This Selected Issues paper analyzes the specialty about the cyclical expansion in the United Kingdom. The paper reviews the labor market data, wage inequality, and implications for inequality. The study examines the contributions of information and communications technologies investment to output and labor productivity growth in the United Kingdom. The paper provides a qualitative assessment of the impact of the appreciation of sterling on U.K. foreign trade, and its effects on competitiveness, and the sectoral composition of production.

Abstract

This Selected Issues paper analyzes the specialty about the cyclical expansion in the United Kingdom. The paper reviews the labor market data, wage inequality, and implications for inequality. The study examines the contributions of information and communications technologies investment to output and labor productivity growth in the United Kingdom. The paper provides a qualitative assessment of the impact of the appreciation of sterling on U.K. foreign trade, and its effects on competitiveness, and the sectoral composition of production.

II. The “New Economy” in the United Kingdom1

A. Introduction

1. Of the three elements most frequently identified with the “new economy”—sustained output growth, subdued inflation, and an acceleration of labor productivity—only the third is conspicuously absent in the United Kingdom. Labor productivity growth has fallen in the latter part of the 1990s compared to the first half. This fact is puzzling when one observes that investment in Information and Communications Technologies (ICT) is relatively high (almost on par with the United States) when measured comparably.2

2. This paper examines this conundrum by looking closely at the contributions of ICT investment to output and labor productivity growth in the United Kingdom. Several studies have attributed the absence of evidence of the new economy outside the United States to mismeasured ICT capital and associated output, and particularly, to deflators that fail to account for quality improvements in ICT goods. To be able to distinguish the impact of different deflators, both U.S. and U.K. price deflators for ICT investment and output are used to decompose labor productivity growth into the contributions from capital deepening (that proportion of labor productivity growth related to the growth of the capital stock available to each worker) and total factor productivity (that proportion measured as a residual component and not attributable to the capital or labor inputs, usually associated with better techniques or organizational aspects of production).3 Each of these two components, in turn, is disaggregated into elements representing the ICT sector (computers and office equipment, software, and telecommunications) and the rest of the economy. Though the data are not completely comparable, the analysis broadly suggests that the cause of the relatively low productivity growth in the United Kingdom does not lie with insufficient use or production of ICT goods, as these elements of productivity growth are similar to those of the United States, but with sub par total factor productivity growth in the rest of the economy.

3. It should be acknowledged at the onset of this examination that the use of U.S. deflators, while becoming the standard for deflating nominal ICT investment and output, is controversial. The use of the U.S. deflators for the United Kingdom is not necessarily the “correct” method—as it depends on a number of assumptions—but simply one that shows the difference in the magnitude of the results when a more dramatic decline in prices is assumed. The adoption of U.S.-type deflators will depend upon the costs and benefits of maintaining the different methodology and its appropriateness in the U.K. context.

4. Though new research is now providing more insight into the role of ICT in the U.K. economy, most commentators had subscribed to the following explanations for the lack of evidence of the “new economy” in the United Kingdom.

  • Much of the higher productivity in the United States has been generated by the actual production of computers and semiconductors. The computer production industry in the United Kingdom amounts to only half the size that it is in the United States and thus the United Kingdom cannot expect to have the type of acceleration in labor productivity that the United States has experienced.

  • The recent dramatic fall in the unemployment rate in the United Kingdom is thought to be, in large part, due to a fall in the unemployment rate of unskilled workers. Adding unskilled workers to the labor pool (relative to skilled ones) could lower labor productivity for a time.

  • Possible data mismeasurement may have introduced a downward bias in labor productivity. Output could be “mismeasured” if deflators for ICT output do not adequately account for quality improvements and thus underestimate price declines. And the output of the service sector, where adoption of ICT has been very heavy (e.g., the financial services industry), is also thought to be under recorded because the output of this sector is difficult to define and measure. For these two reasons, recorded labor productivity may not pick up the “new economy” effects that are present.

5. Yet, even allowing for measurement problems, there is some circumstantial evidence that productivity in the ICT sector is on the upswing and thus these answers are not fully satisfying. First, while the production of computers may be a relatively small part of the economy compared to the United States, the total ICT sector in the United Kingdom, including the production of telecommunications equipment and software, is not inconsequential. In terms of final output, the broad categories representing computers and office equipment, telecommunications, and software produced some 4¼ percent of nominal GDP. Second, analysis from the New Earnings Survey shows that labor composition effects cannot account for the subdued productivity growth—the employment share of high-skill workers actually increased during this expansion rather than the reverse as is typical in business cycle expansions.4 Third, although data problems are present in the measurement of productivity, there is evidence of a pickup in productivity in some of the manufacturing industries most closely related to ICT production (optical equipment, engineering) and, as well, in business services, another heavy user of ICT equipment. Though the potential mismeasurement of output in the services industry also exists in the United States and thus cannot by itself account for the cross country difference in productivity, the fact that the financial sector (one of the most heavy users of ICT) represents a larger share of United Kingdom GDP than the United States suggests that the two countries may be less disparate than the numbers would imply.

6. Precise estimates of the contribution of ICT to output growth and labor productivity are elusive, but the results provided below are an attempt to gain a more formal understanding. The paper uses the work of Oliner and Sichel (2000) as the basis for the analysis and applies their techniques, with some adaptation, to the U.K. data. To examine whether mismeasurement of quality improvements of ICT goods plays a large role in the analysis, U.S. deflators, which show larger quality improvements than the U.K. deflators, are used. The basic results of the analysis are as follows.

  • Using U.S. output and investment deflators as proxies for the decline in the prices of ICT goods rather than the U.K. deflators, adds nearly a percentage point to average output growth between 1997 and 1999. When chain-weighting is additionally employed (as done in the United States) to permit the changes in the quantities of goods that would normally accompany the large price declines in ICT goods, output growth in the United Kingdom increases by about ½ percent between 1997 and 1999 rather than a full percentage point.

  • In terms of labor productivity, the contribution of ICT to capital deepening amounts to about one-quarter of the labor productivity growth in the 1997–1999 period. That is, output per worker has increased by increasing the amount of ICT capital per worker. Moreover the contribution of ICT to capital deepening has increased three-fold compared to the early 1990s. Total factor productivity (TFP), the remaining component of labor productivity, falls in the latter 1990s. TFP is where the network benefits and spillovers of ICT are thought to reside. As yet, then, such benefits are not evident in the United Kingdom.

  • When an analogous decomposition of TFP is conducted for the United Kingdom as done for the United States by Oliner and Sichel, the growth of TFP for the computer and office equipment industry is more or less the same as the computer hardware industry in the United States. One of the most striking results of Oliner and Sichel’s paper is the high TFP growth evident within the semiconductor industry. However, since there is no data isolating semiconductor output in the U.K. official statistics, it is difficult to disentangle the possible contributions of this industry to U.K. TFP growth. In this study, any TFP gains for semiconductor production would reside in the residual “other” TFP growth category. Though because the semiconductor industry is thought to be much smaller than in the United States and most semiconductors are imported, any effect is likely to be limited.

7. Given these results it is hard to know what lies ahead for the United Kingdom since even as of the mid–1990s the United States had yet to show any dramatic improvement in labor productivity: there was no evidence of a pickup in capital deepening nor TFP at the time. But, given the current economic environment one could expect that bottlenecks for higher-skilled labor (and thus higher wage bills) may imply that firms will attempt to substitute capital (possibly ICT capital) adding to capital deepening and productivity growth. This is where most of the productivity growth has come from to date. Further in the future, better use of ICT through better organized work processes and network externalities are likely to increase TFP particularly given its low level currently. Given that the share of ICT in output and investment is likely to rise, even with no change in the productivity of the ICT sector, measured productivity at the economy-wide level is likely to increase.

8. The paper will be presented as follows. The next section provides the motivation for the puzzle—if the United Kingdom has been investing in ICT at similar rates of growth as the United States why hasn’t it seen the acceleration in labor productivity? The following part of the paper presents the methodology and results from a formal examination of the U.K. data. Various data measurement issues are discussed though the technical details of the composition of various series are relegated to an annex. Additional, circumstantial evidence of the “new economy” in the United Kingdom is then presented briefly.

