was estimated to amount to US$177.8m. It was also not clear whether the past service cost could be charged off against the rate base or would be required to be written off against earned surplus by state regulators. George Buck persuaded the firm to measure these costs separately on the assumption that they would be amortized over a 30-year period. This would have raised the total pension cost accrual to 4.0 percent. While Buck could not anticipate the outcome of future regulatory cases with respect to past service costs, he believed it was imperative that the firm accurately measure these balances to support future planning as to their eventual disposition. He also believed that the firm should seek permission to increase its normal cost by an amount necessary to cover the four percent annual charge assessed by the firm on its total pension obligation. This would stop any further compounding of the prior service liability. As we shall see later, however, the firm's slowness in implementing Buck's recommendation concerning the funding of interest charges created regulatory difficulties during the 1930s (FCC, Exhibit 583b, 1937, Appendix 15).

Tax policy also influenced pension planning. The Revenue Act of 1928 facilitated the adoption of actuarially based systems. This law required the firm to file both a copy of its pension plan and the actuarial calculations on which it is based with the Internal Revenue Service. Qualifying plans could deduct both current payments to the pension fund and pension expense accruals. Pre-existing accumulated reserves such as those that existed on AT&T's balance sheet since 1913 could also be deducted for tax purposes over a 10-year period (Latimer, 1932, pp.660-4; Sass, 1997, pp.102-9).

The financial management of the fund sought both to stabilize its interest rate and, thus, facilitate the calculation of the accrual and to subsidize the firm's business activities. Under the agreements with Bankers Trust, AT&T and its associated companies retained the right to structure their individual corporate portfolios. AT&T fully funded its pension liability by depositing its own demand notes in the trust fund maintained at Bankers Trust. These notes initially paid annual interest of 5.5 percent. Later, during the Great Depression in 1932 the firm reduced the interest on these notes to four percent. This practice helped to simplify the projection of future pension liability. It was also close to the long-term rates that the firm paid on its debentures and provided an implicit subsidy to the firm whose long-term returns on net book value of fixed assets averaged 6.67 percent during the period 1913-35 (FCC Exhibit 582, 1936, pp.65-74). The funded pension balance increased from US$6.6m in 1927 to US$151m in 1935 (FCC, 1936, Exhibit 582, p.16).

During the Great Depression, AT&T began to diversify the assets in its pension fund by beginning to purchase its own debentures and those of its subsidiaries and US Treasury securities. The firm began to acquire its debentures partly as an effort to maintain a market for these securities during a period of severe financial stress. Since 1920, AT&T had made a strong effort through the formation of its financing subsidiary, the Bell securities Company, to broaden public holding of its securities, particularly bonds (AT&T, Annual Report 1922, p. 16; FCC Exhibit 250, 1936, pp.1-16). Ownership had been heavily concentrated in New York and Massachusetts. The sale of debentures for the regional subsidiaries would broaden support for the companies and its policies nationwide and in that way also help to smooth its relationships with state regulatory authorities in the South and the West whose members may have harbored fears about the dominance of "Eastern" business and financial interests. Without the market support efforts of the firm, such relationships may have been jeopardized because of the capital losses experienced by bondholders in otherwise illiquid markets. The shift to debentures in pension finance also better enabled the firm to deflect criticism of using securities whose values were set administratively rather than through the operation of market forces. Moreover, the firm's trading activities held out the promise of capital gains that would contribute to lower pension costs. Later, the Illinois subsidiary began to acquire US Treasury securities (FCC, Exhibit 582a, 1936, Appendix 6). This latter choice may have reflected a growing concern about investment risk and liquidity with the long continuation of the Great Depression. It may have also reflected the diminishment of opportunities to purchase corporate securities with the economic slowdown and the inter-corporate financing restrictions imposed by the Public Utilities Holding Company Act of 1935.

