Two-Tiered Demand Structure in Japan’s Biomass Energy Market: Evidence from Wood Pellet Imports Under the Feed-In Tariff Scheme

Abstract

Japan’s import market for wood pellets has expanded rapidly since the introduction of the feed-in tariff (FIT) scheme in 2012, with imports exceeding six million tonnes in 2024, positioning Japan as the world’s second-largest wood pellet importer. Despite this expansion, empirical evidence on its demand structure remains limited. This study employs a Dynamic Linear Approximate Almost Ideal Demand System (Dynamic LA-AIDS) model incorporating demand inertia stemming from long-term fuel supply contracts to analyze Japan’s wood pellet import demand from 2012Q1 to 2025Q3. The results reveal a distinct two-tiered structure: North American pellets behave as a strategic necessity, exhibiting price-inelastic demand and a tendency toward a stable long-run procurement pattern following price and expenditure shocks, suggesting procurement practices that prioritize supply security under long-term contracts. In contrast, Vietnamese pellets behave as a price-sensitive commodity, displaying price-elastic demand and relatively sustained responsiveness following such shocks. These results indicate a dual procurement strategy under the FIT scheme that balances stability and cost flexibility. Importantly, the Japanese demand structure differs from the more uniformly price-inelastic patterns observed in the EU and South Korean markets, providing new insights into how institutional frameworks shape biomass allocation and market responsiveness in renewable energy systems.

1. Introduction

Wood pellets are characterized by high bulk density and homogeneity, which improve transport efficiency [1]. In addition, they are widely available, relatively low-cost, and have low ash content, contributing to their suitability for international trade. Consequently, they are widely used not only for residential heating but also for industrial applications, such as biomass power generation and co-firing in coal-fired power plants. The implementation of renewable energy promotion policies as part of climate change mitigation efforts has driven a substantial expansion in the production, consumption, and international trade of wood pellets since the late 2000s [2,3]. By 2024, global production reached 48 million tonnes. While production remains concentrated in Europe (47%) and North America (28%), the Asia-Pacific region has shown a recent upward trend, accounting for 22% of global output [4,5]. Approximately two-thirds of this production is traded internationally, with around 75% of imports concentrated in five countries: the United Kingdom, Japan, South Korea, Denmark, and Italy.

Within this global context, Japan’s demand for wood pellets has expanded rapidly. This growth was driven by the Feed-in Tariff (FIT) system introduced in July 2012, following the suspension of nuclear power plants after the 2011 Great East Japan Earthquake [6]. The FIT supports renewable energy by guaranteeing a fixed purchase price for up to 20 years for projects approved by the Ministry of Economy, Trade and Industry (METI) [6,7]. Under this system, electric utilities are mandated to purchase electricity from certified renewable sources, with costs recovered through a surcharge on consumers’ bills. Purchase prices vary by fuel type; for example, electricity generated from “general wood,” including imported wood pellets, was set at 24 JPY/kWh and was eligible for a 20-year fixed purchase period under the FIT [7]. By reducing investment risks and improving the profitability of biomass projects—including co-firing in existing coal plants—the FIT created strong incentives for the rapid proliferation of biomass power facilities across Japan [6].

As demand for wood pellets surged with the expansion of biomass power capacity, domestic production failed to keep pace. The self-sufficiency rate for wood pellets declined sharply from 57.7% in 2012 to only 2.4% in 2024 [8], forcing power generators to rely on overseas fuel supplies. Consequently, Japan’s wood pellet imports increased dramatically after the introduction of the FIT, surpassing 6 million tonnes in 2024. This represents an approximately 89-fold increase compared to 2012, making Japan the world’s second-largest importer [4,5]. In terms of source countries, Canadian pellets dominated the market at the start of the FIT in 2012. However, imports from Vietnam have surged since 2016, surpassing Canada in 2019 to become the largest supplier, while imports from the United States have gained prominence since 2022. In 2024, these three countries accounted for 87.8% of Japan’s total wood pellet imports by weight (Figure 1).

