Divestiture and its implications for innovation and productivity growth in U.S. telecommunications.

Datta, Anusua

1. Introduction

The Schumpeterian Hypothesis (1) suggests the social cost of monopoly may be a price that must be paid for rapid technological advancement. Thus, there is a tradeoff between "static" efficiency losses, reflected in higher prices and lower output due to monopoly, and "dynamic" progressiveness (Nelson and Winter 1982). Over the long run, however, the gains to society from continuing innovation may vastly outweigh those achieved through competitive pricing. This purported tradeoff has been central to the controversy surrounding antitrust policy in the United States, including the recent antitrust case against Microsoft, which had raised serious possibilities of a breakup of the company. Even though the Microsoft case seems headed for a settlement, (2) it has brought to the fore important questions about the antitrust policy and its effects.

In that context, the landmark breakup of American Telephone & Telegraph Company (AT&T), less than two decades ago, represents a classic example. It was an experiment aimed at increasing competition and productive efficiency in the U.S. telecommunications industry. Divestiture of AT&T in 1984 split the industry vertically between local and long-distance operations. While AT&T's local operations were divided among seven regional holding companies, AT&T itself was left with a competitive long-distance service. Divestiture also resulted in the breakup of Bell Labs, the centralized research division of the Bell System into two parts--AT&T Bell Labs, which was a part of AT&T, and Bellcore, jointly owned by the seven regional holding companies. Dismantling the monopoly structure of the U.S. telecom industry provides a unique opportunity to examine the effects of antitrust policy on competition, innovation, and productivity.

Kwoka (1993) and Crandall and Galst (1995) provide preliminary assessments of the impact of deregulation on productivity. These studies use a total factor productivity decomposition approach to study the effects of competition and scale economies on productivity growth. Other productivity studies include Staranczak et al. (1994), a cross-country study including the U.S. telecommunications industry, and Resende (1999), which looks at productivity growth and regulation in local telephones. Almost all these studies show that competition had a positive effect on productivity growth, although divestiture itself seems to have affected productivity negatively.

There are few known studies that statistically test the effects of divestiture on innovation in the telecom industry. Noll (1987) provides a nontechnical examination of the impact of divestiture on AT&T's research and development (R&D) activities, while Noam (1993) gives a broad overview of the impact of divestiture on prices, service, productivity, and research. These studies indicate that AT&T's R&D outlay, in terms of the number of people employed and the volume of investment, has done well from divestiture. Despite a large body of empirical literature on productivity in U.S. telecommunications, studies have generally ignored the potential link between R&D and productivity growth.

The present study uses a simultaneous-equations framework to examine the interrelationship between market structure, R&D, firm size, and productivity before and after the breakup of AT&T. This allows us to capture the possible simultaneity that may exist between R&D and productivity and also to statistically test the effects of divestiture on productivity and R&D activities of AT&T.

Prior to divestiture, AT&T was involved in providing both interexchange toll services and local services. However, since divestiture, AT&T no longer provides local access. This raises the question, is it still meaningful to compare pre- and postdivestiture AT&T? (3) The objective of this article is to analyze the impact of government policy, namely divestiture, on the behavior and performance of AT&T. Government policy action brought about sweeping changes in the telecommunications industry, reducing AT&T from a protected monopoly providing every possible telecom service to one providing only long-distance services. In addition, the market for long-distance services was opened to competition even as local services remained protected. A major argument in favor of deregulation of AT&T was that market forces and competition would improve AT&T's productive efficiency (Noam 1993). At the same time, however, some argued that breaking up AT&T would have dire consequences on research (Noll 1987) as the company was re duced to a fraction of its former self. However, the results from this study show evidence to the contrary. Average productivity for the period 1985-1994 remained below the averages for the previous two decades. On the other hand, despite its smaller size, AT&T's spending on R&D was appreciably higher than before the breakup, which is significant!

The structure of the article is as follows. Section 2 provides a brief background of the deregulation of U.S. telecommunications and reviews the evidence. Section 3 discusses the analytical framework and the measurement of total factor productivity for AT&T. Section 4 presents the regression model. Section 5 provides estimation and results, and section 6 concludes.

2. Deregulation of U.S. Telecommunications

The telephone industry in the United States was changed radically on January 1, 1984, when AT&T was divested of its local operations and left only with competitive long-distance services. This marked the end of a century-old monopoly of the Bell System. AT&T served as a vertically integrated end-to-end monopoly, providing almost all long-distance service and much of the local service in the United States. Its monopoly status was fully protected and regulated by the government. However, development of new microwave technology in the 1960s opened the doors to competition, with new companies like MCI entering the long-distance telephone market. AT&T tried its best to stifle competition using its vast resources and market power. This led to antitrust action and its historic breakup in 1984.

Prior to divestiture, research and development in the Bell System had been centralized at Bell Labs, which had established itself as one of the world's richest sources of private telecommunications research. With the divestiture of AT&T, Bell Labs was broken into AT&T Bell Labs and Bell Communications Research, Inc. (Bellcore). AT&T retained the former, while Bellcore met the needs of the divested Bell operating companies. The divestiture of AT&T raised concerns about the future of research and productivity in the telecom industry. It was feared that fragmentation of the Bell System would result in loss of economies of scale besides greatly limiting the scope and mission of Bell Labs.