4. Computational methodology: normal pension cost

The data employed in calculating the pension obligation derived largely from an annual census that the firm had originally started to compile in 1920 under the direction of Walter R Gifford. Donald T. Belcher, the assistant statistician in charge of pension measurement, decided that firm-wide averages derived from the census represented a more efficient way of developing statistically relevant information rather than undertaking individual calculations for each subsidiary entity. In addition, differences in employment patterns apparent in the census induced Belcher and his cohorts to stratify the analysis of data on the bases of gender and age at time of employment. They discovered a reasonable degree of homogeneity when they disaggregated the data between men and women and between six employment age strata: 15-20,20-5,25-30,30-5,35-40 and 40-5. The specific factors from whence the annual accrual rate for normal pension cost derived included estimates of: (1) the future separation from employment for all causes; (2) future wage payments; (3) the amount of the pension; and, (4) the present value of future pension payments. For these and other calculations the comptroller's department used five year moving averages in order to smooth any unusual fluctuations. The firm regressed this data through a process of "graphical gradation" for projecting to future periods what was termed the "observed rates" of change for total costs or any of its component elements. The estimates of the mortality of retired employees derived from a standard insurance industry reference, The American Annuitant. In addition, the estimate also incorporated an annual interest rate assumption that amounted to four percent on fund balances beginning in 1932 (FCC, Exhibit 583g, 1937, Appendix 31).

In estimating probable service periods, Belcher's department employed a methodology similar to that used by insurance companies to create mortality tables from a record of insured lives. Table 1 projects a hypothetical separation rate for each year for a class of male employees of 10,000 entering service at age 25. The annual separation rate for each year derives from the smoothing using graphical gradation of five years of data. Implicit in this analysis is an exceptionally high rate of turnover during the first five years of service (over 65 percent), the period under current Employee Retirement Income Security Act (ERISA) regulations required to vest fully in a pension plan. The example also indicates that only 14.6 percent of the employees beginning careers at age 25 were expected to receive a pension (FCC, Exhibit 581a, 1936, Appendix 1).

A similar pattern was also followed in projecting lifetime wages for the group. Table 2 reports total employees who joined at age 25 and their expected average wages over the course of their careers. Like the separation rates, the wage data resulted from a projection based on a graphical regression of five years' data. Each year's total wage is the year's average wage times the number of employed workers (FCC, Exhibit 581a, 1936, Appendix 1).

The determination of the amount of pension payments for each class can be assessed by multiplying the number of employees retiring each year (Table 1, Column 4) by their last 10 years' average wage (derivable from Table 2, Column 4) times a percentage equal to the total number of years of company service (FCC, Exhibit 581a, 1936, Appendix 1).

The present value of the pension at the grant date depends on the amount of the pension, the duration of the pensioner's life and the interest rate used to discount the annuity (see Table 3). The Bell System used life expectancies derived from the American Annuitant, a standard reference used in the insurance industry. The discount factor was four percent, which besides being a standard in the insurance industry, also was equivalent to the rate used by AT&T to accrue interest annually on its pension obligation. The present value of the pension income at date of grant is equivalent to the total projected pension payment times a four percent annuity factor (FCC, Exhibit 581a, 1936, Appendix 1).

The determination of the annual accrual rate for pension cost is derived from dividing the present value of all wage payments by the present value at the date of entry of the projected pension benefit discounted at four percent (see Table 4). The resultant rate when applied to each year's wages should accumulate a balance sufficient to fund the entire pension costs for the class of employees at their retirement (FCC, Exhibit 581a, 1936, Appendix 1).

5. Computational problems and their effect on management and regulatory compliance

While the creation of the actuarial-based system represented a great achievement for the Bell System, the data and methodologies that it employed to measure pension liabilities in some cases complicated management and regulatory relationships.

Problems of data quality adversely affected liability estimation. The firm had rejected as too costly a recommendation made by Woodward, Fendiller and Ryan that all 350,000 Bell System employees be required to submit annual information cards to support the pension accrual. Instead, it relied on a general annual census that later was discovered to be fraught with many compilation errors. One such problem was that of "negative separation". This involved the subsequent discovery that previous years' separation data was misstated. Some of these errors resulted from employee carelessness and inattentiveness. In other cases the discrepancies arose because of the inability of the census documents to track accurately special employment circumstances. One such condition was the transfer of personnel between various corporate entities within the Bell System. Another called "service breaks bridged" related to the inability of the census reporting system to account adequately for credit earned from previous company service for employees that were subsequently rehired (FCC, Exhibit 581, 1936, pp.47-57).

The system-wide averages used to determine pension costs for the parent and its associated companies also at times deviated significantly from the experience of individual operating subsidiaries. The resultant accrual rates, derived from averaging the experience of many different subsidiaries, often seemed at a variance with experience in particular operating units. Moreover, questions about the reliability of such rates also could weaken a subsidiary's claim in rate hearings as to the reasonableness of such charges for inclusion in the allowable rate base. Because of this and other factors, top management contemplated centralizing the pension function at AT&T and charging back to the subsidiaries their allocable costs through the annual management fee (FCC, Exhibit 581, 1936, pp.66-8).