B. The Puzzle

9. Investment in ICT has been strong in the United Kingdom for an extended period of time. As a share of nominal GDP, investment in computers and office equipment has been above one percent since 1986, with the exception of a marginal dip during a couple of recession years in the early nineties (Figure 1). ICT investment in the United States has been higher, but not appreciably so. The proportion of total nominal investment in other machinery and equipment devoted to computers and office equipment averaged 18 percent during the period from 1991 to 1999. The growth rate of real investment in computers has also been relatively high, averaging 17 percent in the 1990s and 26 percent in the 1980s (Figure 2).

Figure 1.
Figure 1.

U.K. and U-S. Nominal Investment in Computers as Share of GDP (In percent)

Citation: IMF Staff Country Reports 2001, 124; 10.5089/9781451981445.002.A002

Sources: U.S. Survey of Current Business; and U.K. Blue Book.
Figure 2.
Figure 2.

U.K.: Real Computer Investment

Citation: IMF Staff Country Reports 2001, 124; 10.5089/9781451981445.002.A002

Sources: U.K. Input-Output Supply and Use Balances Tables; U.S. Bureau of Economic Analysis; and International Finance Statistics.

10. Though investment in computers has been relatively high, labor productivity has not shown a commensurate increase. After averaging 2.6 percent from 1990ql to 1995ql, labor productivity (output per job) fell to 1.5 percent in the latter part of the 1990s. While U.K. productivity growth was higher than that in the United States in the earlier period, it has fallen significant below the United States since 1994 (Figure 3).5

Figure 3.
Figure 3.

U.K. and U.S.; Labor Productivity

Citation: IMF Staff Country Reports 2001, 124; 10.5089/9781451981445.002.A002

Sources: Output per filled job, whole economy, U.K. Office of National Statistics; and U.S. business sector output per hours of all persons, Bureau of Labor Statistics.

C. Formal Results for the “New Economy” in the United Kingdom

The Data

11. To examine the contribution of ICT capital to growth of real output and labor productivity, a series representing the ICT capital stock in real terms is required. For the United Kingdom, as an initial benchmark, the following four broad industrial categories are used from national statistics to obtain investment data: “office machinery” (referred to below as computers and office equipment) is used as a proxy for computers; “computing services” is used as a proxy for software; and two categories “transmitters for TV, radio, and phone” and “receivers for TV and radio” are considered proxies for telecommunications. These data, though overestimates of a strictly defined ICT sector, will provide an “optimistic” estimate of the contributions.6

12. The net capital stock for each of the three components of the ICT capital stock is individually constructed from investment in these three categories of equipment. The underlying investment data come from the Input-Output Tables and estimates for other parameters come from Oliner and Sichel (2000) (see Annex I for details). The basic approach is the “perpetual inventory method” (PIM) and is used by the United Kingdom and the United States in the construction of their net capital stocks. In its most basic form, it takes gross investment, subtracts depreciation over the period to arrive at net investment, and adds this to the existing capital stock at the beginning of a period. Notably, a constant depreciation rate of 30 percent per annum on a nominal basis was assumed as a first approximation for computers and office equipment and software and a constant depreciation rate of 11 percent per annum was assumed for the telecommunications category to construct the capital stocks.7

13. The assumption used for investment price deflators for the nominal capital stocks is an important one as the declines in prices for ICT have been large in the last several years. The rapid price declines imply high real capital stock levels and higher growth rates of ICT capital than for the remaining “other” capital stock. Because U.S. price deflators are thought by some to be representative of the price declines in ICT, they are used in the first set of results. The deflators used in the United Kingdom are then employed to observe the sensitivity to the results in section E below.

14. The differences between the U.S. deflators and the U.K. ones arise primarily through different methodologies to control for quality improvement—although note that other limitations, for instance, due to small survey samples may be present. In the United States hedonic methods for computers are used and in the United Kingdom option costing and manufacturer costing, both of which adjust a price quote for a new computer to be comparable to an older model, are used. In the hedonic method, various features of the computer are assigned an indicator—e.g., RAM size, hard disk size, and the presence of a CD drive. A regression is run with the price of the whole computer as the independent variable and the indicator variables as dependent variables. Then fitted values of the regression using the estimated sensitivities to the indicators are used to adjust price quotes so that quality changes are accommodated. In the manufacturer costing method used in the United Kingdom, manufacturers are asked to estimate the cost of any change in the product specification. Because costs are usually overestimated—due to economies of scale in production, for example—half of the manufacturers’ costs are taken as the value of the quality change. This is then used to adjust the producer price indices. The option costing methodology bases prices on various options costs and the prices of particular technical specifications as advertised in computer magazines and then applies the 50 percent rule.

15. Even though the exact categories differ the closest matching U.S. deflator is used for adjusting nominal values in the United Kingdom. Because the “office machinery” category includes more than just computers, a weighted average of U.S. price indices that cover various parts of the office machinery category are used. Software deflators in the United Kingdom are not adjusted for quality. The U.S. software deflators use a combination of hedonics and a matching model to incorporate quality changes. For telecommunications equipment, both the United States and the United Kingdom use quality adjustment. In the United States, the hedonic techniques are only used for some the components of this industry (e.g., electronic switches) but not for others (e.g., fiber-optic cables) and thus the deflator does not decline as fast as for computers and is thought to underestimated quality changes. The telecommuncations deflator in the United Kingdom is quality-adjusted as described above. The U.S. deflators are adjusted for the dollar/pound exchange rate before being applied to the U.K. data. Implicit in this adjustment is that ICT goods compete in a global market and that the U.K. prices quoted in pounds reflect the price declines represented by U.S. dollar deflators. The above-described deflators are used to obtain the real capital stocks for the three ICT capital stocks. The remaining net capital stock uses the implicit deflator obtained from the total real and nominal net capital stocks. The real “other” net capital stock is the total nominal net capital stock less the nominal capital stocks for the three ICT categories divided by the implicit deflator for the entire capital stock.

16. The output series is adjusted to account for the different deflators associated with ICT equipment as well. In this case, output deflators (as opposed to investment deflators) from the United States are used—again matching the industrial categories as closely as possible for the deflators. The nominal output of the three ICT categories is obtained from the Input-Output Table 3 (net of intermediate input that goes back into these categories). The nominal ICT output categories are deflated by the associated U.S. output deflators to obtain the real ICT output. Then the same procedure as for the capital stock is used to reconstruct a “U.S. ICT price adjusted” measure of U.K. GDP assuring some comparability between the net capital stock and output measures.

17. The income shares, also necessary for the exercise performed below, are calculated using national income statistics from the Office of National Statistics (ONS). First, the labor share is taken as the proportion of income earned by labor in the economy each year. This is subtracted from one to obtain the share attributable to capital. To obtain the individual shares for the various ICT categories, the relative nominal proportions of ICT capital in the total nominal capital stock are calculated, year by year. These are weights multiplied by the total share attributable to capital to obtain the income shares of ICT and “other” capital.8

Contributions of ICT to Real GDP Growth

18. The first set of exercises utilize a framework first proposed by Robert Solow (1957) to attribute the growth of output to various inputs. Starting with a Cobb-Douglas, constant returns-to-scale production function, one can derive the following:

y=α c kc+α s ks+α T kT+α o ko+α Ll +a ,(1)

where y and / are the growth rates (expressed in log differences) in output and labor (workforce filled jobs), the growth of the three capital stocks, kc, ks, kr, represent those associated with computers, software, and telecommunications equipment. The contributions from the growth of the remaining capital stock is denoted, k0. The final term, a, is TFP and is calculated as a residual. This formulation differs slightly from that used by Oliner and Sichel (2000) as they include a labor “quality” factor which is unavailable for the United Kingdom. Because this factor is absent, improvements in labor quality are captured by TFP. The α terms represent income shares. Under neoclassical assumptions, these shares equal the output elasticities for each input and they sum to one.9 Thus, the methodology uses a simple disaggregation: the contribution of computers, for instance, is the share of national income which is attributable to the net real stock of computers multiplied by the growth rate in that stock of computers.

