Published papers

This chapter reviews the history of policymaking on low-carbon development and decarbonization, examining distortions in economic and technological assessments and the uncertainty that remains in their scientific basis. Under the “two uncertainties” of climate change and the energy transition, it argues for the importance of not betting on any single path and of keeping diverse options open.

Using SBI-BIP, this chapter portrays the Japanese economy of 2040 along two scenarios: a Business-as-Usual (BaU) path extending current trends, and a Technology and Societal Implementation (TSI) path in which the societal implementation of technological innovation advances. It analyzes the sources of growth under TSI—where nominal GDP reaches approximately ¥1,000 trillion—and its consequences for employment, income distribution, AI utilization, and energy constraints, showing that realizing this potential depends not on technological progress alone but also on the economic and social foundations that support its implementation.

This paper examines the rebuilding of Kourakuen’s Asaka Store, a ramen restaurant in Fukushima Prefecture, Japan, as a ZEB Ready building, as reported in July 2026. Using publicly available primary sources—including real estate registration records, building confirmation documents, municipal budgets, and corporate disclosures—it reconstructs the conditions under which this case came to circulate as a decarbonization policy “achievement.” Neither national nor Fukushima Prefectural subsidies covered a small, newly constructed ZEB Ready building of this type, and the Koriyama City subsidy is unlikely to have induced the initial investment decision. The road-widening project explains the opportunity and timing of the rebuilding, but not the adoption of the ZEB Ready specification. The project combined a small scale, which could limit total incremental costs, with the publicity, learning, and rollout value associated with a store in a restaurant chain. For Tohoku Electric Power, the expected value of early market entry, learning, and the formation of a first-project track record may also justify its involvement as the ZEB Planner. Yet land and building ownership, construction, operation, and ZEB support were divided among multiple parties, while cost sharing, contractual relationships, and service remuneration were not disclosed. The economic arrangement that satisfied the parties’ participation constraints therefore cannot be identified. This non-identification arises from institutional design. The ZEB system records compliance with a design-based technical standard within a prescribed assessment boundary, but does not link outcome information to operating performance, financing and contractual relationships, counterfactual benchmarks, or cost incidence. Actual energy savings, private viability under ordinary market conditions, policy additionality, and replicability across projects therefore cannot be verified. At the institutional seam created by this informational disconnect, a “decarbonization achievement” is generated through two stages of selection. In the inner selection, project-specific opportunities and benefits allow the relevant parties’ participation constraints to be satisfied, thereby forming a ZEB Ready project. In the outer selection, projects with symbolic attributes—such as first-of-its-kind status or regional collaboration—are made visible through corporate announcements and media coverage. Narratives emphasizing technical achievement and inter-organizational collaboration while omitting project-specific formation histories may thereby bias public perceptions of policy effects. Appropriate policy evaluation requires examining the incentives that outcome indicators and subsidy-eligibility categories create for project and technology choices, while establishing a recordkeeping system that links project formation histories, public support, financing sources, related-party relationships, and post-completion operating performance to building-level outcome information. Additional resources committed by private actors in response to policy incentives should also be recognized as policy costs arising outside the fiscal process, and their incidence should be assessed.