Radical changes in market circumstances also undermined pension accrual estimation. The projection of trends developed from regressing base period data implied the continuance of past trends into the future. The turning point was the Great Crash and Depression, which was not adequately adjusted for in the firm's development of separation and wage rates. During the early years of the 1930s the firm continued to base these projections in part on pre-crash data. This probably reflected a belief that the economic downturn would not be as deep and as long lasting as it turned out to be. Consistent with that perception was the corporation's promise to restore all laid-off workers to full pension benefits if rehired within two years. Because of the slowness of the recovery, regulators came to believe that the firm's methodologies contributed to an over accrual of pension costs through 1934 (FCC, Exhibit 581,1936, pp.63-6).

The culmination of the process of extending functional capabilities at AT&T for accruing pension costs experienced an ironic twist when it surfaced as a public issue during what some historians have termed the "second New Deal". In contrast to the first New Deal where the policy emphasis had been on the stabilization of prices and markets, the focus of the second New Deal shifted about 1935 to concerns about monopoly, social welfare, workers' rights and income stabilization. This latter era witnessed the passage of the Social security Act of 1935, which funded old age benefits for retirees, the Wagner Act, which strengthened the bargaining position of labor, the formation of Temporary Economic Commission, which focused on the problems of industrial concentration and tax rules that created penalties for corporations that failed to recirculate their profits by making either internal investments or payments to shareholders (Leuchtenberg, 1963, pp.132-3,150-2,257-9).

The Social Security Act helped to alleviate the pension burden. Like many other corporations with such plans, the Bell System decided to reduce pensions paid to employees beginning in 1942 by an amount equivalent to the company's contribution to Social security. That year the 1RS implemented Commissioner Beardsley Ruml's recommendation of collecting taxes at the source. This mandated employers to withhold taxes and social security payments on behalf of the federal government. From that point the pension benefit received by retirees would be reduced by that half of the social security payout that had been financed by the company's contribution.

Concerns about a broad range of policies at the Bell System were also high on the agenda of the Federal Communications Commission formed in 1935 to take over among other duties national responsibility for regulating the telephone industry from the ICC. The FCC's investigation, which was financed by a special appropriation from the House of Representatives beginning in 1935, was contentious and led to the issuance of two reports. The first entitled Proposed Report: Telephone Investigation appeared in 1938 and was followed the next year by a final version entitled, Report on the Investigation of the Telephone Industry. In addition, a staff expert, Nestor R. Danielian (1939), also published a book critical of the Bell System based on the findings of this investigation entitled, A. T. & T.: The Story of Industrial Conquest. The main criticisms that emanated from these sources principally related to the adverse impact that portfolio financing policies and the problems of future expense estimation had on the amounts of recoverable cost that the firm could add to its rate base. In addition, these critics also called for a strengthening of employee rights in pension contracts and noted the uneven distribution of benefits between executive cadres and other lower ranks within the corporation (FCC, Exhibit 136, 1936, pp.11-39, 40-65).

The failure of the firm to implement on a more timely basis George Buck's 1926 recommendation of funding the interest on the past service obligation contributed to another controversy with the new federal regulatory agency. Several Bell System subsidiaries elected to fund these balances in 1936, which by then amounted to about US$1 Im annually, by charging them to operating expenses. The delay in initiating this practice meant that its propriety for regulatory purposes would not be decided by business-friendly bureaucrats of the Hoover administration but, instead, by the more critical staff of the New Deal's FCC.

In Docket 5188 that was finally adjudicated in December 1942, the FCC disallowed AT&T's inclusion of interest payments on the unfunded liability carried over from the 1913-27 period. This essentially disqualified these items for reimbursement through the rate structure. The FCC ruled that for regulatory purposes these charges had to be applied directly to surplus and, thus, borne exclusively by shareholders. The FCC was suspicious of any retroactive adjustments that would materially affect the rate base, especially in this case where the firm had waited nine years after it had originally been advised to make this change. The federal agency also objected that the firm's assessment of a four percent interest charge seemed arbitrary and believed that the justification for any such changes should be incorporated formally within the context of its actuarial plan (FCC, 1942, Docket 5188).

The outbreak of World War II disrupted the momentum of further reform. The resolution of many of the problems confronting the Bell System and the rest of US industry experienced a long hiatus. The comprehensive reform of employee pension rights had to await the passage of the Employee Retirement Income Security Act of 1974 (Sass, 1997, Ch.8). The promulgation of accounting standards for actuarially based pension systems, which resolved the question about the treatment of past service costs and myriad other factors affecting pension liabilities did not emerge until 1966 with the issuance of Accounting Principles Board Opinion 8. Yet in spite of the slowness in institutional development, the Bell System had played an important role not only in developing its own internal management system for pension costs but also in identifying the many technical issues that would have to be addressed in the promulgation of formal financial accounting standards.