19. For the United Kingdom, using data from 1987 through 1999 and U.S. deflators to adjust real output and capital stocks, the contribution from ICT grows from negligible in the 1987–1990 period to 0.6 percent of GDP growth per annum by the 1997–1999 period (Table 1).10 Examining only the component proxying for computers, the contribution to GDP growth calculated here was about 0.5 percent in the final three years. The dominance of this ICT component within the total ICT contribution is probably a result of the larger price declines in this sector relative to the price declines in software and telecommunications equipment. Examining the same ICT category, Kneller and Young (2000) find similar results, computers contribute 0.5 percent to real GDP growth, although they cover slightly different set of years (1996 and 1997). The contribution is probably overstated because semiconductors are not included in the four industry subcategories used to construct the ICT capital stock. Because semiconductors are used as inputs for computers and most of the semiconductors are imported, their inclusion would probably subtract from the overall ICT contribution.

Table 1.

Contributions to Growth of Output in the United States and the United Kingdom, 1987–1999

article image
Sources: Oliner and Sichel (2000); and staff estimates.

All numbers are average annual log difference for years shown multiplied by 100.

Oliner and Sichel’s category is termed “hardware.”

20. The revaluation of ICT output using U.S. deflators raises real GDP growth by a full percentage point in the final three years of the sample, though considerably less in the earlier years. Using similar techniques to examine the effect of hedonic price indices on business sector real GDP growth, Goldman Sachs finds they raise U.K. business sector real GDP by slightly more than ½ percent. The discrepancy with these results could be due to their focus on business sector GDP rather than the broader GDP measure used here, but it appears that much of the difference disappears when output growth is recalculated using chain-weighting. In the early 1990s, the use of U.S. deflators makes little difference in either study.

21. One of the reasons for the large increase in real GDP is that the two sets of deflators (for the United States and the United Kingdom) show a slightly widening discrepancy in the latter half of the 1990s though income shares also play a role. One should keep in mind, however, that when using price deflators that show large declines, as in the case of ICT goods, it is reasonable to expect quantities to adjust. The technique of chain-linking used in the United States and France allows these changes to occur, but the use of base-weighted (Laspeyres) indices in the United Kingdom does not.11 This implies that the fixed value weights of ICT output assigned in 1995 (the base year) will understate the (true) relative weight of the ICT sector for the United Kingdom as time progresses, leading to an overestimate of the residually calculated real quantities. The reason this is the problem of “substitution bias”: categories with declining relative prices are associated with faster growth in quantities. In a Laspeyres index, the (price times quantity) weights are not permitted to adjust so all the impact of lower prices is seen as “real” quantity increases. Thus, the estimates in Table 1 of real GDP growth are biased up and the more so since prices fell dramatically in the latter years.

22. Table 2 presents the same analysis but with chain-linked real GDP and, as expected, real GDP growth falls relative to the non-chain-linked results.12 In the last subperiod (the one with the largest potential impact) the growth rate drops from 3.73 percent per year to 3.22 percent per annum. Alternatively viewed, the growth rate of real output increases by about ½ percent using U.S. deflators and chain-weighting. As does the real GDP growth rates, the contribution of the ICT capital to output growth similarly declines slightly from 0.59 to 0.44 percent in the 1997–1999 and slightly less in the earlier periods. The pattern across the sample period remains the same: increasing ICT contributions to output growth.13

Table 2.

Contributions to Growth of Output in the United States and the United Kingdom, 1987–1999 Chain-weighted U.K. Output using U.S. Deflators for ICT Adjustments

article image
Source: Oliner and Sichel (2000); and staff estimates.

All numbers are average annual log difference for years shown multiplied by 100.

Oliner and Sichel’s category is termed ’ “hardware.”

Contributions to Labor Productivity Growth

23. One can also examine the contribution of ICT to labor productivity. Using equation (1), one can subtract the log of the labor input from both sides and arrive at the following equation:

y-l-[αC(kC-l)+αS(kS-l)-αT(kT-l)+αO(kO-l)]+a,(2)

This formulation divides labor productivity growth, (y–l), into capital deepening, the bracketed term, and total factor productivity (TFP), a. By examining the growth in the ICT capital stocks relative to the total numbers of employees in the economy, the amount of capital deepening attributable to each type of capital can be calculated. And, by adding them, the contribution of ICT capital to capital deepening can be examined. Recall that the final term, a, is calculated as a residual after the contributions of three types of ICT capital and “other” capital are subtracted. The contribution of changes in the ICT capital/labor ratio, [αc(kc-l)+ αs(ks-l)+ αT(kT - l)] an be interpreted as the contribution of changes in the use of ICT equipment by the economy at large.

24. Table 3 shows a less dramatic decrease in labor productivity growth than previously reported using U.K. deflators in the 1997–1999 period. Note that once real GDP is adjusted for real ICT output using U.S. deflators, the drop in the labor productivity growth rate cited in earlier papers and official statistics is not as dramatic—this is simply a reflection of the increased growth rate for real GDP, which is most striking in the latter 1990s, noted above. For comparison purposes, the “unadjusted” productivity growth rates are given as a memo item in Table 3.

Table 3.

Contributions to Labor Productivity for the United States and the United Kingdom, 1987–1999

article image
Sources: Oliner and Sichel (2000); and staff estimates.

Average annual log difference for years shown multiplied by 100. GDP revalued with U.S. deflators for ICT sector.

Percentage points per year.

For Oliner & Sichel, this category includes hardware and semiconductors.

25. Capital deepening attributable to the use of ICT capital amounts to over one-half of the total capital deepening in the 1997–1999 period, 0.56 percent of the total 1.19 percent. This component also increases relative to the earlier periods from a mere 0.08 percent to 0.56 percent in the final period. Comparing the first half of the 90’s with the second, Goldman Sachs reports that the ICT contribution to capital deepening increases from 0.37 to 0.84, slightly higher contributions than the numbers reported here.

26. Within the ICT sector, capital deepening through the use of computers and office equipment is most prominent, accounting for most of the capital deepening regardless of which period is examined. It is somewhat surprising that the telecommunications capital stock contributes so little given the public perception regarding the amount of investment in this area. Two factors are responsible. First, although throughout most of the 1990s the growth of the real telecom capital stock (and the underlying investment growth) has been as robust as for computers and office equipment, the much faster relative declines in computer prices in the last three years has meant faster relative real growth in the computer sector—the real telecom capital stock, while growing, has fallen somewhat behind the real growth of the computer sector. Telecom prices (U.S. deflators) have declined by about 1 percent per year since 1991 and have not seen the dramatic fall in recent years along the lines of computer and office equipment prices (telecom prices declined 1.5 percent per year since 1997 compared to 14.8 percent for computers and office equipment.) Second, the total size of the telecom capital stock is smaller than for computers and thus its income share is also commensurately smaller.

27. Despite a full percentage point increase in TFP relative to unadjusted output figures in the 1997–99 period, TFP still drops compared to the early 1990s—falling from 1.8 percent per year to 1.1 percent. This decline contrasts with the U.S. results in which TFP increased by about ½ percent in the latter period. The results from Goldman Sachs for the United Kingdom also suggest that TFP has also fallen to just over 1 percent per annum in the 1996–99 period. Overall, the labor productivity growth rate declines in the latter period can be viewed as declines in TFP and capital deepening associated with “other” capital which do not match the significant gains made in capital deepening from investment in ICT.

28. The contributions of ICT to labor productivity are also calculated with the chain-weighted output and capital stocks with the results showing similar, but slightly muted, effects (Table 4). The lower GDP growth rates calculated when chain weighted is employed are associated with a lower TFP contribution compared to Table 3—from 1.14 percent to 0.73 percent per annum in the 1997–1999 period. As well, the contributions of ICT capital per worker to productivity are also slightly lower, dropping from 1.19 percent to 1.09 percent per annum in the latter period. Proportionately, however, around one-quarter of the labor productivity growth can be attributed to ICT capital deepening using either of the two GDP growth calculations.

Table 4.

Contributions to Labor Productivity for the United States and the United Kingdom, 1987–1999 Chain-weighted U.K. Output using U.S. Deflators for ICT Adjustments

article image
Sources: Oliner and Sichel (2000); and staff estimates.

Average annual log difference for years shown multiplied by 100. GDP revalued with U.S. deflators for ICT sector.