Divestiture and Competition

Tables 1 and 2, respectively, show the pre- and postdivestiture market shares (4) of AT&T and other long-distance carriers. AT&T had complete monopoly over long-distance services until 1970, when for the first time the market was opened to competition. However, between 1970 and 1983, AT&T's market share fell only marginally, from 100% to 92% (an 8% decrease), as AT&T retained its incumbency advantages with complete control of the networks. Following divestiture, however, the decline in AT&T's market share was rather dramatic. Between 1984 and 1997, AT&T's share of long-distance market share fell from 90% to less than 45%, a loss of almost 50%. In addition, AT&T faced competition from resellers, whose numbers, according to Noam (1993), rose from 42 in 1982 to 597 in 1991. Following divestiture, the provision of equal access to local networks for all long-distance competitors further aided entry and competition in the long-distance telecommunications market (Taylor and Taylor 1993). A look at four-firm concentr ation ratios, however, shows that the telecom industry was still highly concentrated, with the largest four firms accounting for about 80% of the market in 1997. As MacAvoy (1998) notes, there were substantial reductions in concentration from 1984 to 1997; nevertheless, the Herfindahl Index remained between 0.32 and 0.38, representing a market dominated by three large firms.

Divestiture and Research

As noted in the previous section, divestiture and equal access encouraged entry, which significantly increased competition faced by AT&T. Spence (1986) points out that competition can have two conflicting effects on R&D and efficiency. It lowers profit margins for firms as entry puts a downward pressure on product prices, thereby creating incentives for firms to carry out R&D in order to lower cost. But competition also causes market shares to fall, which reduces scale economies necessary for firms to reap the benefits of R&D. Thus, the cost savings from R&D may be eliminated by revenue losses due to declining market shares and prices.

AT&T's toll service rates fell by as much as 45% in real terms between 1984 and 1991 (Noam 1993). The Federal Communications Commission (1991) was quick to attribute reduced rates to competition. However, other studies (5) have argued that the reduction in prices is largely explained by reduction in carrier access charges paid to local telephone companies by the long-distance carriers. This suggests that lower prices did not necessarily translate into lower profits for AT&T. Moreover, despite losing considerable market share during this period, AT&T remained the market leader, with a 45% market share. The effect of these factors is reflected in the postdivestiture R&D spending of AT&T.

Table 3 presents the figures on AT&T's operating revenues, inflation-adjusted R&D expenditures, (6) and R&D as a percentage of operating revenues (to control for firm size) for the years 1962-1994. A look at these figures, before and after divestiture, shows that R&D has fared well through divestiture. Total R&D spending increased consistently. Most of the increase before divestiture was in keeping with increases in operating revenues. Following divestiture, however, AT&T's operating revenue was reduced to a fraction of its value prior to 1984, but what is striking is that AT&T's total spending on R&D remained equal to or was higher than before. This is reflected in the values for R&D intensity. R&D expenditures as a percentage of operating revenue (7) increased from an average of 2.6% between 1962 and 1983 to an average of about 7% between 1984 and 1994. To further test if the difference in total R&D spending before and after divestiture is statistically significant, an analysis of variance was performed, the results of which are presented in Table 4. The averages for the two periods and a significant F-value (F = 131.9) indicate that the postdivestiture R&D expenditure figures are significantly higher. (8) Employment figures reflect a similar story. Total R&D employment at AT&T Bell Labs rose from a total of 5.3% of total AT&T employees in 1984 to almost 10.3% in 1989 (Noam 1993).

Whether all of the increased R&D spending can be attributed to divestiture alone is questionable, as R&D spending registers a marked increase after 1979. Crandall and Galst (1995) point out that some of it might have taken place anyway, as the period marks the beginning of rapid technological change in the telecommunications sector.

It is interesting to compare research spending of the divested Bell operating companies (BOCs), centralized at Bellcore, with that of AT&T Bell Labs after 1984. According to Federal Communications Commission (FCC) reports, between 1984 and 1994, the regional Bells invested only around 2% of their revenues on R&D, with Bellcore's total budget being almost 50% less than that of AT&T Bell Labs. Crandall and Galst (1995) blame the slow pace of research and development at the local Bell companies partly on public policy, which put restrictions on manufacturing and provision of other services such as cable television.

With the passage of the Telecommunications Act of 1996, the structure of the telephone industry was once again in flux. The new act, in seeking to amend the Communications Act of 1934, removed all barriers to entry in the local exchange and access service. This has opened up the local exchange market, the last bastion of monopoly, to competition for the first time. Moreover, this also allows the regional Bells to enter into the long-distance market that had hitherto been forbidden by the 1982 decree. (9)

3. Analytical Framework

The next step is to statistically test the impact of divestiture and competition on research intensity and productive efficiency of AT&T.

Productivity measures the growth in output not explained by the growth in inputs. Following Denny, Fuss, and Waverman (1981), a typical production function for a telecommunications firm can be written as

f(q,x) = 0 (1)

where q is a vector of outputs (telecom services) and x is a vector of all inputs (capital, labor and materials) used to produce the output. Total factor productivity (TFP) is measured as the ratio of aggregate real output (Q) divided by aggregate real input (X), as TFP = Q/X. The instantaneous rate of growth of TFP is then given by

TFP = Q - X. (2)

Multiple outputs and inputs are aggregated using a Divisia index, (10) which uses changing revenue/cost shares as weights. For data that are available in discrete intervals, a Tornqvist approximation to the continuous Divisia index is used. This index can account for compositional changes in aggregate output and inputs, which makes it very desirable for empirical work.

