From: michael perelman (michael@ECST.CSUCHICO.EDU)
Date: Wed Dec 22 2004 - 20:37:40 EST
Jerry, here is a section of the book that discusses the problem. Gerald_A_Levy@MSN.COM wrote: > >Michael P, > >In what countries? >Over what period of time? For what recessions? >What exactly is being measured? > >Sorry I don't have your book, but I'd like to look at the data. > >In solidarity, Jerry > > > _Empirical Limits to the Study of the Replacement Investment_ The exact nature of the replacement decision should be a matter of interest for those who frame economic policy, especially considering the lip service paid to economic modernization. Yet, despite the elegant mathematical tools and the extensive data bases available to economists, we know astonishingly little about actual replacement practices. Common sense suggests that equipment should be replaced once it becomes obsolete, but this idea is tautological. On a superficial level, the term 'obsolete' conveys the idea that the article in question should be replaced. Upon closer scrutiny, the idea of obsolescence becomes considerably vaguer. For example, Strassman has catalogued a number of supposedly obsolete techniques which firms continued to adopt long after more advanced technologies had become available because they were serving particular markets or using special inputs which made these technologies a profitable proposition (Strassman, 1959a). For example, charcoal blast furnaces continued to be built long after anthracite and coke furnaces were becoming widespread, because they produced the purer output which blacksmiths needed (Strassman, 1959b; also see Ono, 1981). Despite the difficulty in clearly defining obsolescence, this concept must be part of any theory of investment. In Griliches' words: [A] theory of investment without obsolescence is like Hamlet without the prince ... [Yet] we have very little information about the expected life of different machines, or about the factors that determine the relative prices of different ages of machines in the used machinery markets. Without this knowledge, we don't know how to measure any kind of capital. (Griliches, 1963, pp. 123 and 135‑6) Replacement investment follows a different course from net investment (Klaasen _et al*.*_, 1961, p. 227), but Feldstein and Foot rightly complained that all the attention had been given to net investment (Feldstein and Foot, 1971, p. 49). Their call for more empirical analysis of replacement investment went largely unheeded. In fact it came after a brief flurry of interest in the subject had already come to an end. This debate began when Joan Robinson initiated the Cambridge Controversy by asking how the capital stock could be measured (Robinson, 1953‑4; see also F.M. Fisher, 1969). Although she did not directly broach the question of replacement investment, the relationship between the difficulty of measuring capital and replacement investment is self‑evident. This connection is most obvious in the theory of vintage capital models (Whitaker, 1966). Economists then began wrestling with the notion of vintage capital models until Tobin and his co‑workers cleverly applied some unrealistic assumptions, including the ubiquitous notion of perfect knowledge, to develop a vintage model consistent with neo‑classical theory (Solow _et al_., 1966). Robinson might have conceded that with perfect knowledge of future prices and discount rates, appropriate shadow prices for capital goods could be theoretically calculated. Her quarrel was with the realism of that approach. Once Tobin and his associates showed how vintage models could be stuffed back into the neo‑classical bag, interest in the subject died off rapidly. Only a handful of articles have attempted to analyse the phenomenon of replacement investment. Feldstein, along with Foot, attempted to analyse scrapping behaviour empirically (Feldstein and Foot, 1971). He returned to the subject again in another joint article, where he gave a rigorous mathematical argument for dropping the assumption of proportionate scrapping (Feldstein and Rothschild, 1974). No data accompanied this theoretical exercise. It stressed the importance of changes in the assumed constant rate of decay of equipment rather than the actual scrapping decision. In some instances, capital goods are surprisingly long‑lived. The steel mills of Youngstown represented a prime example, at least until their recent demise. Many of these plants had remained in operation for three‑quarters of a century or more. In other cases, capital goods become obsolete in a short period of time. Micro‑computers offer a popular example. The difficulty of pin‑pointing the act of scrapping adds a further complication to the analysis of replacement investment. Even with as familiar an asset as a bus, dating a piece of equipment is all but impossible because of the complex history of partial improvements and repairs involved (D.C. Holland, 1962, p. 417); in Paul David's words: '[A]lthough textile machinery survived 'in place' to legendary ages, in many instances the equipment had been rebuilt in piecemeal fashion ‑‑ sometimes more than once ‑‑ so that by the end of its formal service life it contained scarcely a bit of metal dating from its debut on the mill floor' (David, 1975, pp. 177‑8). For the aggregate economy, replacement of capital goods is not equivalent to