Percentage points per year.

For Oliner & Sichel, this category includes hardware and semiconductors.

The ICT Production Component of TFP

29. Another way to the examine the effect of ICT on labor productivity and growth is to examine the productivity of the ICT sector itself. The above examination of labor productivity and the associated capital deepening component results in an examination of the “use” of computers: the effects of the “production” of ICT is addressed in this section. In principle, one could examine labor productivity, and thus TFP, of the ICT sector directly by applying the same technique used above on a sectoral basis, using capital and labor within the ICT sector. However, employment statistics are not provided for the same industry breakdowns as for investment and output at this level of specificity in the U.K. statistics and thus this approach is foreclosed. Another approach, used by Oliner and Sichel, who build on the work of Hulten (1987), Triplett (1996), and Whelan (2000a), attempts to solve for TFP in the ICT sector using the so-called “dual” method. Their approach is modified somewhat to better fit the U.K.’s ICT sector.

30. The model of the U.K. economy for the TFP analysis is assumed to have two industrial sectors: an ICT sector and all other sectors.14 Each sector has its own production function. The ICT sector produces final output, but also intermediate inputs to the rest of the sectors. Additionally, the ICT sector utilizes some of its own output as intermediate inputs, perhaps in other firms in the industry. The growth of TFP can be represented by:

TFP =μcTF PICT+μ oTFP o,(3)

where aggregate TFP growth equals the weighted sum of the TFP growth in each of the two sectors. The weights are the share of gross output in the ICT sector as a share of total output (in current pound prices). Gross output includes final output, but also intermediate inputs to the rest of the sectors. The intermediate inputs of ICT goods to the ICT sector are subtracted as they are already included as final output of the ICT sector. The “dual method” used to estimate sectoral TFP relies on the prices of outputs and inputs, rather than their quantities. Very loosely, the basic intuition is that if output prices are falling in a sector, but input prices are relatively stable, then (assuming firms are able to cover costs), the falling output prices are evidence of TFP gains. Said another way, more output is able to be produced with the given inputs, i.e. TFP gains, and these gains allow price declines without firms failing. The derivation of the method and a further discussion of the data are presented in Annex II. Using the annual TFP residuals from the above exercise and the constructed parameters as described, TFP for the ICT sector and the rest of the economy is calculated.

31. Although the growth rate of economy-wide TFP falls in the latter 1990s, an increasing contribution comes from production in the ICT sector (Table 5). Most of this contribution comes from a dramatic rise in the growth rate of TFP in this sector rather than the size of the sector. In the earlier period, 1993–1996, TFPICT is 3.9 percent per annum, higher, but not dramatically higher, than the 2.1 percent of the rest of the economy. By the 1997–1998 period, TFPICT is 14.7 percent per annum while that of the rest of the economy drops to 0.2 percent per annum. The result depends on the relative difference in the logged difference in the output deflator for ITC goods versus the same variable associated with rest of the economy. Given that output prices in the ICT sector are falling and other prices are rising the gap is non-negligible. The final two years of the sample saw ICT prices fall by 8 and 10 percent, respectively, while the rest of the economy witnessed price increases. Overall, then, it is the large price declines in the ICT sector that are reflected in the large increases in TFP in this sector.

Table 5.

Sectoral Contributions to Growth in United States and United Kingdom TFP

article image
Sources: Oliner and Sichel (2000); and staff estimates.

ICT sector is analagous to “computers” from Oliner and Sichel.

For Oliner & Sichel, ITC sums shares of computers and semiconductors. Since semiconductors treated as an input, shares do not sum to 100. Computer shares were 1.63 and semiconductor shares were 0.94 in 1997–98.

32. The TFP growth for the entire ICT sector recorded in the 1997–1998 period is only slightly below that recorded for the computer hardware sector in the United States as computed by Oliner and Sichel (2000)—14.7 percent versus 16.7 percent The numbers are not directly comparable, however, as Oliner and Sichel’s computer sector does not include the software and telecom sectors included here. To make the comparison closer, the exercise is repeated using only the office equipment and computers sector. In this case, real GDP in the United Kingdom is readjusted by only the price deflator for this single sector using the procedure outlined earlier. Thus, the total economy TFP estimates change as well as the growth rates and output shares. In this case (Table 6), the contribution to total economy-wide TFP is lower than in the ICT case, as expected, but is very close to that of the United States. Moreover, the share of this sector’s gross output in GDP is slightly larger in the United Kingdom, also as expected, since it includes a slightly larger definition of the computer sector. Table 6 also shows that TFP in the U.K. office equipment and computer sector is growing at 22.1 percent on average over the 1997–1998 period, while for the United States, the growth rate is about 16.7 over the same period. Probably not too much should be made of this discrepancy since over the entire sample 1993–1998 TFP in the respective U.K. and U.S. computer sectors were 15.3 percent and 15.2 percent—basically the same. The TFP growth in the remaining sectors of the economy are relatively low in both countries, particularly in the latter period.15

Table 6.

Sectoral Contributions to Growth in United States and United Kingdom TFP

article image
Sources: Olinerand Sichel (2000); and staff estimates.

For Oliner & Sichel, ITC sums shares of computers and semiconductors. Since semiconductors treated as an input, shares do not sum to 100. Computer shares were 1.63 and semiconductor shares were 0.94 in 1997–98.

For Oliner & Sichel, semiconductor TFP growth was 22.3 percent in earlier period, 45.0 percent in latter.

33. The TFP growth rates for the U.K. ICT sector (or computer sector) may be biased because they do not include the semiconductor industry and the degree and direction of the bias is not easily assigned. The semiconductor industry is in a different national accounts industrial category (electronic components) that has not been included in the ICT sector. If the U.K. semiconductor industry were to be as productive as the U.S. semiconductor industry and much of the input to the computer sector came from the U.K. semiconductor industry then the huge price declines for semiconductors would add to TFP there and increase the TFP of the entire ICT sector. However, most of the semiconductors used in U.K computer production are imported and thus final domestic ICT output may be lower with their inclusion, which will be reflected in lower economy-wide TFP growth. The split in TFP between the ICT sector and the rest of economy will depend on the output shares, which may be lower for the ICT sector under these circumstances, the input shares, which will be greater, and the relative price changes between the ICT sector, the rest of the economy and the economy as a whole. Without further analysis regarding the size of these offsetting effects the size and direction of the bias to TFP in the ICT sector is uncertain.

34. The output shares presented in Table 5, however, represent total gross output shares and do not account for the fact that much of the ICT sector’s output is used as intermediate inputs to other sectors. When shares are calculated to exclude intermediate inputs, they are not nearly as large (Table 7). Excluding these intermediate inputs, the share of total final output is less than half as large. Table 7 also utilizes the chain-weighted method for GDP to calculate the residual TFP (though this has no effect on the output shares in equation (4) above), showing that the contribution of TFP in the ICT sector to total TFP drops to slightly less than half—still a surprising showing considering the now smaller relative size of the sector. Moreover, while the TFP growth in the ICT sector is somewhat smaller than recorded in Table 5—12.2 percent per annum versus 14.7 percent—it is still quite high.

Table 7.

Sectoral Contributions to Growth in United States and United Kingdom TFP Chain-Aggregated U.K. Output using U.S. Deflators for ICT Adjustments

article image
Source: Oliner and Sichel (2000); and staff estimates.

ICT sector is analagous to ’ “computers” from Oliner and Sichel

For Oliner & Sichel, ITC s sums shares of computers and semiconductors. Since semiconductors treated as an input, shares do not sum to 100. Computer shares were 1.63 and semiconductor shares were 0.94 in 1997–98.

35. Thus, an breakdown in the economy-wide TFP for the United Kingdom shows that the ICT sector is contributing to productivity and output by its high level of TFP growth. It is producing more output per unit of labor and capital inputs as demonstrated by falling prices in the output of this sector relative to its input prices. In fact, over the 1993 to 1998 period, the productivity of the United Kingdom’s computers and office equipment sector was on par with that of the United States.