Looking ahead, the upward trend in fuel demand for biomass power generation is expected to continue. As of March 2025, the cumulative installed capacity of biomass power generation in Japan is estimated at 8.4 million kW (adjusted for biomass share), representing a 3.7-fold increase from 2.3 million kW prior to the introduction of the FIT scheme [10]. In terms of electricity generation, this corresponds to approximately 6.0% of total power supply. The “7th Strategic Energy Plan,” approved by the Cabinet in February 2025, forecasts that renewable energy will account for 40–50% of the 2040 power mix, with biomass expected to maintain a share of 5–6% of total generation [11]. This suggests that the current installed capacity of biomass power generation is already sufficient to achieve the level envisioned by the government, in terms of its share of total electricity generation. Amid this rapid expansion, the FIT system has recently sought to curb new investments in large-scale biomass—for instance, projects over 10 MW will be excluded from its support after FY2026 [12]. However, many certified projects have yet to begin operations. As of March 2024, 279 out of 523 certified plants were not yet operational, with sequential commissioning expected in the coming years [13]. Accordingly, a report by the United States Department of Agriculture Foreign Agricultural Service (USDA FAS) forecasts continued growth in wood pellet imports [14]. Notably, between 2024 and 2025, 34 new facilities, including two coal co-firing projects, are reported to have commenced operations [15]. Japan has thus emerged as an important importer whose expanding demand is likely to shape global market dynamics in the years ahead.

Despite the rapid expansion of the global wood pellet market, empirical analyses of import demand remain limited. Previous studies applying a system-wide approach with multiple demand equations include Sun and Niquidet [16] and Atasoy and Zhang [17] for the European Union (EU), and Oh and Suh [18] for South Korea. These studies estimate elasticities for country-specific wood pellet import demand and provide insights into the effects of both price and non-price factors on procurement decisions.

In the EU, approximately half of wood pellet consumption is used for residential and small-to-medium commercial heating, while the remainder is allocated to industrial purposes, including power generation. Italy—where residential demand is particularly high—is one of the main importers, whereas countries dominated by power-generation demand, such as the Netherlands and Denmark, account for a large share of imports [19]. The primary import sources are the United States, Canada, and Russia. Existing research indicates that import demand from these countries is consistently price-inelastic, suggesting non-price factors such as pellet quality, supply reliability and stability, and sustainability likely play an important role [16,17]. Sun and Niquidet [16] further suggest that the existence of long-term supply contracts may contribute to the low statistical significance of estimated price elasticities.

In contrast, in South Korea, wood pellet demand is largely driven by power generation. Under the Renewable Portfolio Standard (RPS) introduced in 2012, power generators with a capacity greater than 500 MW are required to achieve a minimum share of renewable energy. If this share cannot be met through their own renewable generation, the shortfall can be covered by purchasing Renewable Energy Certificates (RECs) on the market [20,21]. Due to limited domestic production, South Korea relies heavily on imports from Southeast Asian countries. Oh and Suh [18] find that, although demand from all major source countries remains price-inelastic, increases in the mandated renewable share appear to lead to higher imports from Vietnam and Indonesia, the country’s main suppliers.

In Japan, the FIT system guarantees a long-term fixed purchase price, in contrast to Korea’s quota-driven RPS. Under the FIT framework, project approval for biomass power generation places particular emphasis on securing a stable, long-term fuel supply and ensuring its sustainability. Accordingly, power generators are required to submit detailed fuel procurement and usage plans, along with agreements or contracts with fuel suppliers, at the time of application. Furthermore, separate documentation verifying the legality and sustainability of the biomass resources—such as forest certification or other recognized third-party certification—is also required [22]. This institutional setting may reduce the flexibility of short-term adjustments in procurement sources, which may cause import demand to be partially influenced by past procurement patterns. However, the three prior studies mentioned above do not appear to have examined this potential effect.

This study contributes to the literature by systematically estimating Japan’s wood pellet import demand using the Dynamic Linear Approximate Almost Ideal Demand System (Dynamic LA-AIDS) proposed by Blanciforti and Green [23]. Unlike previous static approaches, this study explicitly models the demand adjustment process. Using quarterly data from 2012Q1 to 2025Q3, the analysis captures long-term demand behavior since the introduction of the FIT system. By incorporating lag terms to account for demand inertia stemming from long-term contracts, this dynamic approach enables us to differentiate short-term and long-term elasticities, thereby distinguishing immediate responses from longer-term equilibrium adjustments.