retirement. Many capital goods find their way to second‑hand markets. Plant and equipment, no longer used for their original purpose, are frequently put to other uses or worked less intensively (Foss, 1981a, 1981b and 1985). The inclusion of the multitude of options for redeploying replaced capital goods blurs the boundaries between replacement and expansion. For example, British firms report difficulty in distinguishing the replacement from improvements in technique (Barna, 1962, p. 31). Even when old capital is no longer actively used, it may continue to serve as an inventory of capacity to meet possible future peak load needs (Oi, 1981). Terborgh illustrated the changing uses to which capital can be put with a delightful history of a fictional 1890 vintage locomotive: It began in heavy main‑line service. After a few years, the improvement in the art of new locomotives available and the development of the art of railroading made the unit obsolete for service, which was taken over by more modern power. It was thereupon relegated to branch‑line duty where the trains were shorter, the speeds lower, and the annual mileage greatly reduced. For some years it served in that capacity, but better power was continually being displaced from main‑line duty and 'kicked downstairs' onto the branch lines, and eventually the locomotive was forced out at the bottom, to become a switcher in one of the tanktown yards along the line, but the march of progress was relentless, and in the end, thanks to the combination of obsolescence and physical deterioration, it wound up on the inactive list. For some years more it lay around, idle most of the time, but pressed into service during traffic peaks and special emergencies. Finally, at long last, the bell tolled and it passed off the scene to the scrap heap. (Terborgh, 1949, p. 17; see also 1945, pp. 102‑3) Indeed, Terborgh reported that in 1927‑29 new locomotives averaged about 50,000 miles per year. Thirty-five year-old units only ran about 14,000 miles (Terborgh, 1945, p. 105). Even from the scrap heap the locomotive might have made a modest contribution to the economy. Parts might be used on other locomotives. Imaginative second‑hand dealers profit from selling components for purposes totally unrelated to their original use. A firm might pay $80,000 for the right to demolish an obsolete chemical plant in the expectation of selling the plant and materials for $2,000,000 (Deigh, 1987). Firms may hold what appears to be excess capacity one moment only to face backlogs the next (De Vany and Frey, 1982; see also Steindl, 1976, p. 8). The incentive to use capital rather than labour to meet sudden increases in demand depends, in part, upon the ratio of capital costs to labour costs. For example, in the textile industry where labour is a relatively small fraction of total costs, surges in demand tend to be met by increasing labour inputs rather than capital. By contrast, in coal mining where labour costs are relatively high, industry meets demand shocks by moving pits on‑ and off‑line (Rowe, 1928, pp. 10‑2). In addition to the complications associated with the various possibilities for replacing obsolete capital goods, the common empirical measures of the existing capital stock are seriously flawed. Until recently, imported capital goods were not included in the measures of current investment in the US. When this omission was discovered, the investment account had to be revised upwards for more than $20 billion in 1978 alone. Additionally, purchases of home computers are included in business investment (Anon., 1983). Although economists are trained to count in terms of monetary values, satisfactory monetary values are often unavailable except for newly purchased capital goods. Once capital is installed, indicators of its value become increasingly unreliable. Accepted accounting practices are almost entirely based on cost rather than the actual values of installed plant and equipment (Beidleman, 1973, p. vii). Thus, accounting values bear little relationship to the actual economic values (Beidleman, 1976; F.M. Fisher and McGowan, 1983) unless depreciation rates track the actual economic deterioration of plant and equipment. To construct such depreciation formulae is all but impossible. It would require the relative values of two vintages of machines to be independent of changing price ratios. Even if such depreciation formulae could be devised, they probably would not be used. Allowable depreciation is determined as much by political as economic considerations. The book value of much of the old plant and equipment will tend to be insignificant, having been either all, or mostly depreciated away. Such discrepancies are of major importance in the study of scrapping. Econometricians continue to debate about the appropriate depreciation formula (Hulten and Wykoff, 1981a and 1981b). Does capital depreciate evenly or does it lose its value more quickly in the early years? Some estimates suggest that capital, once installed, may actually appreciate over the first few years, since one‑ or two year‑old investment seems to have a greater effect on aggregate production than new investment (Pakes and Griliches, 1984). Sylos‑Labini went even further, estimating that current investment tends to decrease productivity because of what he calls the 'disturbance effect' (Sylos‑Labini, 