D. Data Problems and Measurement Issues

36. As mentioned above, the U.K. national accounts data quality-adjusts its producer prices indices using two methods: option costing and manufacturer costing. These methods are considered by some to be inferior to the use of hedonic methods in situations in which the quality of the product is changing rapidly, as in the ICT sector. However, the costs and benefits of using hedonic methods need to be considered carefully: the upfront and maintenance costs of hedonic models are larger and the statistical office needs to assure itself that all the indicators of quality are “valued” by consumers assuring that these elements are priced and, as these elements of quality change, they are included in the model. There are also other limitations that may be present in the U.S. and U.K. indices, for example due to sampling error.

37. Without taking a stand on the relative desirability of hedonic methods, there is, nonetheless, a difference between U.S. and U.K. deflators for computers, (strictly defined) with the United States’ deflators showing sharper declines (Figure 4). Even the weighted average of U.S. output deflators representing a somewhat broader office equipment category shows a difference with the U.K. output deflator applied to the office equipment and computer sector (Figure 5).

Figure 4.
Figure 4.

U.S. and U.K. Computer Deflators (1995=100)

Citation: IMF Staff Country Reports 2001, 124; 10.5089/9781451981445.002.A002

Sources: U.K. Office of National Statistics; and U.S. Bureau of Economic Analysis.
Figure 5.
Figure 5.

U.S. and U.K. Computers and Office Equipment Deflator (1995=100)

Citation: IMF Staff Country Reports 2001, 124; 10.5089/9781451981445.002.A002

Sources: U.K. Office of National Statistics; U.S. Bureau of Economic Analysis; and staff calculations for comparable U.S. index.

38. It is useful to emphasize that the case for using U.S. deflators to examine other countries’ ICT sectors relies on several assumptions. First, price declines—and the quality improvements they represent—are assumed to be the same in the United States and in the other country. This is more likely if computers are a freely tradable good and international competition encourages computers produced in different countries to have similar features. If all U.K. computers were to be imported then the quality improvements imbedded in U.S. price deflators would be the same in the United Kingdom as in the United States. Further, it would seem appropriate to adjust the (annual) U.S. deflator for the average exchange rate (pound/dollar) during each year. This is done in the above analysis. However, the U.K. also exports computers and many computers are purchased domestically. One must assume, then, that markets are competitive so that, imported computers cost the same as domestically produced ones, after accounting for the exchange rate. This assumption may be reasonable for the computer industry, but it appears less likely that software and telecommunications markets are as competitive.16

39. Aside from different price index methodologies, the United Kingdom does not yet use the chain-weighted methodology for national accounts, though it is expected to do so by 2003. Thus, as mentioned above, estimates of growth are biased to the extent that the 1995 weights have changed. A positive bias in the level of real GDP is likely to result given that the volume weights are likely to be too small when prices are falling in areas like the computers, software, and telecom. The continually falling prices imply that the real GDP growth rates are upwardly biased, even without considering the effects of deflators. Though an attempt was made in this paper to discern the effect of chain-weighting, a final answer regarding the degree of the bias would require the chain-weighting exercise to be applied to all components of GDP.

40. The estimates of the real net capital stocks in the three ICT sectors depend on the assumed depreciation—a difficult parameter to nail down. Measuring real net capital stocks is tricky even when price deflators are “correct” since the depreciation of computers, software, and high-tech equipment is not due to normal wear and tear, but is in terms of obsolescence, which is changing through time. For simplicity, little attempt is made in the above calculations to distinguish between the depreciation across the three categories, nor is there any attempt to vary these parameters across time.17 The parameters chosen were the ones used by Oliner and Sichel for 1997 for personal computers and by Jorgenson and Stiroh for software and telecommunications.18

41. A general problem that becomes obvious when this sort of study is undertaken is that the industrial classifications and the collection of data on investment is not precise enough to isolate output of the ICT sector nor investment in ICT goods. The United Kingdom is not the only country for which such data limitations exist. The results of this study, thus need to be viewed tentatively, as “back of the envelope” calculations. Moreover, the nature of the calculations does not allow one to assign statistical errors or to determine the statistical significance of the “estimates”—making comparisons with other countries, e.g. the United States, difficult to assess.

E. “Robustness” Tests

42. To investigate the effect that using U.S. deflators have on the analysis, the first two exercises above are repeated using U.K. deflators for computers and office equipment and software. The U.S. deflator for telecom investment continued to be used because there is no U.K. deflator that is appropriate to apply to the aggregation of the two U.K. telecom sectors used. The software deflator provided by the ONS represents an implicit deflator constructed with “new” estimates of actual software investment, rather than the category “computer services” as used above. The new estimates of nominal and real software investment are used in conjunction with the U.K. deflator. These investment numbers are substantially smaller than the ones used in the study so far and, since there is no corresponding data from which to construct final output, intermediate demand, and net exports, a recalculation of the decomposition of TFP is not done.

43. Not surprisingly, the growth rate of real GDP using the U.K. deflators slows relative to the situation in which the U.S. deflators are used—from 3.73 percent annual growth in the 1997–1999 period to 2.91 percent growth (Table 8). The contributions from ICT capital similarly drop. The ICT capital stock contributes 0.40 percent to the overall output growth rather than the 0.59 percent with the U.S. deflators. The contribution made by computers and office equipment falls from 0.49 percent to 0.34 percent and software falls from 0.05 to 0.01. The miniscule contribution of software investment reflects the now smaller stock of software capital (and its much smaller growth rate) as calculated by the ONS relative to the use of the broad category “computer services.” Since there was no adjustment to the telecom sectors’ deflator, the contribution remains the same at 0.05 percent per annum. The changes for the earlier periods are less dramatic since the difference in the U.S. and the U.K. deflators is smaller in these periods. However, despite the smaller contributions compared to the U.S. case, the ICT sectors’ contribution to output growth is growing over time. For instance, capital deepening associated with computers and office equipment triples between the 1991–96 period and 1997–99.

Table 8.

Contributions to Growth of Output in the United States and the United Kingdom, 1987–1999 Using U.K. Deflators for Computers and Software

article image
Source: Oliner and Sichel (2000); and staff estimates.

All numbers are average annual log difference for years shown multiplied by 100.

Oliner and Sichel’s category is termed’ ’hardware.”

44. Again, not unexpectedly, using the U.K. deflators results in lower labor productivity growth rates across the sample period, with a larger deceleration in the 1997–99 period, than when using U.S. deflators (Table 9). Economy-wide productivity growth drops from 2.59 percent per year to 1.51 percent per year, on average, when comparing the first and second part of the 1990s. However, even with the U.K. deflators, and the decline in labor productivity generally, capita] deepening associated with ICT increases in the last three years of the decade—from 0.84 percent per annum to 1.10 percent. The obverse of this feature is the steep deceleration in TFP growth—from a respectable 1.75 percent per annum in 1991–96 to 0.41 in the 1997–99 period. It is the very low TFP growth that has worried policy-makers in the United Kingdom. Although use of the U.S. deflators makes TFP look better, the phenomenon does not disappear and is more strikingly demonstrated when examined against the U.S. numbers which show an increase of around ½ percent between the two periods.

Table 9.

Contributions to Labor Productivity in real GDP for the United States and the United Kingdom, 1987–1999 Using U.K. Deflators for Computers and Software

article image
Source: Oliner and Sichel (2000); and staff estimates.

Average annual log difference for years shown multiplied by 100.

Percentage points per year.

For Oliner & Sichel, this category includes hardware and semiconductors.

F. Other Evidence Regarding the Prospects for “New Economy” Productivity Gains

45. The above analysis suggests that the ICT sector has contributed to output and labor productivity growth, but does not explain why the other components of labor productivity have fallen or whether they are likely to pick up in the future. Other evidence for the future of the U.K. “New Economy” focus on whether the lackluster growth in TFP in the rest of the economy is likely to reverse and whether the capital deepening attributable to ICT will continue at its present pace.