Traditionally, the excess growth in output over the growth in inputs, alternatively reduction in unit costs, was attributed to technical change. However, Denny, Fuss, and Waverman (1981) have shown that, for large, capital-intensive, and regulated industries such as telecommunications, cost savings can result from realization of scale economies and increased competition as well as technical change.

Before presenting the analytical model used to account for the determinants of total factor productivity growth, measurement of TFP growth for AT&T is discussed.

TFP Measures for AT&T

Christensen, Cummings-Christensen, and Schoech (1981) provide the earliest and the longest index of total factor productivity for AT&T, extending from 1947 to 1979. This is the most widely accepted study for the period, which was used during the antitrust case and also for purposes of regulation later. Crandall (1989) provides a comparable series up to 1987, used in the postdivestiture analysis by Kwoka (1993).

A drawback of previous studies on divestiture is that the postdivestiture period considered is rather short. Kwoka (1993), a study closest to the present one, only includes data up to 1987. However, in order to meaningfully study the effects of divestiture, it was important to incorporate data for the more recent years. To do so, a TFP index for the period 1985-1994 was constructed. (11) This series includes the latest year for which data were available after divestiture. AT&T stopped reporting input and output information to the FCC after 1994. A longer time period is expected to reduce some of the noise associated with the initial years after the breakup.

Detailed data on outputs and inputs were obtained from the public reports of companies to the FCC, contained in the Statistics for Communications Common Carriers (SOCC). The output series was constructed as a Divisia weighted index of four categories of service (interstate toll service), that is, inward only, outward only, private lines, and miscellaneous services. (12) The weights were based on revenue shares for each service. The different output categories were deflated using price indexes for appropriate telephone services, taken from the Bureau of Labor Statistics (BLS). Input series were constructed in a similar fashion. Number of workers and total compensation were used to approximate labor input and labor cost, respectively. (13) Following Christensen, Schoech, and Mietzan (1994), material cost was measured as total operating cost less depreciation and compensation to labor. Material input was deflated by a materials price index (defined in the Christensen, Schoech, and Mietzan study), which is based on those categories of expenditure that relate more closely to materials purchased by the communications industry as opposed to the commonly used gross domestic product (GDP) deflator.

Capital input consists of six asset categories: buildings; general support equipment; office equipment and vehicles, etc.; central office equipment (including operator systems); transmission equipment; and information origination/termination equipment and cable and wire. The quantity of capital stock is calculated using the perpetual inventory method for each class of assets, similar to Christensen, Schoech, and Mietzan (1994). Depreciation rates and price indexes were obtained from the Bureau of Economic Analysis and the BLS.

Figure 1 presents the percentage growth in TFP for AT&T for the period 1962-1994, which includes figures from earlier studies by Christensen, Cummings-Christensen, and Schoech (1981) and Crandall (1989). The figure shows substantial variation in growth from year to year. (14) It also shows a sharp decline in TFP growth in the years immediately following divestiture. However, TFP starts picking up after 1989 and remains in the positive range after that time period. The initial decline in productivity can be largely attributed to sharp decreases in output growth. While output growth stabilizes after 1989, some of the TFP gains after that resulted from big decreases in the size of capital stock. This could indicate an effort on the part of AT&T to divest itself of inefficient or excess capital, a legacy from the previous era.

Table 5 presents the average growth in TFP for AT&T in the last three decades. The figures indicate a slowdown in TFP growth after 1983. When 1984 is included in the calculations for the postdivestiture years, the average further decreases to 2.63. As the evidence suggests, decline in productivity after divestiture could be a reflection of many factors, such as restructuring, overstaffing, excess capital stock, and a slowdown in output growth as competition increased.

4. Model

Theoretical Framework

The purpose here is to determine the sources of productivity growth in telecommunications. Denny, Fuss, and Waverman (1981) have demonstrated that TFP can be decomposed into scale effects, market competition effects, and the pure technical change as

TFP = [1-[summation over (Q)][[epsilon].sub.cq]] [Q.sub.C] + [[Q.sub.P] - [Q.sub.C]] + [B]. (4)

In the above expression, [summation over (q)] [[epsilon].sub.cq] refers to cost elasticity with respect to output and ([summation over (q)] [[epsilon].sub.cq])Qc measures economies of scale. [[Q.sub.P] - [Q.sub.c] measures the deviation in actual output from output under marginal-cost pricing or competition, denoted by [Q.sub.P]. Under perfect competition, the two measures are equal. [B] measures the residual shift in the cost function not accounted for by the above factors and is generally said to capture pure technical change.

A limitation of the above framework is that it does not provide a detailed characterization of technology. It fails to explicitly model the effect of endogenous R&D on productivity growth by treating technical change as an exogenous process [B]. Various cross-section studies, for example, Griliches (1980), Terleckyj (1980), and Scherer (1983), have shown that a crucial positive link exists between R&D investment and productivity, as also exists between market concentration and R&D. However, none of the recent studies in telecommunications, including Crandall and Galst (1995), Kwoka (1993), Staranczak et al. (1994), or Resende (1999), accounts for these relationships.

The Regression Model

In this article, the traditional TFP decomposition, Equation (4), is extended to explicitly include R&D effects, Further, to test the Schumpeterian notion that market concentration promotes innovation, we specify a relationship among R&D, market structure, and productivity.