Labor productivity growth in Singapore that has grown at a rate of over 3.0 percent per year since 1970s considerably slowed down to 0.5 percent on average per annum in the latter half of the 2000s. The purpose of this paper is to ask, first, to what extent Singapore's labor productivity performance is explained by the changes in the characteristics composition of its workforce, and then, what the prospect may hold. Using our newly constructed cross-classified labor dataset, we estimate a volume index of quality-adjusted labor input and an aggregate measure of labor quality change in the Singaporean economy for the period of 1974–2011. Having understood the current dynamics of the workforce, we further project labor productivity and potential economic growth for the coming two decades. In this paper, workers are distinguished by the five characteristic dimensions: gender, educational attainment, age, employment status, and residency. Our findings establish the role of labor quality changes in Singapore's economic growth as highly significant over the long run. During the period of 1974–2011, labor quality improved at a rate of 2.19 percent on average per annum, accounting for 37 percent of labor input growth to the 6.78 percent average yearly economic growth. Our estimates of recent labor quality growth, however, are considerably lower than what have been shown in some previous studies. Moreover, our projections suggest that its prospect in the foreseeable future remains bleak. The downward trend of labor quality growth since the mid-2000s is mainly due to the sharp increase in the number of low-skilled foreign workers. For the next two decades, our projections in the business-as-usual scenario imply a further decrease of labor quality growth. Consequently labor productivity growth will also slow down, from 2.04 percent on average per year in the 2000s, to 1.68 percent and 1.19 per cent in the 2010s and 2020s respectively. Coupled with the downward trend in hours worked, potential GDP growth is projected as 3.10 percent on average per annum in the 2010s and 1.86 percent in the 2020s. Compared with the past experience of 5.47 percent in the 2000s, this represents a considerable slowdown in Singapore's economic growth for the next two decades, if there is no appropriate policy response or boost to TFP growth. Policies, which successfully upgrade the resident workers' skills and/or induce the substitution of IT capital investment for the low-skilled non-resident workers, will lift the projection of labor productivity growth to 2.64 percent and 1.82 percent on average per annum in the 2010s and the 2020s respectively. However, the projected growth rate of potential GDP is little improved. At 3.22 percent in the 2010s and 2.06 percent in the 2020s, our projections even with successful policies are closer to the lower end of the target range set by the Economic Strategies Committee of Singapore's government for the 2010s. We conclude that the room to boost growth by improving labor quality is very limited, and policies targeting TFP growth may be more fruitful.

We analyze the widely used methodology of international price harmonization by comparing the sources of IT price gaps between the U.S. and Japan at the SIC 3-, 4-, 5-digit level. By careful examination of the most detailed published data and methodology used in both countries, we can conclude; (i) At the SIC 5- and 4-digit level, we find almost no differences in computer output prices during 1995-2003. (ii)At the SIC 3-digit level, Computer & Peripheral Equipment prices fall almost twice as fast in the U.S. as in Japan during 1995-2001. Over the long-term, the price gap is similar. This gap mainly results from the higher production weight of the peripherals, which have relatively moderate price declines in Japan. (iii)Moving to the other IT goods, Electronic Component prices fall more rapidly in Japan before 1990, but faster in the U.S. after 1990. Communication Equipment prices fall faster in Japan. Price gaps for IT goods are unique and should be considered separately from the Computers & Peripherals price gap.

This paper examines international energy cost competitiveness using real price conditions rather than exchange-rate-based nominal comparisons. Focusing on nine major economies—Japan, China, Korea, India, the UK, Germany, France, Italy, and the U.S.—we construct a Real Price Level Index for energy (Real PLI) based on value-added prices using the ECM Database and extend the measurement through the fourth quarter of 2025. The results indicate that, in the post-pandemic period, real energy price differentials have persistently widened in Japan, Korea, and several European economies relative to the U.S. and to China and India. We further investigate the relationship between Real PLI and the output of energy-intensive and trade-exposed (EITE) industries. Panel threshold regressions reveal significant nonlinear effects: once Real PLI exceeds estimated critical levels, EITE production declines sharply. In contrast, no statistically significant threshold effect is detected for China and India. The paper also shows that recent improvements in Gross Energy Productivity (GEP) and the apparent stabilization of Real Unit Energy Costs (RUEC) in advanced economies partly reflect structural shifts—specifically, the relative contraction of EITE sectors—rather than purely technological efficiency gains. While such structural adjustment may reduce macroeconomic sensitivity to energy price shocks, it simultaneously implies declining domestic basic-material production capacity and greater external dependence. These findings suggest that persistently adverse real energy price conditions act as structural constraints on industrial location and investment decisions.