1983‑4, p. 173). I will return to this subject later, in discussing the subject of 'learning by doing'. In reality, the depreciation rates implicit in actual market prices are highly irregular. For example, the annual rate of market price decline for two‑year old Ford F600 trucks ranged from 7.8 per cent in 1976 to 25.8 per cent in 1971 (Bulow and Summers, 1985, p. 27). Government estimates of capital stocks are unsatisfactory. They are constructed on the basis of the perpetual inventory method, which determines the capital stock each year by adding the difference between the value of new investment and an estimate of the annual depreciation of existing capital goods. Unfortunately the assumed pattern of depreciation is based on a predetermined economic life for each category of investment goods, usually based on Bulletin F of the Internal Revenue Service (first published in 1931) or Winfrey's 1935 study, developed from mortality curves compiled by workers at the Engineering Experiment Station of Iowa State College during the 1920s and 1930s (Winfrey, 1935). The capital is then assumed to depreciate according to some fixed pattern, such as double‑declining balances or straight line depreciation. The estimated lifetime of capital introduces further bias in the aggregate depreciation figures. An error of one‑third in the assumed asset life of capital alters the estimated size of the capital stock by about one‑third (Redfern, 1955, pp. 142‑7). The economic lives used for tax purposes often are unrelated to the actual economic lifetime of the plant and equipment. By the time a machine tool is scrapped, three entire generations of tools can be written off (Beidleman, 1976). Feldstein and Rothschild note that the basis for the 1942 edition of Bulletin F lives was never published, although the estimated lives were based on Winfrey's work, as well as conferences with industry and statistical studies (Feldstein and Rothschild, 1974). Even so, these estimates of capital goods lives are not sufficient. For example, Hickman questioned the accuracy of Bulletin F, speculating that the standard of obsolescence applied during the 1930s was atypical because plant and equipment might be less readily scrapped during a depression (Hickman, 1965, p. 241). In fact, a depression may actually shorten the life of capital goods, creating more incentive for scrapping and less for the renewal of plant and equipment (Boddy and Gort, 1971; and Eisner, 1978, p. 182). In effect, the permanent inventory method of calculating capital stock suggests that the retirement of capital is a wholly technical decision, unaffected by prevailing economic conditions. It presumes that no matter what sort of shocks occur, the relative prices of different vintages of capital goods will remain unaffected. The weakness of such assumptions is obvious. You do not have to be a firm believer in the Kondratieff cycle to suspect that technical change does not always evolve regularly, but often seems to come in spurts. Sometimes it will be concentrated in specific industries. At other times, it will be more evenly distributed among industries. In addition, capital decays faster when it is utilized more intensively (Keynes, 1936, pp. 69‑70; Marx, 1977, pp. 527‑9). Certainly statistical evidence indicates that failure rates do rise with capital use (Davis, 1952; Jorgenson, McCall and Radner, 1967). Are we to believe that capital equipment is kept in operation for a fixed period of time, regardless of the prevailing long‑run macro‑economic conditions? Otherwise, the permanent inventory procedure is unjustified for estimating both the capital stock and models of investment. Yet every empirical study of which I am aware suggests that the capital stock decays irregularly. Most observers agree current economic conditions affect the scrapping decision. For example, a number of British firms were asked to give the percentage of assets of different vintages surviving in 1957, as well as the percentage of assets of different vintages scrapped in the same year. The first question elicited survival rates and the second, mortality rates. Survival rates constructed from current mortality rates were higher than the rates obtained directly, leading to the conclusion: 'This is consistent with the hypothesis that the rate of scrapping varies with the trade cycle and in boom years scrapping is postponed' (Barna, 1957, p. 88). Parkinson's data on the pattern of scrapping ships supports Barna's finding of variable scrapping rates: The first conclusion that emerges from a general survey of scrapping rates is that the service life of a ship is not rigorously determined by factors of a technical nature governing structural strength ... The decision to scrap a vessel will depend very greatly on anticipated movements of freight rates and the rising trend of repair costs with increasing age ... A sudden increase in the demand for tonnage at a time when new building cannot be increased is likely, therefore, to give rise to some postponement in the scrapping of those types of tonnage in demand, almost irrespective of age. (Parkinson, 1957, p. 79) Ryan found that in the Lancashire cotton industry between 1860 and 1838, 38 per cent of the machinery replacements occurred in the boom years 1906‑1908, 1912‑14, and 1919‑21 (Ryan, 1930, p. 576). Feldstein and Foot