In addition to estimating own-price, cross-price, and expenditure elasticities, the model provides both short-run and long-run elasticity measures. This framework allows for a detailed examination of the demand characteristics of country-differentiated wood pellets imported into Japan, including substitution patterns across exporting countries and the responsiveness of total import expenditure.

The objective of this study is to elucidate how Japan’s wood pellet import demand adjusts under the unique institutional constraints of the FIT system. Given global concerns regarding supply–demand tightness and the carbon footprint of wood products, a precise characterization of Japan’s demand response provides insights for both energy policy design and the sustainable international allocation of biomass resources, thereby contributing to the development of a more sustainable and carbon-neutral wood market.

2. Materials and Methods

2.1. Data

To ensure a sufficient sample size for model estimation, this study employs quarterly data rather than annual data. While monthly data are also available, they contain a non-negligible number of zero-import observations. These intermittent flows can lead to estimation instability, particularly in logarithmic specifications. Consequently, the data were aggregated to a quarterly frequency to eliminate zero observations and ensure a more robust estimation framework. The sample period spans from the first quarter of 2012 (2012Q1) to the third quarter of 2025 (2025Q3), yielding 55 observations. Data on import quantities and values were obtained from the Trade Statistics of Japan published by the Ministry of Finance [9]. The wood pellet import data are classified under the Harmonized System (HS) 6-digit code 440131, corresponding to “wood pellets” within Chapter 44 (wood and articles of wood). Import prices are calculated as unit values (import value divided by import quantity in tonnes) on a cost, insurance, and freight (CIF) basis. Budget shares are computed as the ratio of expenditure for each origin to total import expenditure. Nominal unit values were deflated using the Import Price Index (JPY-based) for “lumber and wood products and forest products” (2020 = 100), obtained from the Corporate Goods Price Index published by the Bank of Japan (series code 2600420001). Although this index does not specifically include wood pellets, it covers major wood-related products such as plywood, laminated wood, wood chips, fiberboard, and logs. Compared with other available product-specific indices or the more aggregated Import Price Index, it is considered a relatively appropriate proxy for deflating wood pellet prices in the absence of a dedicated price index for wood pellets. Total import expenditure was deflated using the same index to maintain consistency. This adjustment ensures that the analysis reflects the real cost considerations of importers—primarily electric power generators and general trading companies—whose procurement decisions are based on inflation- and exchange-rate-adjusted prices.

As described in Section 1, Japan’s wood pellet imports are predominantly sourced from Canada and Vietnam. Although imports from the United States have increased since 2022, they were negligible in earlier periods and occasionally recorded as zero at the quarterly frequency. To ensure stable model estimation and to avoid distortions arising from zero observations in price and expenditure share variables, this study specifies a three-good system comprising North America (aggregating Canada and the United States), Vietnam, and the Rest of the World (RoW), as shown in

Table 1. Descriptive statistics of variables.

Variable	Unit	Mean	Std. Dev.	Min	Max
Quarterly: 2012Q1–2025Q3
Market Share—North America	%	60.1	23.8	18.0	97.2
Market Share—Vietnam	%	30.7	22.0	1.1	66.9
Market Share—Rest of the World (RoW)	%	9.1	7.7	1.1	36.1
Import Price—North America	JPY/t	21,406.5	3743.9	13,692.8	36,016.2
Import Price—Vietnam	JPY/t	16,805.6	2122.2	11,034.4	23,410.3
Import Price—RoW	JPY/t	19,517.0	6908.8	11,450.9	55,253.7
Total Expenditure (E)	Thousand JPY	585,476.6	661,267.4	9313.0	2,351,668.0

Note: Market shares are expressed as percentages (0–100%) and sum to 100% in each period. N = 55. These variables are employed in the estimation of the dynamic LA-AIDS model.