6. Conclusion

This experience relating to the development of the pension accounting and management system provides insight into two aspects about the nature of corporate learning at AT&T. The first was concerned with the ways that the activities relating to the implementation of several new forms of specialized knowledge gradually became integrated within the firm's organizational structure to facilitate operational efficiency. The second involved the manner in which corporate actions pertaining to pension matters shaped the relationship between the firm and the broader societal context in which it operated.

The primary focus of firm-specific learning relating to pension administration centered in the comptroller's department in the statistics division. This unit initially functioned at a disadvantage because of its lack of competency in actuarial science, a shortcoming that directly contributed to the short-sighted decision to operate the pension system on a pay-as-you-go basis. Initially, the firm tried to manage these activities by relying on accounting information, a type of knowledge that the comptroller's department had deep experience with, in preparing financial reports and undertaking cost analyses. But the historical perspective that permeated accounting measurement practices proved to be inadequate in anticipating rapidly changing future cost patterns associated with rising employment levels and the need to attract and retain qualified employees. The reduction of risk in trying to control contingent outcomes required a measurement system that could provide reliable measures of prospective trends. Although the firm's engineering department had long used probability theory to confront operational uncertainties, the comptroller's department did not acquire such capabilities until after World War I with the expansion of the statistics division under the leadership of Walter Gifford. This led to the extension of the learning base through the recruitment of seasoned statisticians and mathematicians that made possible the type of projective analysis necessary to support an actuarially based system.

While such learning that relied heavily on the application of probability theory helped to reduce the risk that the firm would be unable to anticipate and to fund future liabilities, it had some shortcomings. The problems of inaccurate data compilation marred pension calculus. A more serious difficulty derived from the way that regression analysis was used to project future conditions. The utility of the regression line slopes used for projective purposes depended on the continuance of the past patterns from whence they were derived. The usefulness of such estimates, however, declined markedly when the firm confronted the great economic discontinuity that began with the Crash of 1929. Lacking an effective stochastic mechanism that could accommodate extreme variability and uncertainty, the corporate estimating process only gradually was able to capture in its calculations the profound economic changes that emerged in the 1930s. Fortunately, the phenomenon being measured extended over a very broad time horizon, thus making possible periodic adjustments until the discovery of new trend lines that better reflected contemporary circumstances.

The learning and information costs associated with the development of the actuarially based pension system were doubtless lower at the Bell System than similarly situated unregulated firms because of the operation of the rate base. The costs incurred in acquiring knowledge of measurement practices and associated organizational changes could generally be applied to the rate base and charged back to consumers through the amounts they paid for telephone service. The ability to defray such expenses, doubtless, made the Bell System management more inclined to experiment in the development of new modes of management as well as technology.

AT&T's receptivity to the development of mathematical solutions to complex management challenges such as those inherent in pension administration was reflective of its more general tendency to compensate for the lack of competitive pricing information in the markets it monopolized (Miranti, 2002,2005). The company had sought to overcome the lack of useful market pricing data by placing greater reliance on very detailed statistical and accounting analyses of many business functions. In this way the enterprise resembled a command economy that, lacking the information benefits provided by competitive markets, relied on quantitative models in making resource allocation decisions. Probability theory represented a key tool in this analytical regime, which had been successfully applied by the firm in confronting the uncertainties and risks encountered in earlier initiatives both in network planning and manufacturing quality control. The successful application of probability theory to enhance operational efficiency and reduce risk encouraged the later adaptation in the firm of probabilistic methods that had been used in actuarial management since the eighteenth century.

During this era the Bell System had also learned much about the limitations of pension plans as a means for stabilizing business by building employee loyalty and reducing turnover. While the promise of deferred compensation benefits had strong appeal to workers in an urban-industrial economy who otherwise might be destitute in old age, the plan was imperfect to the extent that only a relative few benefited from its provisions. About a quarter of million employees who qualified for inclusion in the plan during this era did not receive pensions. Part of this was due to the adverse effects of the Great Depression on employment levels. Part of this also had to do with circumstances that bred alienation and dissatisfaction in the work environment. The latter factor eventually motivated new learning in the form of firm-sponsored labor studies such as the one conducted by Professor Elton Mayo and his colleagues from the Harvard Business School at the Hawthorne plant in the 1920s and 1930s.