This paper develops a high-resolution economic model, BIP, designed to evaluate the macroeconomic impacts of the social implementation of technological innovation in Japan. The BIP framework visualizes how technological innovations propagate across economic activities through structural transformation and aims to capture multiple dimensions of the economy—production, employment, income distribution, and energy consumption—within a consistent general equilibrium framework. Two scenarios are examined. The first is a business-as-usual (BaU) trajectory that extends the stagnation observed in the Japanese economy, while the second is a technology and social implementation (TSI) scenario in which technological innovations are widely adopted across society. The analysis traces the divergence between these scenarios toward 2040. Under the TSI scenario, the service sector expands significantly; however, the core growth mechanism remains the reconstruction of an industrial foundation centered on manufacturing. Relative to the BaU baseline, nominal value added increases by ¥257 trillion, of which ¥132 trillion is attributable to productivity improvements and market expansion associated with the adoption of artificial intelligence. These productivity gains simultaneously enable shorter working hours and rising wages, while also promoting domestic capital accumulation supported by affordable and stable electricity supply. The results suggest that technological innovation of this kind could fundamentally alter the long-term growth trajectory of the Japanese economy.

This paper presents new data on the sources of growth for the Japanese economy over the period 1960–2000. The principal innovation is the incorporation of detailed information for individual industries, including those involved in the production of computers, communications equipment, and electronic components as information technology equipment. We show that economic growth is dominated by investments and productivity growth in information technology, both for individual industries and the economy as a whole. We also show that the revival of total factor productivity growth accounts for the modest resurgence of the Japanese economy since 1995.

This paper constructs a multi-sector model to take explicit account of the very sharp change in the relative price between non-IT and IT goods. The model is calibrated to the Japanese economy, and its solution path from 1990 on is compared to Japan's macroeconomic performance in the 1990s. Compared to the one-sector analysis of Japan in the 1990s in Hayashi and Prescott (2002), our model does slightly better or just as well in accounting for Japan's output slump and does worse in accounting for the capital-output ratio. We also show that, to revive a 2% long-term growth in percapita GDP, Japan needs to direct 10% of private total hours to the IT sector.

This paper constructs a measurement framework for commodity-specific relative price ratios consistent with the 1990 Japan-US International Input-Output Table (164-sector classification), with the aim of enabling real comparisons of production structure and cost structure between Japan and the United States, and reports the resulting estimates. Existing PPP surveys by OECD-Eurostat and the ICP focus on purchasers' price differentials for final consumption goods and are not suitable for use as industrial input prices for intermediate goods. The earlier estimates produced by the joint project of MITI and Keio University (the MITI-IRI estimates) also suffered from several unresolved problems: information loss in the conversion from expenditure classification to input-output commodity classification, conflation of domestic producer prices with composite prices that include imported goods, and failure to account for price differentials in the transportation and trade sectors themselves. The framework developed in this paper is grounded in a nominal balance system for non-competitive and competitive import-type input-output tables that explicitly describes the bilateral trade structure between Japan and the United States. Within this framework, domestic production relative price ratios (at producers' prices, separately for household and industrial demand), domestic demand relative price ratios (composite goods prices), and domestic demand relative price ratios at purchasers' prices — including commodity-specific PPPs — are all defined in a mutually consistent manner. By defining commodity-specific PPPs starting from composite prices that include imported goods, all relative price ratios are derived coherently under the interdependence arising from bilateral commodity-specific import shares, tariff rates, and international transport margin rates. Domestic transport and trade margin rates are computed from Japan's 1990 input-output table and the US 1987 input-output table for all 164 commodity sectors, covering five transport margin categories (rail, road, water, air, and other) and two trade margin categories (wholesale and retail), each disaggregated by household and industrial demand destination. For evaluating the validity of the estimates, a statistical hypothesis testing criterion is constructed based on whether deflating the input-output tables using the estimated relative price ratios reduces the gap between Japanese and US real input coefficients relative to the gap observed in nominal input coefficients. This criterion is used both to select among multiple relative price estimates derived from different data sources for the same commodity and to detect and exclude outliers. Final estimates of domestic production relative price ratios at producers' prices (for both industrial and household demand) and domestic demand relative price ratios at purchasers' prices are provided for all 164 commodity sectors.