46. The overall productivity trends shown above hide some sectoral changes that can potentially shed light on future productivity performance. Until mid–1999, productivity growth had been below its long-run trend in the manufacturing sector as a whole while the service sector, specifically business services and finance, enjoyed higher productivity growth (Figure 6). Despite the relatively low productivity growth for the manufacturing sector, recent increases have been noted in some areas: (1) electrical and optical equipment and (2) chemical and man-made fibers (Figure 7). The manufacturing sector only represents 22 percent of gross value added, but the fall in productivity growth has been quite pronounced. The service industries represent 66 percent of GDP of which the business services and finance industry represents 21 percent of gross value added.19 Thus, some of the sectoral productivity growth, since it is occurring in sectors that incorporate elements of the ICT sector and other high-tech sectors, such as biotechnology, is suggestive that the most recent set of productivity gains may be related to the new economy.

Figure 6.
Figure 6.

United Kingdom: Labor Productivity Growth of Selected Industries

Citation: IMF Staff Country Reports 2001, 124; 10.5089/9781451981445.002.A002

Figure 7.
Figure 7.

United Kingdom: Labor Productivity Growth in Selected Industries: Manufacturing

Citation: IMF Staff Country Reports 2001, 124; 10.5089/9781451981445.002.A002

47. There is a positive correspondence between the sectors that are heavy users of ICT equipment and those in which there is a pickup in productivity growth. Such sectors in the United States are financial markets, banking, education, insurance, discrete manufacturing, utilities, and communications and media. While there is not a one-to-one match between these industries and data with which to compute labor productivity growth in the United Kingdom, a glance at some of the broad service industries find some productivity improvement in some of these industries (Figure 8).

Figure 8.
Figure 8.

United Kingdom: Labor Productivity Growth in Selected Industries: Services

Citation: IMF Staff Country Reports 2001, 124; 10.5089/9781451981445.002.A002

48. Divining the future of productivity growth in the ICT production sector is more difficult but the sector’s superior TFP performance is unlikely to degrade rapidly. For the narrow computer-producing sector, the above analysis suggests that TFP growth has been equivalent to that in the U.S. during the 1993–1998 period. The newest findings from the United States, however, suggest that TFP growth in this sector is declining though the share of output from this sector continues to increase.20 A similar phenomenon could be expected in the United Kingdom implying a somewhat lower contribution of ICT production to productivity growth. The 1TC sector more generally appears to have lower TFP growth than the narrow computer sector, but still quite healthy rates of growth. No additional information is available from the United States, but it appears that the price declines that underlie this estimate have been fairly stable in the last few years and might be expected to continue.

49. Labor productivity in the rest of the economy is likely to rise over the next few years, though probably due to more technical reasons than to improvements in TFP. Although the fall in economy-wide labor productivity in the latter 1990s cannot be pinned on an influx of low skilled workers during the cyclical upswing, the low unemployment rate and the several consecutive years of very robust employment growth suggests that available labor resources will not be as plentiful in the future. Assuming the expected slowdown in overall investment is not substantial, the non-ICT sectors of the economy should experience increases in capital deepening. Regarding capital deepening arising from ICT investment, the investment growth rate in ICT may fall off somewhat from its very high levels in recent years. However, as long as ICT capital stock growth exceeds employment growth, capital deepening is still likely to increase from this source as well. Moreover, as the share of ICT investment and output increase in the economy, measures of the contribution to output and labor productivity growth will continue to increase.

50. Recovery of the low TFP growth in the non-ICT sector is likely to be a longer-term project. In part because current wisdom regarding TFP growth improvement relies mainly on the upgrading of the structural underpinnings of the economic environment—innovation, research and development, competition, and entrepreneurship—and the enhancement of the education and skill levels of the workforce, which are difficult to effect in the short-run. In particular, the low past investment in physical and human capital cannot be remedied by a surge in investment in ICT capital. The ability to integrate this type of capital in the production processes takes the time and energy of qualified workers and an environment that is conducive to doing so. Previous studies of technology diffusion in the United Kingdom, though, show that the United Kingdom quickly absorbs new ideas, technologies and work practices from abroad.21 In the areas of the economy unrelated to the New Economy, the low overall level of basic skills of the workforce and the previous low investment levels will need to be improved before TFP growth is likely to pick up—again it is likely to take significant time for improvements to be evident. The government is committed to the task of improving productivity through a multi-pronged approach, as demonstrated by its recent attention to the topic in its Pre-Budget Report and the accompanying paper outlining its strategy.22

G. Conclusions

51. The results of this study suggest that the contribution of the ICT sector in the United Kingdom to output and labor productivity growth has been non-negligible and not very different from the experience of the United States. Much of the positive impact of the New Economy on labor productivity arises through ICT-related capital deepening—about ¼ of productivity growth can be attributed to the increase in the ICT capital to labor ratio. Labor productivity growth slows in the latter 1990s, unlike the United States, and this presents a “puzzle” given that the rate of investment in ICT goods in the United Kingdom has been on par with that in the United States. This slowing, however, appears to be mostly owing to a slowdown in the TFP growth in the non-ITC sector. The growth of TFP in the production of computers, in particular, is virtually the same as in the United States.

52. The conclusions of the above formal analysis should be viewed as the most optimist case for estimates of the contribution of ICT to output and labor productivity growth, but by the same token, they also highlight a relatively dismal picture of labor productivity outside the ICT sector. The use of U.S. deflators for the ICT sector substantially raises the real net capital stock and real output. Further, the industrial categories used for the ICT sector are quite broadly defined. The category “computers and office equipment” covers not only computers but more generic office equipment and “computing services” entails more than software. Some sensitivity analysis is attempted, but more is necessary before the results could be deemed definitive. Nevertheless, a conclusion that seems warranted is that the productivity doldrums in the United Kingdom are not related to the ICT sector and that, indeed, much of the recent productivity growth can be ascribed to this sector, either through the use of ICT capital or through the production of ICT goods.

Annex I: The Data

53. This annex provides details about the data used in the study. It discusses some of the issues arising from the differences between the desired data and that available. The notation is standard: upper-case letters denote levels of a variable and lower-case letters denote its logged difference.

A. Data for Growth Contributions and Labor Productivity Analysis

Real output for the economy (Y)

54. The data for the U.K. real output come from the National Accounts. The output data are gross value added at basic prices, seasonally adjusted. Oliner and Sichel use non-farm business less housing as their concept for output. In the United Kingdom, it is not possible to exclude the agricultural sector, so the output concepts are slightly different.

Capital Stocks (Kc, Ks, KT, Ko)

55. For the growth accounting exercises above, the appropriate concept of the capital stock is the “productive” capital stock rather than a “wealth” capital stock. Productive stocks measure the income-producing capacity of the current stock of capital during a given period. Wealth stocks are the current market value of the assets in use. Since the exercise is meant to capture the contributions of ICT capital to growth in output and labor productivity, the “productive” capital stock is the desired one—the interest is in how much the ICT capital stock can produce not what it is currently worth. Unfortunately, productive capital stocks are not available yet in the United Kingdom, though there are now ongoing efforts to produce them.

56. To construct the various components of the capital stock, a nominal total capital stock is built up from gross fixed capital formation (investment), capital consumption (depreciation), changes in inventories, and acquisitions less disposals of valuables for the total economy. The nominal stock is constructed under the assumption that the nominal and real stocks were equal in 1995, the base year. Separately, the nominal capital stock for each of the ICT products are built up by using Table 6 of the U.K. Input-Output Annual Supply-Use (“Input-Output”) Tables in which the gross fixed capital formation is provided by industry product. Thus, investment in computers and office equipment is summed across all industries to provide a total investment figure. In this case the stock of capital was assumed to be zero in the year before the first figures were available. The industry classifications used are Office Equipment (69) for computers and office equipment; Transmitters for TV, Radio, and Phone (74) added to Receivers for TV and Radio (75) for telecommunications; and Computer Services (107) for software. The investment data begins in 1970 for computers and office equipment and for software. The investment price deflators (discussed below) begin in 1980 for computers and office equipment and 1989 for both software and telecommunications. Official investment data are available through 1998. Investment in computers and office equipment for 1999 is estimated as the average proportion of office machinery investment in total investment over 1996–1998. For software and telecommunication, the ratio of these components to the office machinery category are maintained at their 1998 level. The capital stock for dwellings is also constructed from investment and depreciation and then subtracted from the total economy nominal capital stock.