Due to the absence of exact information on cost elasticities, marginal cost, and competition, appropriate proxies are used for empirical estimation. Following Denny, Fuss, and Waverman (1981) and Kwoka (1993), these proxies are defined as follows.

(a) Divestiture. The effects of divestiture on productivity growth and R&D investment are captured by the dummy Divest that takes the value 1 for the period starting 1984.

(b) Scale effects. (1 - [summation over (q)] [[epsilon].sub.cq])Q, measures the degree to which output growth lowers unit costs. Both TFP and output growth Q are measured in logs; therefore, the coefficient of Q measures scale elasticity. For a firm that enjoys increasing returns to scale, the estimated coefficient of Q should be positive. The output series is taken from the measure of the output growth variable used for TFP construction by Christensen, Cummings-Christensen, and Schoech (1981), Crandall (1989), (15) and Datta (1999).

(c) Competition. In theory, the higher the competition, the closer a firm's output is to marginal cost output, which in turn increases efficiency and therefore productivity. The gap between actual output and output under marginal cost pricing is therefore a measure of the extent of market competition. For AT&T, this is measured by its market share, MS. (16) Because market share of AT&T has been declining, a positive impact of competition on TFP would be captured by a negative MS coefficient.

In addition, other variables included in the present analysis are the following.

(d) R&D. R&D, which represents R&D intensity, is measured as the ratio of total R&D expenditure (17) to total operating revenues or sales. The data on R&D expenditure and total operating revenues are obtained from the company annual reports.

(e) Capital Intensity. KS represents capital intensity, measured as the ratio of total telephone plant (gross) to total operating revenues. KS is included to test for the presence of the Averch-Johnson (A-J) effect. In a recent study, Nadiri and Nandi (1997) note that the telecom industry demonstrates a considerable overinvestment in physical capital. The authors attribute overcapitalization to the A-J effect of the rate of return regulation, which was in place until 1989. If the A-J effect exists, we would expect the coefficient of KS to be negative.

(f) Total Factor Productivity. TFP measures the growth in total factor productivity. The series used for the study is taken from Christensen, Cummings-Christensen, and Schoech (1981) for the period 1947-1979 and Crandall (1989) for the period 1980-1984. TFP values for the remaining period (1985-1994) are taken from Datta (1999).

The system of equations to be estimated is expressed as

[TFP.sub.t] = [a.sub.10] + [a.sub.11][Divest.sub.t] + [a.sub.12][MS.sub.t] + [a.sub.13]Q + [a.sub.14][R&D.sub.t] + [a.sub.15][KS.sub.t] + [[epsilon].sub.1t], (5)

[R&D.sub.t] = [a.sub.20] + [a.sub.11][Divest.sub.t] + [a.sub.22][MS.sub.t] + [a.sub.23][TFP.sub.t] + [[epsilon].sub.1t]. (6)

Equations (5) and (6) are jointly estimated using the three-stage least squares technique to account for possible simultaneity between R&D and productivity growth. (18) Additionally, this method takes care of cross-equation correlation between residuals and is considered superior to other simultaneous-equations methods. The above specification represents the basic model. Variants of the above model are estimated, including slope dummies, to account for divestiture effects due to some to variables of interest and the role of profits in R&D decisions.

5. Estimation and Results

Equations (5) and (6) are first estimated with an intercept dummy to account for the overall effect of divestiture, (19) using a three-stage least squares method. The results are reported in Table 6. The system has a weighted MSE of 1.06, 54 d.f., and a system weighted [R.sup.2] of 0.97.

The adjusted [R.sup.2] for the TFP equation is 0.44, which is comparable with that of Kwoka (1993). All the variables of interest are significant at the 5% level. The coefficient of the Divest dummy is strongly negative, indicating a decline in productivity following divestiture. The declining trend in productivity found three years after divestiture (Kwoka 1993) seems to persist beyond the initial years. The MS coefficient is negative and significant. A negative coefficient suggests that increased competition, measured by a decline in market share, has led to improvements in productivity. The coefficient of the scale variable (Q) is greater than one and is highly significant, underscoring its importance in explaining productivity growth.

The R&D-TFP link, one of the focuses of this study, is significant and positive. This, combined with positive scale economies, suggests that some of the decreases in marginal cost for the firm can probably be traced to technological improvements. Finally, the KS variable, which tests for the Averch-Johnson effect, is significant and negative, supporting the presence of overcapitalization, which is mostly true for the predivestiture period. The negative sign suggests that a decrease in the size of the capital stock leads to productivity gains.

The R&D equation has an adjusted [R.sup.2] value of 0.98, suggesting a good fit. Further, the t-statistics for all the variables of interest are significant at the 1% and 5% levels. The coefficient of the Divest dummy is significantly positive. The estimates suggest that 4.72% of the increase in R&D investment can be attributed to divestiture effects. This is further reinforced by a significant and negative MS coefficient, which shows that increased competition, measured by a decline in market share, had a positive effect on AT&T's R&D spending. A possible explanation for such behavior may lie in the nature of the market in which AT&T now operates. A competitive oligopoly has two positive effects on R&D: (i) competition forces a firm to innovate to retain its market share and (ii) it affords a firm sufficient market power to appropriate the returns from innovation. Finally, the correlation between TFP and R&D is positive but not significantly different from zero.