This paper examines the theoretical and empirical problems in existing capital stock estimates for Japan, and reports on the time-series estimation of a private/public fixed capital stock matrix (1955–1992) constructed as part of the KEO Database (KDB). The paper first establishes the conceptual framework for measuring capital service input in production theory, formalizing the Age-Efficiency Profile and its dual counterpart, the Age-Price Profile for economic depreciation. It demonstrates that adopting the Best Geometric Approach (BGA) — assuming a geometric depreciation pattern — is particularly useful in practice, as it renders the perpetual inventory method independent of the age composition of the capital stock. Five problems are identified in the Economic Planning Agency's "Private Enterprise Capital Stock Estimates": (1) the use of a gross capital concept that inadequately accounts for physical deterioration; (2) aggregation errors arising from the lack of capital goods disaggregation — in our estimates, the capital growth rate for 1960–65 is revised upward by approximately 6.7 percentage points relative to simple-sum aggregation; (3) lack of transparency in deflators due to publication of real series only; (4) absence of reproducibility in the estimation procedure; and (5) insufficient coverage of the estimation scope. The time-series fixed capital formation matrix is constructed to be consistent with the commodity classification of the input-output tables (78 capital goods categories and 43 industries), and is aligned with SNA gross fixed capital formation through adjustments for scrap generation. For depreciation rate estimation, the paper combines a quantity-based approach using the polynomial-benchmark method — with the 1955 and 1970 National Wealth Surveys as double benchmarks — with price-based empirical analysis. For private housing, a Box-Cox transformation model is applied to prefectural age-stratified rental data from the Housing Survey, yielding depreciation rates of 4.4–5.9% for wooden structures and 3.2–3.8% for non-wooden structures after correcting for censored sample bias. For used automobiles, Age-Price Profiles are estimated from used car price data using a hedonic approach, with economic depreciation rates (under the geometric approximation) ranging from approximately 18–45% depending on vehicle type. The results also reveal a consistent declining trend in depreciation rates over time — implying lengthening service lives — across all vehicle categories examined. These findings broadly support the BGA as a useful approximation, while highlighting the need to incorporate time-varying depreciation rates in future work. The paper concludes by identifying remaining challenges, including reconciling conceptual differences between input-output tables and the SNA, improving the granularity of capital goods composition data, developing dynamic depreciation estimates that reflect rapid technological change, and extending the framework to cover intangible fixed assets.

This paper examines conceptual and measurement issues arising from the treatment of land in the current System of National Accounts (1993 SNA) and proposes directions for extending the framework to consistently incorporate non-produced assets as factors of production. In the production account, the SNA does not recognize the rental of land as a production activity and classifies land rent as property income. This treatment generates an inconsistency between the ownership-based recording of land assets and the user-based allocation of rent in input–output tables. The paper demonstrates that this inconsistency introduces a systematic bias in the measurement of total factor productivity (TFP), leading to an underestimation of approximately 1.5 percentage points per year during Japan’s high-growth period. From the perspective of asset measurement, the paper distinguishes two alternative definitions for separating the quantity and price of land: one that incorporates externalities into quality-adjusted prices, and another that attributes them to quality-adjusted quantities. It argues that the latter is more consistent with the treatment of fixed assets in productivity analysis. The paper also highlights that the treatment of land improvements in the SNA obscures the distinction between produced and non-produced assets and supports the direction of separation adopted in the 2008 SNA revision. In the stock account, it is shown that even when the total land area remains constant, changes in land attributes across regions and uses can generate variations in a Divisia-based quantity index of land. This finding underscores the need to identify, within non-produced assets, the components that function as factors of production. Based on these analyses, the paper concludes that two revisions should be considered in the national accounting framework: (i) extending the production boundary to treat the rental of non-produced assets as production, and (ii) explicitly identifying the portion of non-produced asset stocks that serve as productive inputs.