wrote: 'Expansion investment causes an offsetting fall in replacement investment, supporting the view that firms postpone replacement during periods of expansion investment and accelerate replacement when there is less expansion investment' (Feldstein and Foot, 1971, p. 54). This conclusion must be taken with a grain of salt, based on the justifiable criticisms of their estimates by both Jorgenson and Eisner (Jorgenson, 1971, p. 1140; Eisner, 1978, pp. 175‑88), which I will discuss later. Since replacement does not occur with the regularity assumed by the permanent inventory method, the government's published series for the US capital stock is inaccurate. Realization of the problems resulting from the assumption of the permanent inventory method prompted Feldstein and Foot to write: The rejection of the proportional replacement theory as a description of short‑run behavior has important implications. First, all of the current methods of estimating parameters of net investment may be misleading because they rely on the proportionate replacement assumption either to derive a net investment series or in specifying a gross investment series. Second, understanding and forecasting short‑run variations in gross investment requires a more complex model of replacement investment. (Feldstein and Foot, 1971, p. 57) The implicit widespread acceptance of the perpetual inventory assumption is ironic. Despite the recent emphasis on rational expectations, which purportedly highlights business' supposed immediate reaction to new economic information, the economics profession still assumes away (or at least it relies on data that assume away) any effect of expectations on the retirement of capital. This practice is especially troubling since capital accumulation is generally accepted as central to the functioning of the economy. The unsatisfactory measures of the capital stock substantially complicates the analysis of replacement investment. In contrast to econometricians' solicitude about the characteristics of their residuals, they display no concern about the shaky empirical underpinnings of this measure of capital, although they frequently include this measure of capital as a _key variable_ in their macro‑economic models. Ideally we need measures of present value, but the present values, which are calculated in journals and on blackboards, require knowledge about the future which is impossible to obtain (Hayek, 1941, p. 90). You need only consult Lock's description of the survey of the capital stock of the Dutch cigar industry to appreciate the almost impossible challenge of calculating an adequate, let alone precise measure of the capital stock (Lock, 1985). As Boddy and Gort noted: 'Capital stocks are midway between an observable phenomenon and a state of mind. One can touch and see the tangible assets, but to measure them in constant units requires a theory of production and a host of assumptions' (Boddy and Gort, 1973, p. 245). Mistakes in the depreciation rate can significantly affect the estimated rate of return on capital. The US Department of Commerce estimated that in 1983, profits for non‑financial corporations based on Bulletin F lives ranged between $171 and $260 billion, depending on whether straight line or double declining balances were used in calculating depreciation (Anon., 1984a). To make matters worse, the unfounded assumption of a determinate economic lifetime can also distort measures of productivity (Enke, 1962; Miller, 1983 and 1985a). Considering that national economic policy is predicated upon this measure of the health of the national economy, reliance on this data is especially distressing. In the future, some of the problems resulting from reliance on data calculated on the basis of the perpetual inventory method may be gradually overcome. Since 1977, the Annual Survey of Manufactures of the US Department of Commerce has surveyed business concerning retirements of its capital stock (US Department of Commerce, 1987, p. 4). In 1988, the annual time series consisted of only nine observations, demonstrating a strong positive trend. In 1977, retirements equalled almost 50 per cent of new capital acquistions. _Table 3.1 _ _Retirements and expenditures on used capital ($ billion)_ _Year_ _Retirements_ _Expenditures on Used Capital_^1 _Billion Dollars_ _Billion Dollars_ 1985 41.7 8.2 1984 33.7 5.4 1983 33.2 5.6 1982 29.3 6.3 1981 30.0 5.1 1980 23.6 4.5 1979 18.3 3.4 1977 18.6 3.3 1976 15.0 n.a. ^1 including structures Source: US Department of Commerce, 1987 ______________________________________________________________________________ The figures shown in Table 3.1 are far from perfect. They depend upon voluntary information based on the book values of capital goods replaced. As information accumulates about this time series, more interest may develop in studying the scrapping decisions. Although considerable information is collected regarding aggregate gross investment, scrapping rarely leaves a satisfactory paper trail. One partial exception may be the stock of ships which are listed in Lloyd's Register (Parkinson, 1957). In the absence of any useful public data base, researchers depend on access to private records. Even if firms were required to give a public accounting of the retirement of all plant and equipment, substantial measurement problems would still remain since