Austrian Manufacturing Capital Productivity

The paper begins by establishing the critical need for accurate capital stock estimates in various economic analyses. These estimates are essential for calculating potential output, informing production and investment functions, and, most importantly, analyzing productivity changes, as highlighted by the OECD (2001a). The study focuses on disaggregated estimates of productive capital stock, crucial for understanding structural issues and the dynamics of productivity, investment, and employment at a granular level. The paper aims to provide disaggregated gross, net, and productive capital stock estimates for 20 Austrian manufacturing industries from 1969 to 1994. This specific period is chosen due to data availability constraints and a change in industry classification that prevents extension. The lack of published capital stock data at this level of disaggregation for this period further justifies the undertaking of this research. The paper distinguishes between the wealth concept of capital, suitable for certain purposes, and the need for a capital services measure for productivity analysis, introducing the concept of the volume index of capital services.

The study outlines two fundamental approaches to capital stock estimation: direct estimation for a benchmark year using insurance, book values, or direct data collection; and cumulative methods that involve aggregating historical investment series and deducting retired or written-off assets. The paper emphasizes the use of the Perpetual Inventory Method (PIM), widely adopted by OECD countries (OECD 2001a), which combines long-term gross domestic fixed capital formation estimates with assumptions regarding asset lives and retirements. The PIM assumes the capital stock at a given moment equals past investments still in production. The authors opt for the PIM due to its ease of implementation and cost-effectiveness compared to direct methods. The study explicitly states that the gross capital stock might be sourced from surveys, but the net capital stock always relies on a variant of the PIM, requiring consistent fixed capital consumption estimation. The paper acknowledges limitations of the PIM, including the reliance on assumptions about service lives, depreciation patterns, and initial capital stock, citing Ward (1976) and Mayes and Young (1994) on these drawbacks. The selection of retirement and depreciation functions is highlighted as a critical element impacting results. Furthermore, the paper clarifies the exclusion of intangible assets and natural resources, aligning with the 1993 SNA classification, and explains why inventories are excluded from the capital stock measure.

A significant portion of the methodology section delves into the complexities of depreciation and deterioration. The paper acknowledges the existing debate surrounding these concepts (OECD 2001a), distinguishing between the 'productivity school's' cross-sectional view of deterioration (Hill, 1999) and the inter-temporal System of National Accounts definition (Hill, 1999; Triplett, 1996). The authors discuss the intuitive appeal of geometric depreciation patterns for age-price profiles, referencing the use of hyperbolic decay by the BLS and the Australian Bureau of Statistics (OECD 2001a). However, they also present the arguments of Oulton and O’Mahony (1994) suggesting that, despite individual asset deterioration potentially following different patterns, aggregate depreciation can be approximated by geometric decay due to composition effects and premature scrapping (Hulten and Wykoff, 1981, 1996). The study further reinforces the use of geometric depreciation by citing Hulten and Wykoff's findings that a constant depreciation rate provides a reasonable statistical approximation to underlying Box-Cox rates (Hulten and Wykoff, 1981). The paper also demonstrates that, under geometric depreciation, the rate of deterioration equals the rate of depreciation (Jorgenson, 1996), linking the disaggregated net capital stock to both wealth (asset price aggregation) and capital services (rental price aggregation). The adoption of geometric depreciation in US national accounts (Fraumeni, 1997; Herman, 2000) is cited as further support for this approach. The use of industry-level depreciation rates from Timmer and O’Mahony (2003) is justified, acknowledging the assumption of similar depreciation rates and sectoral capital composition between Austria and the US. This choice is defended as a necessary compromise given limited reliable information on depreciation outside the US.