Comprehensive investigation of asset discard patterns has not been sufficiently implemented in Japan, and the methodology for capital stock measurement in the Japanese national accounts still relies largely on parameters derived from the 1970 National Wealth Survey. As the first step toward reframing this measurement, we examine actual survival distributions and asset service lives using new micro data for directly observed discards from the Survey of Actual Capital Stock and Discard of Private Enterprises (SASD), newly implemented by the Economic and Social Research Institute (ESRI), Cabinet Office, in 2003. We begin by clarifying three distinct capital stock concepts — gross, productive, and net — and the distinction between discard and decay in the context of capital measurement theory. We then introduce the Weibull family of distributions as a flexible framework for approximating discard patterns, and apply it to 66 asset categories covered in the 2002-SASD. Three definitions of actual age-survival distributions are considered, differing in the weighting scheme applied to the sample observations. We also discuss sampling bias inherent in single-period surveys and propose an adjustment procedure based on investment quantity indices. The estimated average service lives vary considerably across asset types, ranging from around 6–9 years for IT-related assets such as personal computers and mainframes to over 20 years for buildings and construction. The shape parameter estimates indicate that almost half of the assets exhibit a progressively increasing hazard rate, suggesting that the probability of retirement rises with asset age — a finding that contrasts with results reported for the Netherlands. To validate our estimates, we compare them against registration data for motor vehicles, and find broadly consistent results. We conclude with proposals to improve the questionnaire design of the SASD for future investigations, including the identification of whether retired assets were originally purchased new or as second-hand goods, the collection of sold asset prices, and the addition of coverage for intangible assets.

In this paper, we measure own-account software investment in Japan as the applications of the OECD Task Force recommendation at the aggregate level and the BEA's methodology at the industry level. We can conclude that the scale of own-account software investment in Japan is 0.60 percent of GDP in 2000. This share is 0.13 percent point lower than that in the U.S. The share of total software investment to GDP is 2.03 percent, which is the almost same as that in the U.S. (2.07 percent), reflecting the larger share of custom software in Japan relative to other countries. By type of software, in 1970, own-account software has the largest share in software investment and prepackaged software is minor in the U.S. and Japan. The composition is very similar between the two countries. In the U.S., the diminution of the share of own-account software is reflected by the rapid expansion of prepackaged software in the 1970s and the 1980s. On the other hand, in Japan, the diminution is mainly reflected by the expansion by custom software. To measure software capital stock, we consider four scenarios. First, for depreciation, the 33 percent and 55 percent geometric depreciation rates are assumed. Second, there are two options for prices, a cost index for all types of software and harmonized indexes for each type of software. When we use 33 percent depreciation rates and the cost index, Japan's own-account software stock is 7.6 trillion yen (evaluated by the 1995 constant prices) estimated using the cost index and 8.1 trillion yen using the harmonized prices in 2000, which amounts to about 0.4 percent of fixed capital stock and about 0.2 percent of total capital stock including land and inventories. Total software stock in Japan is 25.2 trillion yen estimated using the cost index and 27.5 trillion yen using the harmonized indexes in 2000.

This paper focuses on the compilation process of derived statistics, which are indispensable for empirical economic analysis, and examines various issues inherent in the primary statistics that serve as their foundation. In economics, where the verification of theoretical hypotheses relies on observed facts, the accuracy of processed statistics—constructed by combining multiple statistical surveys—is of critical importance.As a specific example, this study addresses the time-series estimation (1955–1992) of the fixed capital matrix, an essential component for constructing multi-sectoral general equilibrium economic models. Since a systematic National Wealth Survey has not been conducted in Japan since 1970, a decentralized estimation method that combines various existing primary statistics has become the practical solution. However, discrepancies in concepts and definitions, as well as variations in precision, exist among the primary statistics used—such as the Financial Statements Statistics of Corporations and the Census of Manufactures—presenting numerous difficulties in the process of integrating them into a consistent framework.This paper details the estimation procedures using the Perpetual Inventory Method (PIM) and the polynomial double-benchmark approach. Furthermore, it clarifies challenges regarding the adjustment of qualitative changes in capital goods and the estimation of depreciation rates. Ultimately, the paper emphasizes the necessity of mutual cooperation between statistical compilers and users, proposing that insights gained during the compilation of processed statistics should be fed back into the design and improvement of primary statistics.