detailed studies of actual business replacement practices, such as Holland's account of the Bristol Bus Company are rare indeed (D.C. Holland, 1962). _Variations on a Theme by Terborgh_ Jorgenson claims, 'There is no greater gap between economic theory and economic practice than that which characterizes the literature on business investment in fixed capital' (Jorgenson, 1963a, p. 47). This admission is especially valuable since Jorgenson has been the most vocal advocate of theories designed to gloss over the complexities of investment (see below). On the simplest level, profitable investment in capital equipment requires that the present value of the investment exceed the costs. This far from trivial comparison occurs within a very complex calculus of risk considerations. In Keynes' words, business invests in capital goods 'in light of _current_ expectations of _prospective_ costs and sale proceeds' (Keynes, 1936, p. 47). These expectations include the first, second and perhaps higher moments of the expected future returns from an investment; that is, not just expected profits, but the expected variance and skewness of future profits. This elementary proposition has enormous consequences for investment theory. Over and above Keynes' suggestion that rational calculation was all but impossible because uncertainty was so pervasive that probability distributions could not be known, the difficulty of integrating the expected variance of future investments into capital theory is so daunting that it has been ignored in the economics literature. In contrast, finance theorists have analysed the expected variability of future returns since they are concerned with the remarkable array of financial instruments which business has invented to hedge against an increasingly uncertain environment. Just as the mix of financial assets changed as business attempted to cope with uncertainty, the mix of real capital assets purchased should be similarly affected by uncertainty. Let us take a leaf from the financial theorists. Terborgh suggested that purchasing capital goods is akin to participating in a futures market for capital services (Terborgh, 1949, p. 29). A hypothetical futures market for capital services differs from actual futures markets since spot markets for capital services do not generally exist. Futures markets typically develop for widely traded, homogeneous commodities. Most capital services, or even capital goods, are neither widely traded nor homogeneous. Existing futures markets are difficult enough to predict, but the hypothetical futures market for capital services would be even more complex. Broader spot markets for capital services could evolve if capital services were produced by capital goods that had a lifetime of only one period, making the spot market for capital services identical with the capital goods market. Indeed, if all capital goods only lasted one fixed period investment theory would be simple since all capital would be circulating capital. To make the concept of a spot market for capital services more general, consider a notional spot market for capital services measured by what firms would be willing to pay for a flow of capital services during a particular period. Under some sets of expectations, firms may want to lock themselves into a firm contract at existing prices. In Woodward's terminology, they would be willing to pay a solidity rather than a liquidity premium (Woodward, 1983). The existence conditions for a liquidity premium are well known ‑‑ the most important one being that present conditions convey significant information about the future (Day, 1986). In addition, the expected cost of producing capital services would have to grow more rapidly than the discount rate. Brownian motion with drift might reasonably describe the pattern for the value of capital services. It imples that the variance of expected future prices t‑periods ahead is proportional to the value of t. Consequently, the risk of investing in future capital services increases the more distant their delivery. In the absence of inflationary expectations, 'normal backwardation', would be expected to occur in this notional futures market because the owners of capital goods have to bear the risk of taking a long position on capital services (Keynes, 1923a). Some students of futures markets deny that normal backwardation is the rule. They argue that in a balanced commodity market, those ‑‑ such as millers or feed processors ‑‑ who intend to purchase commodities in the future may also want to protect themselves from risk. Their preference for a short hedge tends on average to balance out the suppliers' demand for a long hedge, thereby eliminating any tendency for normal backwardation. Hicks discounts the likelihood of balanced futures markets, referring to the 'congenital weakness of the demand side' (Hicks, 1946, p. 137n.). His case is even stronger for the futures market for capital services. According to Hicks: Now there are quite sufficient technical rigidities in the process of production to make it certain that a number of entrepreneurs will want to hedge their sales for this reason; supplies in the near future are largely governed by decisions taken in the past ... But although the same thing sometimes happens with planned purchases as well, it is almost inevitably rarer; technical