57. To obtain the real capital stocks, for each ICT sector, a corresponding deflator from the U.S. investment deflators are used. To obtain the real capital stock for the remaining capital stock, first the sum of the nominal ICT capital stocks is subtracted from the entire economy nominal capital stock. Then, the deflator obtained by dividing the original nominal capital stock series by the real series (both excluding dwellings) is applied to the “other” nominal capital stock. Ideally, one would have preferred a deflator appropriate to the types of capital left in the “other” category, but this would require all the “other” nominal capital stock components and their deflators. It is unlikely this approximation affects the results much, as the ICT capital stocks are relatively small proportions for most of the years.

ICT Investment Deflators

58. The deflators used for obtaining the ICT capital stocks come from the U.S. National Income and Product Accounts (NIPAs). Since the U.K. computers and office equipment category is broader than the “computers and peripherals” category used in the United States the price deflators for three U.S. categories were averaged using weights equal to the nominal share of private fixed investment to the total private fixed investment in these categories. Implicitly this assumes that the U.K. investments are made in the same proportion across these categories as in the United States. The U.S. categories used are “Computers and Peripheral Equipment,” “Office and Accounting Equipment,” and “Photocopying and Related Equipment.” The deflators for software and telecommunications were used without alteration from the U.S. categories “Software” and “Communications Equipment.”

59. For the analysis using the U.K. deflators, the deflator for computers and the more general computers and office equipment categories were provided by the ONS and represent the producer price indices used to value real GDP. The deflator for software investment was also provided by ONS and is not publicly available. No specific price indices currently exist for computer software and the general investment-based price indices are used to derive real price estimates.

Labor (L)

60. The labor variable is the U.K. workforce jobs, seasonally adjusted, from Labor Market Trends (produced by the ONS). Workforce jobs are calculated by summing employee jobs, self-employment jobs from the Labor Force Survey, HM Forces, and government-supported trainees. Workforce jobs is the concept of the labor force used in domestic measures of productivity. In the United States, by contrast, the labor hours is used in official productivity measures. Oliner and Sichel used the growth rate in hours of all persons in the non-farm business sector from the BLS Productivity and Cost release. Although hours worked is collected and published in the United Kingdom, these data are thought to be imprecisely measured and most researchers have opted for the workforce jobs variable in previous productivity studies.

B. Data for TFP Disaggregation Analysis

ICT Output Deflators (Pc, Ps, PT, Po)

61. The ICT output deflators used in the study come from the database on “shipments of manufacturing industries” produced by the Bureau of Economic Analysis. The data are publicly available at www.bea.doc.gov/bea/dn2/gpo.htm. This database includes chain-type price indices for industries using 2-, 3-, and 4-digit SIC codes. The deflators for the same categories as mentioned above are obtained for the purposes of deflating the U.K. ICT output. The deflator applicable to the remaining output of the economy, p0, is the implicit GDP deflator for the United Kingdom obtained by dividing nominal value added at basic prices by real value added at basic prices.

62. As stated in section C, the U.S. ICT output deflators are used to revalue the output of the U.K. ICT sector for the purposes of adjusting real U.K. GDP to be consistent with the adjustments made to the capital stock. Thus, nominal final output of the ICT sectors are deflated by the U.S. output deflators to obtain real ICT output. Nominal GDP in the non-ICT sector is obtained from subtracting the nominal ICT output from total nominal GDP. This remaining GDP is then deflated by the implicit GDP deflator described above.

Annex II: The “Dual Method” to Disaggregate TFP Growth

63. This Annex describes the “dual method” approach used to disaggregate TFP growth into a component associated with the ICT industry and a component associated with the remaining industries. It outlines the formal model and briefly provides the underlying intuition.

64. As noted in the text, the growth of TFP can be represented as a weighted average of TFP in the two sectors under consideration: ICT and the rest of the economy:

TFP =μcTF PICT+μ oTFP o,(a1)

65. The approach of the so-called “dual method” used to estimate sectoral TFP relies on the prices of output and input, rather than their quantities. The intuition for this approach is as follows: if output prices in a sector are dropping and input prices are relatively constant, then the TFP growth in this sector must have been more rapid than other sectors. If it were not, the lower output prices (with stable input prices) would have driven this sector out of business. Thus, the lower output prices are viewed as representing the ability to produce more output with the same inputs. For the ICT sector, then, one can derive the following:23

TFPICT =TFP o+(p ICT-po )-αIC To(pICT -Z ),(a2)

where the p’s represent price changes in the ICT sector and “other” sectors and the αictº represents the share of current pound input costs associated with computer and office equipment inputs to the “other” sector. The numerator of this share represents the pound value of computers and office equipment used as an intermediate input in the “other” industries, excluding the “own” use of such equipment as inputs in the computer and office equipment sector itself. The denominator is the current pound output of the rest of the economy, the “other” sector. Z is the share weighted growth of capital and labor costs of the economy as a whole. That is, Z equals [(RK/PY) r + (WL/PY)w], where R and W are the rental and wage rate respectively, r and w their logged differences, K is the capital stock and P is the price level for the economy as a whole. The final term in the equation represents an adjustment to TFP0 based on the notion that TFPICT should be lower if the “other” industries are also able to benefit from lower (relative) input costs into their businesses. In this analysis, however, this term is small in comparison to the others.

66. In addition to the standard neo-classical assumptions of using Cobb-Douglas, constant returns to scale technology, other assumptions in this formulation are:

  • The “other” sector only produces final goods.

  • The two sectors employ the same mix of workers and capital goods.

  • The rental rates and wage rates (and their growth rates) are the same across the sectors.

The two equations above represent a system of two equations and two unknowns (TFPICT and TFP°) and can thus be solved algebraically.

67. To calculate the output shares and input cost shares, final domestic output from the three ICT sectors and the associated output price deflators are needed. The Input-Output Tables 2 and 3 are used for the periods 1992 through 1998.24 These tables permit the calculation of final output for each of the three ICT sectors as a sum of four categories: final consumers’ expenditures, gross capital formation (investment and changes in inventories), and exports less imports.25 In the tables, government consumption and investment are included in the first two categories. These final outputs are then summed and used to calculate the output shares, μICT, and μo, and the intermediate demands less own-industry (intermediate) demands are summed across the three categories to allow calculation of the αICTo, the input share ICT going to other industries.26 The output deflators utilize the U.S. output deflators produced by the BEA that are used to deflate industry output for the U.S. GDP and are the same deflators used to adjust the U.K. output above.27 Under the assumptions above, the variable Z (the share-weighted growth of capital and labor costs) can be calculated by using the “dual” which says that the growth in the economy’s output prices (the GDP deflator in this case) equals the growth in the (share) weighted average of input costs, minus the growth rate of TFP, that is, p = Z - TFP, or Z = p + TFP. This completes the discussion of the model and the data used for its implementation.

References

  • Estevão, Marcello, (2000), “The New Economy in France: Developments and Prospects,Chapter III in France: Selected Issues, Staff Country Report No. 00/148, November, International Monetary Fund, Washington, DC.

    • Search Google Scholar
    • Export Citation
  • Hubert, Florence and Nigel Pain, (2000)Inward Investment and Technical Progress in the UK Manufacturing Sector,Economics Department Working Paper, No. 268, October, Organization for Economic Co-operation and Development, Paris, France.

    • Search Google Scholar
    • Export Citation
  • Hulten, Charles R., (1978), “Growth Accounting with Intermediate Inputs,The Review of Economic Studies, October, 45:3, pp. 511518.

    • Search Google Scholar
    • Export Citation
  • Input-Output Annual Supply-Use Tables, 1992–1996, 1997, 1998, Office of National Statistics: London, United Kingdom.

  • Jorgenson,Dale W., and Kevin J. Stiroh, (2000), “U.S. Economic Growth in the New Millinium,Brookings Papers on Economic Activity 1, pp. 125211.

    • Search Google Scholar
    • Export Citation
  • Kneller, Richard, and Garry Young, (2000), “The New British Economy,National Institute of Economic and Social Research, September: London, United Kingdom.

    • Search Google Scholar
    • Export Citation
  • Labor Market Trends, (2000), Office of National Statistics, August: London, United Kingdom.

  • Lynch, Robin, (1996) Measuring Real Growth—Index Numbers and Chain-Linking” Economic Trends, No. 512, June, Office of National Statistics, London, United Kingdom.