The objectives of this paper are to describe the problems with the current estimates for WiP inventories of cultivated assets in Japan's system of national accounts and to propose an alternative method of dealing with these assets. There is a huge di erence between inventory stock measurement that are compiled according to the 1968 SNA versus the 1993 SNA for Japan's national accounts. This leads to non-negligible di erences for productivity measures when inventories are treated as a capital input. We argue that the 1993 SNA recommendations have been inappropriately introduced in the current production and wealth accounts for Japan and that the current 1993-SNA version estimates for WiP inventory stocks of cultivated assets (CAWiP) are implausible. As an alternative to the "perpetual inventory method" used for the current estimates for CAWiP, we propose a "realized inventory method" (RIM), where inventory stocks over the past periods are backwardly estimated based on the realized values of cultivated assets to be sold. RIM can provide alternative estimates that we argue are more realistic and can also help with distinguishing CAWiP and non-cultivated biological resources, with the latter being as a non-produced asset, of which the natural growth should not be described in production account.

On the eve of the first commitment period of the Kyoto Protocol (2008–2012), Japan’s actual CO₂ emissions have substantially exceeded earlier projections, leading to a growing recognition of the difficulty of achieving the target. At the same time, increasing attention has been directed toward the need for a longer-term perspective in designing climate policy frameworks beyond 2013. This paper employs a multi-sector general equilibrium model to examine the marginal abatement costs associated with domestic measures required to meet the targets for the first commitment period, and to provide projections of energy supply and demand through 2030 under medium- to long-term policy scenarios. The results indicate that stabilizing emissions at the 1990 level by 2012 would require a marginal abatement cost of approximately 40,000 yen per ton of carbon under domestic measures. Roughly half of the total emission reductions are attributable to a shift in the power generation mix, driven by a reversal in relative costs between coal-fired and LNG-fired power, leading to increased utilization of LNG-based generation.

This paper collates country practices on compilation of national accounts in 18 Asian countries, based on the metadata gathered from a survey conducted by the Asian Productivity Organization (APO) in joint research with Keio Economic Observatory (KEO), Keio University, between April and July 2008. The paper was prepared as a first step towards the construction of an international database for productivity comparisons among Asian countries. Our observations are summed up as follows. First, although most countries are 1993 SNA compliant, the extent of compliance in terms of coverage still varies. The treatment of FISIM and software is far from standard in this country group, and methods to align GDP coverage need be considered for any cross-country comparisons. Second, employment data are more fragmented and countries are more likely to have a manufacturing census than a census on services. Third, most countries have detailed benchmark supply-use/input-output tables which can be used to supplement their national accounts. Our judgment is that GDP per worker as a labor productivity volume should be feasible for this country group, whereas GDP per hour worked will be more of a challenge due to data limitations. Industry comparative analysis of labor productivity should be feasible; but comparing total factor productivity performance, which requires capital stocks and services estimates, is feasible for only a handful of countries at the whole-economy level.

The paper looks at the contribution of the market sector to changes in Japan's living standards over the years 1955-2006. It considers real income as a measure for living standards and decomposes its growth rate into three components: components due to changes in productivity, in real output prices (including changes in the terms of trade) and in primary input growth. The exact index number approach developed by Diewert and Morrison (1986) and Kohli (1990) (2004) is adapted to this real income context. Finally, the paper switches from a gross output concept to a theoretically preferred net output concept. In the net output context, it turns out that the role of capital deepening as a contributor to higher living standards diminishes and the role of productivity and labour growth becomes more important.

This paper presents a comparison of total factor productivity (TFP) levels between the US and Japan for the period 1960–2004 and allocates the gap to individual industries. We carefully distinguish the various concepts of purchasing power parity (PPP) and measure them within the framework of a US–Japan bilateral input–output table. We also measure industry-level PPPs for capital, labor, energy, and materials inputs and output for 42 industries common to the US and Japan, based on detailed estimates for 164 commodities, 33 assets, including land and inventories, and 1596 labor categories. The US–Japan productivity gap shrank during three decades of rapid Japanese economic growth, 1960–1990. The Japanese manufacturing sector achieved parity with its US counterpart by the end of the period. With the collapse of the Japanese economic bubble at the end of the 1980s, the US–Japan productivity gap reversed course and expanded to 79.5% by 2004. This can be attributed to rapid productivity growth in the IT-producing industries in the US during the late 1990s and the sharp acceleration of productivity growth in the IT-using industries in the US during 2000–2004. Wholesale and Retail Trade emerged as the largest contributor to this gap, accounting for 25.1% of the lower TFP of the Japanese economy.