conditions give the entrepreneur a much freer hand about the acquisition of inputs than about the completion of inputs ... If forward markets consisted entirely of hedgers, there would be a tendency for a relative weakness on the demand side; a smaller proportion of planned purchases than planned sales would be covered by forward contracts. (Hicks, 1946, p. 137) The absence of speculators who are willing to bear the risk of holding long‑lived capital goods contributes to firms' reluctance to invest in them. Think of a long‑lived capital good as an inventory of expected future capital services. The owner must generally bear the risk of holding the inventory. Some exceptions occur. Owners of buildings attempt to transfer risk to occupants by means of long‑term lease agreements (Keynes, 1936, p. 163). Unfortunately no equivalent exists for most capital goods. Alternatively a firm could contract to lease a capital good from the supplier. Ignoring risk considerations, a firm's profitability would be unaffected whether it would enter into an agreement to lease a capital good or simply borrow the funds to purchase it outright. In effect, the lessor would be lending money to the firm to finance the inventory of capital services. Of course, the motive to lease goes far beyond the extension of credit from the lessor. Ignoring the tax consequences, the lessor bears two kinds of risk. First, the lessor will suffer if the earning capacity of the capital good deteriorates. Second, in a straight lease agreement, the lessee has an incentive to overwork a leased capital good. The lessor has to shoulder the loss if the operator fails to maintain it in good working order (Rust, 1985, p. 590). When lease agreements are in force, lessors often tend to be especially suited to bear the risk associated with holding an inventory of capital services (Flath, 1980; see also Bulow, 1986). For example, '[t]he lessor may be active or skillful in dealing in the associated second‑hand market: his specialized knowledge may give him an edge' (Flath, 1980, citing Lewellen, Long and McConnel, 1976, p. 796). The continual threat of obsolescence causes firms to prefer to lease computers (Flath, 1980). Obviously computer manufacturers are better informed than their customers about the future likelihood of obsolescence. Because computers do not wear out from overworking as much as mechanical equipment, they are less subject to abuse. Thus computers are leased more frequently than most capital goods. With the exception of computers and transportation equipment, especially the sort that is subject to government mandated maintenance schedules, capital goods are not generally leased (B. Klein, Crawford and Alchian, 1978). Does the absence of leasing agreements mean that normal backwardation is inapplicable to the notional futures market for capital services? Probably not. What Hicks called the 'congenital weakness' in the demand side occurs because those who purchase capital goods do not have a strong need for a short hedge. Some of those who might require capital services in the future normally can squeeze an extra period of capital services out of old plant and equipment. Those that hold excessive inventories of capital services do not have the same degree of flexibility. As a result normal backwardation might be even more normal in this hypothetical futures market than in more familiar ones. Lessees are generally willing to pay a premium to lease a capital good in the same way Keynes suggested that long hedgers in futures markets, such as farmers, are willing to pay speculators the premium associated with normal backwardation to protect themselves from risk. Where lease arrangements are not available, the willingness to pay for the security of leases manifests itself as a reluctance to invest in long‑term capital. Imagine a Walrasian world without technical change. Each firm is endowed with machines of different durabilities and a cash balance. Firms trade until they are satisfied with their portfolio of money and machines. Then trading commences. If the producers believe that they are entering a stable, golden age with a commonly expected rate of interest, then the price of each capital good will equal the expected discounted value of the capital services that it will deliver. To eliminate the problem of discounting, the price of each capital good will be measured in terms of the annual payout of an annuity over the lifetime of the capital good, with an interest rate based on the commonly expected rate. If capital goods prices are expected to rise, then contango will exist. Firms will be willing to pay more for the certainty of a future flow of capital services than they would pay for immediate capital services (say, the value of an old capital good with only one more period of service). If firms have a high liquidity preference, they will hoard their money. Sellers of capital goods will have to lower their price to entice other firms to part with their funds. Future values of capital services, reflected in the values of newer capital goods, will be very low relative to the value of current capital services. -- Michael Perelman Economics Department California State University michael at ecst.csuchico.edu Chico, CA 95929 530-898-5321 fax 530-898-5901
This archive was generated by hypermail 2.1.5 : Fri Dec 24 2004 - 00:00:01 EST