Next, the overall divestiture effect is decomposed into its various sources by adding slope dummies for market share, R&D, TEP, and other variables to the base model. This allows us to see the differential effects of these variables on productivity growth and R&D before and after divestiture. In addition, Profit is included as an explanatory variable in the R&D equation in order to examine the relationship between R&D spending and profitability before and after divestiture. For instance, it is contended that divestiture has forced AT&T to move away from basic research to more profit-oriented research (Noll 1987). Several variants of the model are estimated, the results of which are reported in Table 7.

Inclusion of slope dummies significantly improves the [R.sup.2] values for the TFP equation, from 0.44 to 0.68. The results show that the pure Divest effect on TFP is negative. However, as other explanatory variables are added to the model, the net effect of divestiture becomes insignificant. The MS coefficient remains significant and negative, reinforcing the positive link between competition and productivity. When a dummy variable for output growth (Q) is added, scale effects for the predivestiture years remain positive but are now insignificant. The coefficient of the Q dummy, however, is highly significant and positive. The magnitude of the coefficient reveals considerable scale economies in the postdivestiture period. Nadiri and Nandi (1997) found similar evidence of increased scale economies in toll services after divestiture.

The relationship between R&D and TFP again shows interesting results. It is negative and insignificant/significant in the predivestiture period but strongly positive and significant in the period after divestiture. This result supports the positive R&D-productivity link found in the studies on nonregulated industries by Terleckyj (1980) and Griliches (1980). Improvements in technology in long-distance telecom can partly explain the increased efficiency in the production of toll service. However, this could also represent evidence of greater accountability in R&D investment as AT&T moves away from basic research, the benefits of which are not directly appropriable, to more proprietary research (Noll 1987).

Finally, the coefficient of the KS dummy is negative and significant, indicating the possibility of productivity gains for AT&T through the divestiture of excess capital. As noted in an earlier section, the productivity gains after 1989 were largely the result of a decrease in the size of the capital stock.

The overall explanatory power of the R&D equation remains at 0.98. The results show that the Divest dummy remains significantly positive. The coefficient for MS is also negative and significant, reinforcing the positive effect of competition on R&D. However, when a profit dummy is added, the MS coefficient loses significance. No significant relationship exists between Profu (20) and R&D in the predivestiture years. However, the postdivestiture Profit dummy is positive and significant, suggesting that research at AT&T became more profit-driven after divestiture.

The relationship between TFP and R&D reveals some interesting facts. The coefficient of TFP is negative before divestiture but becomes positive thereafter. However, when one considers the net effect (measured as the difference in coefficient value between TFP and TEP dummy), it is not different from zero. Combined with the results from the TFP equation, this suggests that R&D is important in explaining productivity growth, but the reverse is not necessarily true.

6.Conclusions

This study has sought to determine the impact of policy changes that led to the breakup of the monopoly structure of U.S. telecommunications on productivity and research activities of AT&T. A simultaneous-equations model is used to examine the relationship between productivity, R&D, market share, and other variables. The results show that, contrary to popular belief, transition from a protected monopoly to a competitive oligopoly led AT&T to increase and not decrease its R&D. The effect of competition on productivity is strongly positive. Interestingly, scale economies remain a significant factor in productivity growth even after divestiture, suggesting thereby that the telecommunications market continues to favor large firms. The net effect of divestiture on productivity growth remains disruptive.

The study finds an interesting relationship between R&D and TFP. While it is insignificant prior to 1984, it becomes significant and positive in the postdivestiture period. Technological change in long-distance telecommunications can partly explain the increased efficiency in the production of toll service. However, this could also be the result of greater accountability in R&D spending after divestiture as AT&T moves away from basic research, the benefits of which are not directly appropriable, to more proprietary research. The profit-R&D relationship further supports this view. While the link is insignificant before divestiture, it is strong and positive postdivestiture. Finally, evidence supports overinvestment in capital stock, especially in the predivestiture years.

There are a few important issues that require further study that have not been addressed in this article. As data become available, it would be of interest to examine the postdivestiture R&D activities and productivity growth of the other long-distance carriers, such as MCI, Sprint, and the local Bell operating companies.

[FIGURE 1 OMITTED]

Table 1

AT&T's Share of Toll Revenues before Divestiture

Year AT&T's Market Share

1962 100
1963 100
1964 100
1965 100
1966 100
1967 100
1968 100
1969 100
1970 99.4
1971 98.7
1972 98.8
1973 98.9
1974 98.9
1975 98.8
1976 98.5
1977 98.5
1978 98.6
1979 98.1
1980 97.6
1981 96.5
1982 94.6
1983 92.1

Source: Kwoka (1993).

Table 2

Post Divestiture Market Shares of Long Distance Carriers

 Other Long
Year AT&T MCI SPRINT WorldCom Distance Carriers CR4

1984 90.1% 4.5% 2.7% 2.6% 0.97
1985 86.3 5.5 2.6 5.6 0.94
1986 81.9 7.6 4.3 6.3 0.94
1987 78.6 8.8 5.8 6.8 0.93
1988 74.6 10.3 7.2 8.0 0.92
1989 67.5 12.1 8.4 0.2% 11.8 0.88
1990 65.0 14.2 9.7 0.3 10.8 0.89
1991 63.2 15.2 9.9 0.5 11.3 0.89
1992 60.8 16.7 9.7 1.4 11.5 0.89
1993 58.1 17.8 10.0 1.9 12.3 0.88
1994 55.2 17.4 10.1 3.3 14.0 0.86
1995 51.8 19.7 9.8 4.9 13.8 0.86
1996 47.9 20.0 9.7 5.5 17.0 0.83
1997 44.5 19.4 9.7 6.7 19.8 0.80

Source: FCC Reports.