The research draws upon several key data sources. Investment data, crucial for the perpetual inventory method, comes from the ISIS-database. This database provided industry-wide nominal and real investment data series used to calculate industry-specific deflators, although these deflators were not industry-specific. For the period 1969-1975, a lack of separate time series for machinery and vehicles necessitated disaggregation of aggregated series using investment ratios from 1976-1986. The nominal data was converted to constant (1983) prices using appropriate deflators from the ISIS-database. The initial gross capital stock for 1969 was obtained from Hahn (1983). Data on tax rates, tax credits, commercial interest rates, and depreciation rates for tax purposes—necessary for computing the user cost of capital—was provided by Serguei Kaniovski from WIFO. Kaniovski (2002) used this data for Austrian manufacturing sector user cost estimations, which were compared to the estimations used in this study. The paper notes that the implied service lives from the Bureau of Economic Analysis (BEA) were shorter than those reported by Schenk and Fink (1976), leading to an adjustment of average service lives for structures and machinery, with the adjusted service lives reported in Table 1. These service lives were then used to calculate gross capital stocks, exploring the impact of service life assumptions on results. Further details are provided in the data appendix.

Industry-level depreciation rates for the three capital stock categories (buildings, machinery & equipment, vehicles) were primarily sourced from Timmer and O’Mahony (2003) (Table 2). The reliance on these rates involves the assumption that capital stock composition and deterioration are similar between Austria and the US. Given the strength of this assumption, and the empirical basis of service lives in Schenk and Fink (1976), the researchers adjusted depreciation rates for machinery & equipment in four industries where the implied service lives from Timmer and O’Mahony differed significantly from Schenk and Fink's estimates. Specifically, adjustments were made for the mining, food, metal (excluding steel), and clothing industries to bring the implied service lives closer to those of Schenk and Fink. The modified depreciation rates are detailed in Table 2. The paper highlights the calculation of the average weighted depreciation rate at the industry level using geometric depreciation rates, weighted by average annual investment in the three asset categories (1969-1974). A comparison with an average net-gross ratio of approximately 66%, as estimated by Böhm et al. (2002), is included in Table 3. The sensitivity analysis examined the impact of varying the age of the initial capital stock, specifically the average age of structures (baseline: 20.5 years, variations: 16.5 and 24.5 years), concluding that differences in the age of the initial capital stock have less of an effect on the medium-run development of gross capital stock.

Estimating the initial capital stock, representing accumulated past investments, presented challenges due to the unknown vintage and composition. The paper employs the assumption of equal shares of past investment for the initial capital stock, with the 1969 initial gross capital stock obtained from Hahn (1983). Since this data was aggregated, the authors had to disaggregate it into three capital categories (buildings, machinery, vehicles) at the industry level. They used the end values of the calculated net capital stocks (excluding initial capital stock) to derive these disaggregated estimates. This approach acknowledges a potential bias stemming from the strong assumption that investment shares across capital categories remained relatively stable over time. The influence of the initial net capital stock on net capital stock development was examined by varying the initial net capital stock by ±5% (Figure 4), demonstrating that variations in the initial net capital stock have only a temporary effect, with any errors reducing over time. In contrast, the impact of alternative depreciation rates (and implicitly, service lives) for machinery on net capital stock development showed a larger, growing impact over time (Figure 3), reinforcing the importance of the chosen depreciation rates and their underlying assumptions.

The paper extends beyond traditional capital stock measures (gross and net) by calculating a volume index of capital services. This index, derived from the work of Jorgenson and Griliches (1967) and detailed in Jorgenson (1989), quantifies the flow of services from capital assets. Unlike a wealth-based capital stock measure, where assets are weighted by their price, the volume index weights assets by their price multiplied by their rental price. This approach gives more weight to assets with high rental prices relative to asset prices, better reflecting their contribution to production. The methodology acknowledges the heterogeneity of assets, following Jorgenson, Gollop, and Fraumeni (1987), by defining capital service flows individually for each asset type and aggregating them using asset-specific user cost shares as weights. The rationale is that user cost shares approximate the relative marginal productivity of assets, thus incorporating the varying productive contributions of heterogeneous capital assets into an overall capital input measure. The calculation of rental prices (user costs of capital) considers tax rates and tax credits, and uses the market interest rate as the rate of return due to unavailability of profit income data at the industry level. The resulting estimated volume indices of capital services for the industries are reported in Table 11.