    • Search Google Scholar
    • Export Citation
  • Oliner, Stephen D. and Daniel E. Sichel (2000), “The Resurgence of Growth in the Late 1990s: Is Information Technology the Story?” Federal Reserve Research and Statistics Working Paper Series, Federal Reserve Board, Washington, D.C.

    • Search Google Scholar
    • Export Citation
  • Oliner, Stephen D. and Daniel E. Sichel (2000), “The Resurgence of Growth in the Late 1990s: Is Information Technology the Story?Journal of Economic Perspectives, 14:4, pp. 322.

    • Search Google Scholar
    • Export Citation
  • Pre-Budget Report, (2000), HM Treasury, November: London, United Kingdom.

  • “Productivity in the U.K.: the Evidence and the Government’s Approach,” (2000), HM Treasury, November: London, United Kingdom.

  • “Review of Short-Term Output Indicators” (2000), “National Statistics Quality Reviews”, Series No. 2, October 2000, Office of National Statistics, (www.statistics.gov.uk/nsbase/methods_quality/quality_review/economy.asp), London, United Kingdom.

    • Search Google Scholar
    • Export Citation
  • Rizki, Uzair, (1999), “Investment in Computer Software and Large Databases,” Office of National Statistics, (www.statistics.gov.uk). London, United Kingodm.

    • Search Google Scholar
    • Export Citation
  • “Shipments of Manufacturing Industries,” Bureau of Economic Analysis, (http://www.bea.doc.gov/bea/dn2/gpo.htm).

  • Solow, Robert, (1957)Technical Change and the Aggregate Production Function,Review of Economics and Statistics, August, 39:3, pp. 6594.

    • Search Google Scholar
    • Export Citation
  • Triplett, Jack E., (1996), “High-Tech Industry Productivity and Hedonic Price Indices,” in Industry Productivity: International Comparison and Measurement Issues, Proceedings of May 2–3, 1996 OECD workshop. Paris, Organization for Economic Cooperation and Development, pp. 119142.

    • Search Google Scholar
    • Export Citation
  • Walton, David, Ben Broadbent, and Peter Sullivan, (2000), “Technology and the New UK Economy: Following in the Footsteps of the United States,” Goldman Sachs, December: London, United Kingdom.

    • Search Google Scholar
    • Export Citation
  • Whelan, Karl (2000a), “Computers, Obsolescence, and Productivity,” Federal Reserve Board, Finance and Economics Discussion Series Paper 2000–6, February: http://www.bog.frb.fed.us/pubs/feds/2000/index.html, Washington, D.C., United States.

    • Search Google Scholar
    • Export Citation
  • Whelan, Karl (2000b), “A Guide to the Use of Chain Aggregated NIPA Data,” Finance and Economics Discussion Series Paper 2000–35, June, Federal Reserve Board, http://www.bog.frb.fed.us/pubs/feds/2Q00/index.html. Washington, D.C., United States.

    • Search Google Scholar
    • Export Citation
1

Prepared by Laura E. Kodres. The author is grateful to the U.K. Office of National Statistics, particularly Prabhat Vaze, as well as Garry Young, Daniel Sichel, Ben Broadbent, Nicholas Oulton, and participants at a round-table discussion in the HM Treasury for their provision of data, results, valuable insights, and comments. These individuals are absolved from all remaining errors in this paper.

2

Another characteristic of the “new economy” is reduced volatility of output and inflation. This attribute has been evident in the United Kingdom as well.

3

A variable measuring possible improvements in labor quality is not available and thus any productivity gains related to labor quality will be subsumed in total factor productivity.

4

See accompanying United KingdomSelected Issues paper “What’s So Special About This Cyclical Expansion in the U.K.? A New Perspective from Labor Markets” for more details.

5

In the United Kingdom, labor productivity is traditionally measured as output per worker whereas in the United States the measure is output per hour worked of all persons.

6

Though the series includes anything from staplers to Pentium 4 computers, most of the growth in the series is thought to reflect the growth of computers and peripherals.

7

These rates were also used by Jorgenson and Stiroh (2000). A measurement of the “productive” capital stock is the preferred concept, but as such a measure is unavailable, the PIM is used as an approximation. See Annex I for further discussion of the notions of productive capital stock versus wealth stocks.

8

Oliner and Sichel (2000) provide a more precise method for calculating income shares that include the depreciation rate, the capital loss, the real rate of return on capital, and a tax factor. Further work will need to be done to reproduce these weights for the United Kingdom with data provided by the ONS and others.

9

Such assumptions include perfect competition (implying prices are equal to their marginal costs), constant returns-to-scale technology, and that the economy is on the production possibility frontier (implying businesses are at or near their optimal long-run levels of labor and capital inputs).

10

The breakdown into the 1991 to 1996 and 1997 to 1999 periods is undertaken to accord with the time periods represented by negative and positive output gap periods.

11

See Chapter III, “The New Economy in France: Developments and Prospects,” in France: Selected Issues, Staff Country Report No. 00/148 for a similar exercise to that executed here using French hedonic deflators.

12

The real ICT capital stock was also chain-weighted using a method called a “Tornqvist” index approximation, which is close to the ideal “Fisher formula” used in the United States. In this approximation, the growth rate of the index is a weighted average of nominal shares in the current and previous periods. This growth rate is then used to construct levels by assuming a base year (here assumed to be 1995). Inventories are ignored in the construction of this chain-weighted capital stock. For a discussion of the issues regarding chain-weighted in the United States and possible pitfalls in the area of ICT investment and output see Whalen (2000b). For a review of these concepts by the Office of National Statistics, see Lynch (1996).

13

The higher contribution of the telecommunications sector in Table 2 is partly due to a change in the assumptions for depreciation. Table 1 assumed that all three categories of ICT used a depreciation rate of 30 percent and a capital loss of 34 percent per annum. In Table 2 the capital loss component was eliminated and the depreciation rate for telecommunications capital was assumed to be 11 percent.

14

The exercise performed here is also reproduced for the computer and office equipment sector alone. Oliner and Sichel (2000) use three sectors, including a semiconductor sector.

15

Updated numbers for the United States for the year 2000 show a marked pickup in TFP in the rest of the economy and a slowdown in TFP of the computer sector.

16

The U.S. Justice Department’s case against Microsoft alleged that it was a dominant firm in the software market and inhibited competition, driving up prices. Even though deregulation of the telecommunications market has taken place there have been various complaints about the level of competition both in the United States and the United Kingdom.

17

A lower depreciation rate, 11 percent, underlies the capital stock for the telecom sector in the calculations for Tables 2, 4, and 7.

18

Oliner and Sichel, however, differentiate between the various ICT sectors and allow the depreciation rates to change through time, for the most part following the assumptions used by the Bureau of Labor Statistics. Further work will be needed to remedy some of these shortcomings.

19

In the United States, the output of the financial, insurance, and real estate (FIRE) sector represented 19.4 percent of GDP in 1997.

20

The productivity growth in the U.S. semiconductor industry is holding steady at around 50 percent per year according to updated work by Oliner and Sichel.

21

See, for example, Hubert and Pain (2000).

22

Chapter 3 of the Pre-Budget Report from the HM Treasury, November 2000 and “Productivity in the U.K.: the Evidence and the Government’s Approach”, HM Treasury, November 2000.

23

For a more formal derivation of this equation see Appendix B of Oliner and Sichel’s Federal Reserve working paper.

24

Consistent data is maintained only for these years as the pre–1992 Input-Output Tables use SIC80 for industrial sector breakdowns.

25

Distributors trading margins and taxes less subsidies are also added to obtain a similar concept of GDP as used in the rest of the study. Since these sector output levels are used to calculate the share of output in the entire economy (to weight the sectoral TFPs) and the total economy TFP was calculated using this GDP concept, for consistency, these variables are added to obtain final output. Distributors trading margins net to zero across the entire economy.

26

Table 5 includes intermediate inputs in the calculations of the shares and thus the shares are not representative of final (net) output but represent gross total output of the ICT sector.

27

Additional information about the data and calculations are in Annex I.

United Kingdom: Selected Issues
Author: International Monetary Fund