Table 3

R&D Expenditures before and after Divestiture

 Totoal Operating R&D as a Percentage of
Year Revenues ($000) Total R&D ($000) Operating Revenues (%)

1962 8,980,208 247,484 2.76
1963 9,568,961 250,843 2.62
1964 10,305,993 277,071 2.69
1965 11,061,783 313,256 2.83
1966 12,138,265 333,601 2.75
1967 13,009,000 364,627 2.80
1968 14,100,000 398,592 2.83
1969 15,683,000 445,684 2.84
1970 16,954,000 525,058 3.10
1971 18,510,000 587,763 3.18
1972 20,904,000 657,062 3.14
1973 23,527,000 542,529 2.31
1974 26,174,000 584,083 2.23
1975 28,957,000 619,679 2.14
1976 32,815,000 641,067 1.95
1977 36,431,000 718,399 1.97
1978 40,993,000 837,757 2.04
1979 45,408,000 977,500 2.15
1980 50,658,000 1,189,500 2.35
1981 58,065,000 1,516,900 2.61
1982 65,093,000 1,790,400 2.75
1983 69,403,000 2,221,100 3.20
1984 33,187,000 2,368,000 7.14
1985 34,417,000 2,370,000 6.89
1986 34,087,000 2,278,000 6.68
1987 33,768,000 2,453,000 7.26
1988 35,276,000 2,572,000 7.30
1989 36,112,000 2,652,000 7.34
1990 38,263,000 2,935,000 7.67
1991 38,805,000 3,114,000 8.02
1992 39,580,000 2,911,000 7.35
1993 41,623,000 3,111,000 7.47
1994 43,425,000 3,110,000 7.16

Source: Annual Reports of AT&T and Western Electric Company, Moody's
Public Utility Manual.

AT&T's total R&D expenditure before divestiture is the sum of Research &
Systems Engineering (R&SE) performed by AT&T and development expenditue
of Western Electric Company (WECO). Following divestiture WECO was
consolidated with AT&T.

Table 4

Comparing AT&T's R&D Spending before and after 1984

Total R&D Expenditure Average Variance

1962-1983 729,088.9 2.72E+11
1984-1994 2,715,182.0 1.08E+11

ANOVA

 Source SS d.f. MS F
 Model 2.89E+13 1 2.89E+13 131.9 (a) ***
 Error 6.80E+12 31 2.19E+11
 Total 3.57E+13 32

Total R&D Expenditure

1962-1983
1984-1994

ANOVA

 Source p-Value
 Model 0.000
 Error
 Total

(a)***Significant at the 1% level.

Table 5

TFP Growth Comparison, 1962-1994

Years 1962-1972 1973-1983 1985-1994

TFP 2.98 3.43 2.76

Table 6

Three-Stage Least Squares Estimates with Intercept Dummy (a,b)

 TFP Dependent Variable R&D

Constant 13.47 3.95
 (1.29) (4.67) ***
Divest -32.98 4.72
 (-2.76) *** (18.24) ***
MS -0.15 -0.02
 (-2.09) ** (-2.34) ***
Q 1.14 --
 (4.95) ***
TFP -- 0.002
 (0.07)
R&D 4.56 --
 (2.11) **
KS -4.11 --
 (-2.11) **
[R.sup.2] (adj.) 0.44 0.98

(a)Figures in parentheses represent t-statistics.

(b)** Significant at the 5% level.

*** Significant at the 1% level.

Table 7

Three-Stage Least Squares Estimates with Intercept and Slope Dummies
(a,b)

 TFP (1) R&D (1) TFP (2) R&D (2)

Intercept 22.24 4.72 27.08 4.57
 (2.19) ** (4.64) *** (3.13) *** (4.46) ***
Divest -59.78 4.31 -23.61 4.36
 (-3.32) *** (12.65) *** (-1.35) (12.67) ***
MS -0.16 -0.02 -0.17 -0.02
 (-1.97) * (-2.56) ** (-2.43) ** (-2.46) **
Profit -- -1.66 -- -0.77
 (-0.92) (0.43)
Profit dummy -- -- -- --

Q 0.42 -- 0.21 --
 (1.40) (0.80)
Q dummy 1.11 -- 0.95 --
 (2.68) ** (2.57) **
R&D -0.35 -- -1.27 --
 (-0.17) (-1.73)
R&D dummy 7.11 -- 7.17 --
 (3.14) *** (3.53) ***
TFP -- -0.06 -- -0.07
 (-1.94) * (-2.14) **
TFP dummy -- 0.06 -- 0.07
 (2.23) ** (2.21) **
KS -1.86 -- -0.74 --
 (-0.96) (-0.44)
KS dummy -- -- -35.14 --
 (-2.99) ***
[R.sup.2] (adj.) 0.59 0.98 0.68 0.98

 TFP (3) R&D (3)

Intercept 28.39 2.25
 (3.38) *** (1.57)
Divest -21.12 4.00
 (-1.26) (10.04) ***
MS -0.17 0.005
 (-2.43) ** (0.33)
Profit -- -3.28
 (-1.42)
Profit dummy -- 13.04
 (2.14) **
Q 0.17 --
 (0.68)
Q dummy 0.93 --
 (2.63) **
R&D -1.28 --
 (-2.16) **
R&D dummy 7.53 --
 (3.75) ***
TFP -- -0.08
 (-2.48) **
TFP dummy -- 0.07
 (2.26) **
KS -0.76 --
 (-0.47)
KS dummy -38.84 --
 (-3.41) ***
[R.sup.2] (adj.) 0.68 0.98

(a)Figures in parentheses represent t-statistics.