The process of institutional discovery eventually revealed that the awarding of retirement benefits over the long term became strongly skewed in favor of top and middle management echelons. Ultimately the pension plan at AT&T benefited relatively few potentially qualifying workers, amounting to only about 15 percent of total plan members. Moreover, as the FCC's investigation revealed, the lion's share of benefits was allocated to Group 1 employees - the managers and supervisors who had the skills and knowledge that the firm relied on to maintain its complex network and to facilitate the transition to a new knowledge economy. In this context the incipiency of the federal social security system proved beneficial to the firm because of its indirect subsidization of the company's plan. It also proved to be better suited for providing old-age benefits for the many who would never qualify for corporate pensions that required a 20-year vesting period.

Implicit in the evolution of the pension system was a significant shift involving the transformation of employees, particularly those in the firm's managerial cadre, into financial stakeholders through their collective claim on pension assets. This transition had two aspects. First, under the plan's original conception in 1913, the accrual of employee benefits took the form of an appropriation of the after tax profits of the firm retained in its earned surplus account. Shareholders became compelled to surrender part of their wealth as reflected in the growing reserve for employee benefits in the earned surplus account. But such an innovation had limits. Rapid increase in the potential size of the appropriation because of workforce expansion could not be sustained without creating a crisis among investor groups, particularly shareholders. The second transition emanated from the conversion to an actuarially based system in 1927. In this latter case employees indirectly became major creditors of the firm through the investment of most of the pension trust funds assets in the debt obligations of AT&T and its associated companies.

With each transition in the evolution of a new pension system, the choice of financial assets to fund pension obligations increased employee risk to the firm. It was not until the 1930s that subsidiaries in Illinois demanded investments in US Treasury securities. This may have resulted from the increasing perception of economic risk deriving from a deepening depression. It may have also responded to the increase in the transparency of fund financing practices through regulatory investigations and concerns about how the public subsidization of these payments through the rate base was being applied. It may also have reflected concerns about the liquidity of the market for the firm's debt instruments as the supply of new debentures began to wane because of the general business decline.

The pension learning process at AT&T also suggests that management financing choices were conditioned less by the changes that they wrought on employee risk to the firm and more on the overall level of business risk and range of investment opportunities open to the firm. During periods of rapid growth and increasing investment opportunity, pension funding decisions increased retirees' exposure to the risk of the firm. From 1913 through 1929 the firm indirectly financed its expanding operations in such key activities as automatic switching and long-distance service through the pension system by funding the obligation using corporate equity, demand notes and debentures. This pattern, however, began to reverse with the onset of the Great Depression. Part of this latter change reflected a decrease in internal investment opportunities to fund because of the general recession. Moreover, growing concerns about fiduciary responsibilities in an environment of increasing economic uncertainty encouraged a shift toward low-risk US governmental obligations.

The long-term learning associated with pension management also revealed a growing primacy of the priorities of managerial capitalism over financial capitalism in firm governance. Control of the modern firm remained firmly in the hands of professional managers. They defined the agenda for the advancement of the pension contract and successfully pushed for its implementation. Although their actions were somewhat constrained by consumers through regulatory boards and by investors through the board of directors, the logic of pension creation proved irresistible. Professional managers had two great advantages in this corporate policy debate. They had better knowledge of the firm than outsiders. They also exercised great discretion in controlling the businesses daily operations. In the following decades the pension system increasingly shifted firm resources from consumers and investors to those employees fortunate enough to enjoy its benefits.

Finally, the experience of AT&T and other contemporary firms provided an important lesson in the inherent limitations of private pension plans as vehicles for overcoming the broad societal problems associated with old-age poverty. From the perspective of prospective beneficiaries the probability of enjoying such benefits eventually were revealed as being low. The uncertainty that the actuarially-based pension scheme could not measure at the Bell System was the pattern of exogenous economic change that directly impacted employment levels. This shortcoming diminished the public image of the firm. The bright promise made during the Progressive era of lifetime economic security for loyal service proved unattainable for most because of the onset of the Depression. The social security programs of the New Deal helped to overcome some of the shortcomings of private pension plans. The federal system did so by making its benefits portable between employers. The federal alternative was also less vulnerable financially because it did not depend on the economic performance of any particular employer organization. Moreover, the public system helped to shore up private plans when the latter reduced their overall contribution to pension fund trusts by the amounts they were required to contribute to social security on behalf of their employees.