Measuring Irish Capital Stock

The paper highlights the absence of information on capital stocks and capital services in Ireland's National Income and Expenditure Accounts (NIE). Accurate estimates of these macroeconomic aggregates are crucial for various modeling exercises. The study aims to provide a detailed, asset-level analysis of Irish fixed assets, their depreciation, and the capital formation flows that determine them. It intends to achieve this through an improved perpetual inventory methodology (PIM) based on US National Income and Product Accounts (NIPA) best practices. The research will demonstrate the relationship between net capital stock, consumption of fixed capital, and capital services using a standard asset valuation equation. Finally, it will present capital service estimates for the Irish economy as a whole and by asset type, emphasizing the importance of the Volume Index of Capital Services (VICS) in accounting for the relative productivity of different assets within the capital stock. Failing to control for this, especially the share of housing, can lead to overestimation of total factor productivity in growth accounting.

Several methods exist for measuring the value of gross capital stock, including an econometric approach based on the production function, the widely used perpetual inventory method (PIM), direct surveys, and a 'balance of fixed assets' method. The PIM, pioneered by Goldsmith, involves accumulating past capital formation and deducting the value of assets reaching the end of their productive lives. A direct survey approach uses exact records of all assets still in use, while the 'balance of fixed assets' method, considered an ideal PIM version, maintains a current quantity record of different asset vintages. This paper adapts a model-based perpetual inventory methodology, aiming to replicate the comprehensive asset vintage information of the survey approach but using data from non-Irish surveys on depreciation and service life parameters, applied at the detailed asset level. Pricing is done using detailed price indices, leveraging Irish National Accounts' asset-type investment cost deflators.

The traditional PIM involves accumulating investment series, which this research expands upon by adding supplementary information. A depreciation function is applied to capture the productive efficiency and age-price profile of assets, along with an asset mortality function to model realistic asset retirement patterns. The PIM differentiates between gross and net capital stock measurements. Key considerations include asset survival rates (also known as relative efficiencies), which represent an asset's life pattern and its productive efficiency over time. Different age-efficiency profiles are discussed, highlighting that the geometric depreciation method best aligns value and efficiency declines over time. The method's weaknesses are identified as data input deficiencies and limitations in its underlying assumptions. Specifically, the lack of Irish data on used asset transactions causes distortions because capital acquisition figures must represent net additions to the capital stock, particularly impacting the government sector. Price differentials between new and second-hand assets are largely ignored, but the use of detailed asset-level gross fixed capital formation data mitigates some of these data limitations.

The paper discusses the complexities and uncertainties in applying the PIM, including the lack of precise information on asset lives and the ongoing debate about what constitutes 'keeping the capital stock intact'. The 1993 System of National Accounts (SNA) recommends a revaluation account to record changes in asset value due solely to holding them over time. This study avoids making estimates of net appreciation or after-purchase revaluation, although this is acknowledged as potentially significant in the Irish context, where housing investment makes up a considerable portion of the capital stock. The study's use of investment price deflators, rather than market indices, is highlighted to address this. Starting values for the PIM, mainly drawn from Henry (1989), present challenges due to data limitations. Left-censoring of capital formation data underestimates the initial capital stock, especially for assets that became prominent after 1980. A steady-state growth approach is used to estimate starting values where Henry's data is insufficient or unreliable. The selection of asset service lives from non-Irish sources (primarily US Bureau of Economic Analysis) is justified, noting that existing methods are subject to substitution and quality biases. The paper mentions adjustments made to Henry's (1989) starting values after identifying over- and under-valuations, and the impact of scaling to match Central Statistics Office (CSO) control totals.

The primary data source for this research is Ireland's National Income and Expenditure Accounts (NIE). The study covers the period from 1980 to 2004, although the most detailed asset-type and sector-specific data is available only from 1998 onwards. This time frame allows for analysis of significant economic shifts, including the 'Celtic Tiger' era. The choice of data sources reflects the limitations of available data. The use of NIE data is justified by its accessibility and established role in Irish economic reporting, although it has limitations as noted below. The Central Statistics Office (CSO) National Accounts provide control totals for Gross Fixed Capital Formation (GFCF) at the whole-economy level, allowing for reconciliation and scaling of data from other sources like the Census of Industrial Production and Public Capital Programme.