The research uses the calculated capital services and capital stock estimates to analyze productivity and technological change in Austrian manufacturing. The study links technological change to shifts in the production function due to new techniques, noting that these shifts can be neutral or affect input-output relationships. Neutrality is assessed using measures such as capital-output ratio (Harrod-neutrality), capital-labor ratio (Hicks-neutrality), and labor-output ratio (Solow-neutrality) (Nadiri, 1982). The researchers aim to examine patterns of technological change to determine whether it exhibits a Marxian bias (labor-saving and capital-using). The paper contrasts capital productivity calculated using capital services and gross capital stock, highlighting that using gross capital stock yields lower capital productivity growth because gross capital stock growth consistently exceeds capital services growth for all industries. The analysis uses this comparison to assess the presence or absence of a Marxian bias, comparing the results to those found by Schreyer and Pilat (2001). The study notes that with a Leontief production function, growth rates in capital and labor productivity directly reflect factor augmentation parameters (Foley and Michl, 1999; Nadiri, 1982), but a neoclassical analysis would require additional consideration of capital-labor substitutability. The analysis of capital-labor and capital-output ratios (Figure 8) further contributes to the understanding of technological change, revealing differences in capital-labor ratio growth across industries.

The results section begins by comparing the calculated gross capital stocks with estimates from Hahn (1983) and the WIFO database for the period 1969-1981 (Table 4). The comparison focuses on growth rates, revealing that the study's baseline estimates are generally close to those of Hahn (1983) and WIFO, except for industries 14, 17, and 21. While some discrepancies exist (some estimates being higher, some lower), the authors conclude there's no systematic error in their calculations, attributing the differences to variations in initial capital stock estimates and composition. The growth rate of the modified gross capital stock is consistently lower than WIFO and Hahn (1983) estimates, except for three instances. Three specific industries (leather manufacturing, foundries, and clothing) showed negative gross capital growth. The authors characterize this modified gross capital stock as a conservative estimate. Tables 8 and 9 in the appendix provide the detailed gross capital stock estimates.

The research analyzes the relationship between net and gross capital stocks, focusing on their ratio's evolution over time and variations across industries. The study observes that the net-to-gross capital stock ratio remains relatively stable for the mining industry between 1969 and 1994 but increases for other industries. This pattern is attributed to the depreciation of new investments in structures (and machinery) in the net capital stock calculation, while no retirement is reflected in the gross capital stock. The impact of the reduced average service lives, used in the calculation, is evident in the increasing divergence between gross capital stock estimates calculated using the original service lives and those with shortened service lives. This analysis highlights the effects of assumptions made about depreciation rates and service lives on the final estimations.

The final part of the results section addresses technological change and its impact on productivity. The study uses a production function perspective, linking technical change to shifts in the production function due to adopting new techniques. The analysis considers the effects of technological change on capital-output ratios (Harrod-neutrality), capital-labor ratios (Hicks-neutrality), and labor-output ratios (Solow-neutrality), as discussed by Nadiri (1982). The researchers investigated whether a Marxian bias (labor-saving, capital-using technological change) was present. A comparison of capital productivity growth based on capital services versus gross capital stock shows a significant difference, with gross capital stock underestimating capital productivity growth. The analysis of capital productivity growth based on capital services reveals that there is no overall Marxian bias across all the industries and that most experienced increases in capital productivity. A scatter diagram (Figure 7) helps visualize the relationship between labor and capital productivity growth rates, using both capital services and gross capital stock as measures of capital. Only 5 industries displayed negative average capital productivity growth, potentially indicative of a Marxian bias. The textile industry is identified as an example of Harrod-neutral technical change, showing positive labor productivity growth and a constant capital-output ratio. The study concludes that all industries experienced Hicks-labor-saving technical change, with significant variations in the growth rates of capital-labor ratios across industries (Figure 8).