(b)* Significant at the 10% level.

** Significant at the 5% level.

*** Significant at the 1% level.

Received July 2000; accepted June 2002.

(1.) Schumpeter (1950).

(2.) Four years after the Department of Justice filed its antitrust case against Microsoft, the federal government, along with half of the 18 states, has decided to settle the case against the company. The final course of action, however, remains to be determined, as the remaining nine states continue their battle in court.

(3.) One way to compare the two periods could be to compare only the long-distance portion of AT&T's business before divestiture with AT&T postdivestiture. However, while separate output measurements for local and toll services are available for the period before divestiture, there is no way to distinguish between inputs use and R&D spending for the two.

(4.) Market shares are based on operating revenues of long-distance carriers.

(5.) See Taylor and Taylor (1993) and Hausman, Tardiff, and Belinfante (1993) for details.

(6.) AT&T's total R&D spending before divestiture includes Research & Systems Engineering performed at AT&T (which covers both local and long-distance operations) and R&D expenditure of western Electric Company (WECO). R&D spending after 1984 includes research spending at AT&T and WECO (which was consolidated with AT&T after divestiture) but excludes research spending by the divested local Bells.

(7.) Access charges were included in operating revenues before divestiture as part of the BOC revenues. Following divestiture, access became equal for all common carriers and the amount of access charges paid by AT&T decreased considerably. After 1984, AT&T reported access charges as part of operating expenses and not operating revenues, as it viewed access charges as a pass-through item, which is paid as an expense to the BOCs. Excluding access charges from revenue was expected to provide a more meaningful picture of AT&T's revenue performance after divestiture (Noll 1987).

(8.) A similar ANOVA test, comparing R&D intensity instead of total R&D spending, yielded a significant F-value of 1080.91.

(9.) Schiesel (2001) notes that the 1996 law has largely benefited the local Bells, as they (e.g., Verizon Communications, previously Bell Atlantic) have used their incumbency advantages in local markets to capture market share in the long-distance telephone market. Yet there is little evidence of the long-distance giants making similar inroads into local markets.

(10.) There are many possible approaches to measuring TFP: estimation of a translog cost/production function; goal-programming, an operations-research methodology (Charnes, Cooper, and Sueyoshi 1988); and the Divisia-index approach. However, unlike the Divisia approach, the former two are sensitive to the precise specification of the production function or the efficiency frontier and perform poorly in in-sample predictions for industries with rapidly changing technology. For details on the desirable properties of the Divisia-Tornqvist index, see Diewert (1981).

(11.) The methodology used for measuring TFP closely follows Christensen, Cummings-Christensen, and Schoech (1981). All the three TFP measures used in this study are based on the Divisia-index approach to measuring TFP developed by Jorgensen and Griliches (1967), discussed in footnote 7. TFP for the predivestiture years includes both local and toll operations data, as no separation exists between the two. Differences with the previous period pertain to changes in the composition of outputs/inputs due to divestiture. Further details on data construction can be found in Datta (1999).

(12.) Output measures for the predivestiture period include both local and toll operations.

(13.) Further break-up of data by type of workers, e.g., skilled, unskilled, was not available for the postdivestiture years to construct a quality-adjusted index of labor input.

(14.) This might reflect the effect of business cycles, which can cause short-term fluctuations in TFP due to fluctuations in capacity utilization.

(15.) The Crandall (1989) data on total factor productivity growth and rate of growth of output for the period 1980-1984 are obtained from Kwoka (1993).

(16.) Market share is measured as AT&T's share of toll revenues of long-distance companies. Data on market shares for the period before 1984 are taken from Kwoka (1993); for the remaining period the numbers are obtained from FCC (1995).

(17.) For the measurement of R&D expenditures, see footnote 3.

(18.) Studies indicate that, while future output and productivity depend on present and past R&D, R&D investments are in turn affected by the level of output, past profits, and productivity, which calls for the use of simultaneous-equations models (Kamien and Schwartz 1982; Griliches 1998). The problem of simultaneity, unfortunately, has very often been ignored in the literature.

(19.) As noted before, TFP growth can be affected by short-term fluctuations such as business cycles. In order to control for business cycles, GDP growth was included in one version of Equation (5). The GDP variable was insignificant; however, results also revealed collinearity between the GDP and output growth variables. Griliches (1998, ch. 2) points out that short-term fluctuations in TFP are captured by the scale variable (output growth) through fluctuations in capacity utilization. This suggests that the output growth variable in the model largely captures the business cycle effect.

(20.) Profit is measured by the rate of profit, figures for which were obtained from annual reports.

References

Charnes, Abraham, William W. Cooper, and Toshiyuki Sueyoshi. 1988. A goal programming/constrained regression review of the bell system breakup. Management Science 34:1-26.