. The aggregation of Canada and the United States into a single North America category is justified by several considerations. First, pellet quality characteristics are broadly comparable between the two countries. Evidence from a previous study analyzing pellet samples produced in both countries concludes most commercially traded pellets satisfy German and broader European industrial standards [24]. Second, supply structures are similar, with exports largely governed by multi-year take-or-pay contracts. Projected long-term contractual quantity was expected to increase from 2.93 million t/yr in 2022 to approximately 4.75 million t/yr by 2026 [25]. Although these figures were projections at the time of publication, they nonetheless indicate that a substantial portion of Japan’s imports from both countries is secured under long-term supply contracts. Third, from a cost perspective, pellets from both countries generally entail higher procurement costs compared to other source countries. Fourth, Canadian and U.S. wood pellets are supplied under internationally recognized forest certification frameworks, including the Forest Stewardship Council (FSC), Programme for the Endorsement of Forest Certification (PEFC), Sustainable Forestry Initiative (SFI), and Sustainable Biomass Program (SBP) [26,27], thereby operating within similar institutional frameworks that support sustainable forest management and biomass supply chains.

The RoW category mainly consists of Southeast Asian exporters, including Indonesia, Malaysia, and Thailand, and serves as a residual group of smaller suppliers outside North America and Vietnam. During the study period, the combined market share of North America and Vietnam accounted for 90.6% of the total import value, validating the use of a three-good model to represent the overall import market. Although domestic wood pellets are excluded due to price data constraints, their impact on the results is likely limited given Japan’s high import penetration.

The presence of seasonality, which is often observed in quarterly data, was examined using a one-way ANOVA on import quantities and prices for total imports and each origin. The results did not indicate statistically significant differences across quarters, as reported in Appendix A. This empirical finding is broadly consistent with the institutional characteristics of the Japanese wood pellet market, where imported pellets are predominantly utilized for year-round power generation rather than temperature-sensitive residential heating demand. Under such conditions, pronounced seasonal demand fluctuations would not necessarily be expected. Given the absence of statistically detectable quarterly effects in both total imports and origin-specific series, seasonal dummy variables are not included in the baseline specification.

Furthermore, the time-series properties of the data were evaluated. Augmented Dickey–Fuller (ADF) tests indicate that most variables are stationary in levels, while one variable exhibits non-stationarity in levels but becomes stationary after first differencing (see

Table A2. Results of the ADF unit root test (p-values).

Variable	Level (with Trend)	1st Difference (with Drift)	Result
$w_{NA}$	−2.877 (p = 0.0059 ***)	−5.951 (p < 0.001 ***)	I(0)
$w_V$	−1.903 (p = 0.0629 *)	−6.763 (p < 0.001 ***)	I(1)
$\mathrm{l}\mathrm{n}{p}_{NA}$	−2.514 (p = 0.0153 **)	−5.934 (p < 0.001 ***)	I(0)
$\mathrm{l}\mathrm{n}{p}_V$	−2.799 (p = 0.0073 ***)	−5.001 (p < 0.001 ***)	I(0)
$\mathrm{l}\mathrm{n}{p}_{RoW}$	−4.263 (p < 0.001 ***)	−6.183 (p < 0.001 ***)	I(0)
$\mathrm{l}\mathrm{n}(E/P)$	−2.338 (p = 0.0235 **)	−8.039 (p < 0.001 ***)	I(0)

Notes: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Values in parentheses represent p-values. The integration order of each variable is determined based on the 5% significance level.

in Appendix B). Since the majority of variables are stationary, the likelihood of spurious regression in the level specification is limited. For completeness, Johansen’s trace test was also conducted as a supplementary check to examine potential long-run equilibrium relationships within the six-variable system. The results indicate the presence of three cointegrating vectors (

Table A3. Johansen cointegration test results (Trace statistics).