A major challenge is the lack of comprehensive Irish data on transactions involving used assets. This absence necessitates reliance on capital acquisition figures to represent net additions to the capital stock, which introduces inaccuracies, especially for the government sector, where capital information is limited. Furthermore, the price differentials between new and second-hand fixed assets are not consistently captured in price indices, potentially distorting the results. The research attempts to mitigate these weaknesses by using highly disaggregated gross fixed capital formation data at the asset level instead of relying on whole-economy aggregates. The need to estimate starting values for the capital stock series before the beginning of the available investment data (1980) is addressed. Data from Henry (1989) are used where possible, but require adjustments and scaling. The limitations of Henry’s (1989) data are highlighted, and a steady-state growth approach is adopted for assets not adequately represented in Henry’s work. This is particularly important because of the left-censoring of data and the emergence of new asset categories post-1980.

The study recognizes that the ideal economic environment for accurate capital asset measurement does not exist. Three factors are highlighted: economic downturns causing asset liquidations and premature scrappage, rapid technological change leading to unexpected asset obsolescence, and government depreciation allowances impacting the useful life of assets for tax purposes. These factors cannot be directly addressed, but the use of an asset life distribution in the model intends to reduce the impact of these indirect economic effects. The ongoing debate regarding the interpretation of 'keeping the capital stock intact' is discussed. This involves whether positive changes in asset value should be considered a net deduction from capital consumption, as seen in the different interpretations by Hayek, Pigou, and Hicks. The study aligns with the PIM convention by deflating asset series according to investment (cost) price deflators, as opposed to market indices, specifically acknowledging the importance of this distinction for housing assets. The data utilized includes current price data assembled for 1970-2004, with sector subdivisions made for households, firms, and the government sector. Specific data tables from the NIE are cited for GFCF, sectoral data, household consumption, and government gross physical capital formation.

A primary challenge is the lack of detailed Irish data on asset lives. The study draws upon US Bureau of Economic Analysis (BEA) data on retirement, efficiency, and service life profiles for detailed asset types to address this gap. These BEA estimates, though based on cross-sectional studies of used asset prices, haven't been adjusted for quality changes. The paper emphasizes the sensitivity of results to assumptions about asset life and discusses the impact of errors in average service lives. Sensitivity analyses are suggested using alternative service life estimates from other statistical agencies, such as the UK Office for National Statistics (UK ONS), whose approach to housing depreciation is noted (approx. 100-year asset life). The study adopts a 5-year life-length for computers and software based on a UK ONS review. The limitations of using investment price indices in the NIE accounts are acknowledged, specifically their susceptibility to substitution and quality biases, as noted by Hwang (1996). The use of a long-run constant price time series to mitigate this is justified and the intention to incorporate chain-linked indices in future work is mentioned. This highlights the ongoing methodological challenges in capturing the true rate of economic depreciation and the impacts of technological change.

The research presents estimates of net capital stock for the Irish economy, broken down by sector (households, firms, government) and asset type. Figure 1a illustrates the net capital stock held by households from 1980 to 2004, showing a marked increase in housing investment from 1998 onward, with new dwellings surpassing home improvements. This distinction is important due to differences in depreciation rates and asset lifespans. Investment in transport equipment significantly increased from 1996, peaking in 2000. The stock of consumer durables also shows steady growth, accelerating from the mid-1990s. Figure 1b, focusing on the industrial sector, reveals a significant reduction in transport equipment between 1990 and 1998, despite relatively stable net acquisitions reported by the Census of Industrial Production. This highlights the importance of considering depreciation rates alongside investment figures. The study relies on data from Henry (1989) for starting values of the capital stock, but these values underwent adjustments and scaling to align with the methodology recommended by Vaughan (1980) and to match CSO control totals. The sensitivity of these initial values and subsequent calculations to choices about the price deflators is acknowledged.