Christensen, Laurits R., Dianne Cummings-Christensen, and Philip E. Schoech. l98l. Total factor productivity in the Bell sys tern 1947-1979. Madison. WI: Laurits R. Christensen Associates (Report).

Christensen, Laurits R., Philip E. Schoech, and Mark E. Mietzan. 1994. Productivity of the local telephone operating companies. Madison, WI: Laurits R. Christensen Associate (Report).

Crandall, Robert W. 1989. Efficiency gains from divestiture. Washington, DC: Brookings (Mimeo).

Crandall, Robert W., and Jonathan Galst. 1995. Productivity growth in the telephone industry since 1984. in The service productivity and quality challenge, edited by P. T. Harker. Dodrecht, Netherlands: Kluwer Academic Publishers, pp. 391-405.

Datta, Anusua. 1999. Three essays in R&D, productivity and growth. Ph.D. thesis, University of Wisconsin, Milwaukee.

Denny, Michael, Melvyn Fuss, and Leonard Waverman. 1981. The measurement and interpretation of total factor productivity in regulated industries. In Productivity in regulated industries, edited by Thomas G. Cowing and Rodney E. Stevenson. New York: Academic Press, pp. 179-212.

Diewert, Erwin W. 1981. The theory of total factor productivity in regulated industries. In Productivity in regulated industries, edited by Thomas G. Cowing and Rodney E. Stevenson. New York: Academic Press, pp. 17-42.

Federal Communications Commission. 1991. Expanded interconnection with local telephone company facilities. CC Docket 91-141.

Federal Communications Commission, Common Carrier Bureau. 1995. Long distance market shares: First quarter 1995. Washington, DC: Federal Communications Commission (Report).

Griliches, Zvi. 1980. Returns to research and development expenditures in the private sector. In New developments in productivity measurement and analysis, edited by John W. Kendrick and Beatrice N. Vaccara. Chicago: The University of Chicago Press, pp. 419-54.

Griliches, Zvi. 1998. Issues in assessing the contribution of research and development to productivity growth. In R&D and productivity: The econometric evidence, edited by Zvi Griliches. Chicago: The University of Chicago Press, pp. 17-45.

Hausman, Jerry, Timothy Tardiff, and Alexander Belinfante. 1993. The effects of the breakup of AT&T on telephone penetration in the United States. American Economic Review 83:178-84.

Jorgensen, Dale W., and Zvi Griliches. 1967. The explanation of productivity change. Review of Economic Studies 34:249-83.

Kamien, Morton I., and Nancy L. Schwartz. 1982. Market structure and innovation. New York: Cambridge University Press.

Kwoka, John E. 1993. The effects of divestiture, privatization and competition on productivity in U.S. and U.K. telecommunications. Review of Industrial Organization 8:49-61.

MacAvoy, Paul W. 1998. Testing for competitiveness of markets for long distance telephone services: Competition finally? Review of Industrial Organization 13:295-319.

Nadiri, M. Ishaq, and Banani Nandi. 1997. The changing structure of cost and demand for the U.S. telecommunication industry. Information Economics and Policy 9:319-47.

Nelson, Richard R., and Sidney G. Winter. 1982. The Schumpeterian tradeoff revisited. American Economic Review 1:113-32.

Noam, Eli M. 1993. Assessing the impacts of divestiture and deregulation in telecommunications. Southern Economic Journal 59:438-49.

Noll, Michael A. 1987. Bell system R&D activities: The impact of divestiture. Telecommunications Policy 11:161-78.

Resende, Marcelo. 1999. Productivity growth and regulation in U.S. local telephony. Information Economics and Policy 11:23-44.

Scheisel, Seth. 2001. Sitting pretty: How baby Bells may conquer the world. New York Times, 22 April, section 3, p. 1.

Scherer, Frederic M. 1983. Concentration, R&D and productivity. Southern Economic Journal 50:221-5.

Schumpeter, J. A. 1950. Capitalism, socialism and democracy. 3rd edition. New York: Harper.

Spence, Michael. 1986. Cost reduction, competition and industry performance. In New developments in the analysis of market structure, edited by Joseph E. Stiglitz and G. Franklin Mathewson. Cambridge, MA: The MIT Press, pp. 475-515.

Staranczak, Genio A., Edgardo R. Sepulveda, Peter A. Dilworth, and Shafi A. Shaikh, 1994. Industry structure, productivity and international competitiveness: The case of telecommunications. Information Economics and Policy 6:121-42.

Taylor, William E., and Lester D. Taylor. 1993. Postdivestiture long-distance competition in the United States. American Economic Review 83:185-90.

Terleckyj, Nestor E. 1980. Direct and indirect effects of industrial research and development on the productivity growth of industries. In New developments in productivity measurement and analysis, edited by John W. Kendrick and Beatrice N. Vaccara. Chicago: The University of Chicago Press, pp. 359-77.

Anusua Datta *

* School of Business Administration (Economics), Philadelphia University, Philadelphia, PA 19144-5497, USA; E-mail [email protected].

This article is a part of my doctoral dissertation research. I wish to thank James Peoples for his helpful advice and Mark Mietzen, Robert Crandall, and two anonymous referees for their valuable suggestions. I am especially grateful to Mark Mietzen for making Christensen Associates' predivestiture TEP data on AT&T available to me. The usual caveats apply.