The analysis shows that depreciation rates vary significantly across asset types, reflecting differences in asset lifespans and usage patterns. Housing depreciation rates remain relatively stable at 2-3 percent. However, the recent boom in the construction sector contrasts with the rate of consumption of non-housing capital in construction; the rapid increase in the capital stock has led to a significant decline in the average depreciation rate for non-housing construction assets, approaching zero by 2004. This dynamic relationship underscores the importance of considering the timing and pace of investment when calculating depreciation. The study acknowledges that the application of the perpetual inventory method is limited to estimating the capital stock at constant (rather than current) prices because of the absence of detailed Irish survey data about asset age-efficiency and age-price profiles. Using investment vintages helps account for the age profile of capital services and produces stock estimates in standard efficiency units. This approach combines average asset service lives, considers asset mortality and retirement, and integrates a detailed age-efficiency profile with a corresponding age-price profile.

The research proceeds to estimate capital services, crucial for assessing productivity and capacity utilization. The paper uses a volume index of capital services (VICS) which addresses the limitations of using simple net or gross capital stock measures by incorporating asset-specific rental rates as weights. This approach accounts for the fact that assets with shorter lifespans must generate a faster return on investment. The VICS, thus, gives higher weight to assets with high rental prices (short-lived assets with rapid depreciation). The study’s approach mirrors methods developed by Jorgenson and Griliches & Jorgenson to account for the heterogeneity of assets. Asset-specific rental rates are used, reflecting marginal productivity under competitive market conditions. This calculation requires data on net capital stock by asset type. The resulting VICS measure is compared with a wealth-based measure of capital stock highlighting the differences between productivity-based and value-based aggregations of capital. The analysis reveals a decline in capital services during the early 1980s (possibly linked to the second oil crisis), followed by accelerated growth during the 'Celtic Tiger' era and a subsequent dip potentially due to the 'dot com' bust. The study then compares the VICS across sectors, illustrating the higher levels of capital service flows in the industrial sector. This is largely due to the higher concentration of high-rental rate assets like machinery and equipment, and software. This contrasts with the household sector's significant expansion in new private dwellings which have a lower rental rate and therefore a lower weight in terms of capital services.

The study compares the VICS measure with a wealth-based measure of capital stock. Both are weighted averages of asset stock growth rates but differ in their weights. The wealth measure uses the share of total wealth as its weight, thus depending on asset prices, while the VICS relies on rental rates reflecting marginal productivity of capital. The ratio of the rental price to the asset price differs across asset types, being lower for assets with long service lives and increasing capital gains. Table 7, illustrating the vast difference between rental and asset price weights for various assets, directly supports the choice of the productivity-based VICS measure over a wealth measure. This is particularly relevant for housing given its significant share in the Irish capital stock. Table 8, comparing the growth rates of both measures, confirms that during 1990-2003, the VICS grew faster than the wealth measure, demonstrating the importance of using the VICS for growth accounting, which weighs asset contributions according to their volume share and productivity. The VICS provides a more accurate measure of capital's contribution to economic growth, showing the overall positive contribution of productive assets despite the housing boom.

The conclusion emphasizes the superiority of the Volume Index of Capital Services (VICS) over simple capital stock measures (net or gross) for growth accounting. The VICS, by weighting asset contributions according to their volume share and contribution to total capital effort, provides a more accurate reflection of capital's role in economic growth. It highlights that the growth of the VICS is a weighted average of growth rates in individual asset stocks, where weights are the shares of the value of total capital services; this contrasts with wealth measures which use market values as weights. The study's findings demonstrate that while housing investment was substantial, highly productive assets such as machinery, equipment, and software also grew rapidly, contributing significantly to the overall increase in Irish capital services. The VICS's growth consistently outpaced a wealth-based measure of capital stock between 1990 and 2003, reinforcing its value as a more efficient measure of capital's contribution to economic growth. The study acknowledges that its dataset, while consistent, relies on assumptions which might be termed 'heroic' due to the scarcity of detailed Irish data, but nevertheless represents a significant improvement over prior estimates.

The implications for capacity utilization measurement are less straightforward. The similar growth rates of capital services and capital stocks until recently suggest that the choice of capital measure might not significantly affect capacity utilization estimates. However, a re-examination of the 1980s data, or a continuation of the trend observed in 2004 (where the wealth-based measure surpassed the capital services measure), could lead to a reevaluation of capacity utilization growth. The study points out that the choice of capital measure has implications for macroeconomic modeling, and the detailed findings warrant further careful analysis to fully assess the impact of this work on capacity utilization estimations, as well as other relevant macroeconomic modeling exercises. This highlights the ongoing need for further research and deeper analysis.