Null Hypothesis	Trace Statistic	5% Critical Value	1% Critical Value	Conclusion
r = 0	171.28	102.14	111.01	Rejected ***
r ≤ 1	108.82	76.07	84.45	Rejected ***
r ≤ 2	67.1	53.12	60.16	Rejected ***
r ≤ 3	33.15	34.91	41.07	Not Rejected
r ≤ 4	14.82	19.96	24.6	Not Rejected

Note: r denotes the number of cointegrating vectors. *** indicates rejection at the 1% significance level. The test was performed on the system containing $w_{NA}$, $w_V$, $\mathrm{l}\mathrm{n}{p}_{NA}$, $\mathrm{l}\mathrm{n}{p}_V$, $\mathrm{l}\mathrm{n}{p}_{RoW}$, and real_exp. A lag length (K = 4) was selected based on the Akaike Information Criterion (AIC). All estimations were performed using R version 4.5.2.

in Appendix B), providing additional indicative rather than conclusive evidence of a possible long-run relationship. Additionally, the Rotterdam model used for comparative validation in Section 2.3 is specified in first differences; given that all variables become stationary after first differencing, its estimation remains econometrically valid.

2.2. Theoretical Model (Dynamic LA-AIDS Model)

To account for the dynamic nature of procurement behavior in Japan’s wood pellet import market and the inertia stemming from long-term supply contracts, this study employs the Dynamic LA-AIDS model proposed by Blanciforti and Green [23]. The AIDS model, originally introduced by Deaton and Muellbauer [28], is a flexible demand system that represents budget allocation across multiple goods using a translog price index and real total expenditure. The model has been widely applied across diverse sectors, such as energy and food markets [29,30,31], and has also found several applications in the analysis of forest products [32,33], including recent studies on wood pellets [16,17]. In empirical practice, a linear approximation of the AIDS model (LA-AIDS) is commonly used to facilitate estimation [29,30,31,32].

Unlike the static AIDS model, the Dynamic LA-AIDS model incorporates the concept of habit persistence, as theoretically discussed by Pollak and Wales [34]. This specification explicitly captures the influence of previous-period budget shares on current demand allocation. By introducing dynamic adjustment, the model allows for the estimation of both short-run and long-run elasticities, thereby distinguishing immediate responses from longer-term equilibrium adjustments. As a recent application of the Dynamic LA-AIDS model, Selvanathan et al. [35] examined consumer behavior in developed and developing countries and reported that, for their dataset, the dynamic specification satisfied theoretical restrictions more effectively than the static model.

2.2.1. Adoption of the Stone Price Index

In this study, the Stone price index ($P_t$) is used as a linear approximation of the non-linear price index in the AIDS model. The Stone price index is derived as a weighted average:

$$\mathrm{l}\mathrm{n}{P}_t={\sum }_{j=1}^n{w}_{j,t-1}\mathrm{l}\mathrm{n}{p}_{jt}$$

(1)

where the lagged budget share ($w_{j,t-1}$) of origin $j$ is used as a weight, the import price ($p_{jt}$) corresponds to origin $j$, and $n$ denotes the number of origins. Since the Stone price index enters the right-hand side of the LA-AIDS share equations, the use of contemporaneous budget shares may lead to simultaneity bias in the estimation of the share equations. To mitigate this issue, some studies construct the Stone price index using budget shares from the previous period as weights (e.g., [36]). Following this approach, this study adopts lagged budget shares for the Stone price index weights.

2.2.2. Model Equations

The basic estimation equation for the Dynamic LA-AIDS model is as follows, following Blanciforti and Green [23] and subsequent studies [35,37,38]:

$$w_{it}={\alpha }_i+\sum _{j=1}^n{\gamma }_{ij}{\mathrm{l}\mathrm{n}p}_{jt}+{\beta }_i\ln \left({\displaystyle \frac{E_t}{P_t}}\right)+{\varphi }_i{w}_{i,t-1}+{\epsilon }_{it}$$

(2)

Here,

• $w_{it}$: Budget share of imports from origin $i$ at time $t$.
• $p_{jt}$: Import price of origin j (JPY-based).
• $E_t/P_t$: Real total expenditure deflated by the Stone price index $P_t$.
• ${\alpha }_i$: Constant term.
• ${\gamma }_{ij}$: Price response parameters capturing substitution effects across origins.
• ${\beta }_i$: Expenditure response parameter.
• ${\varphi }_i$: Degree to which past consumption patterns (i.e., the previous quarter) dictate current demand (habit persistence or inertia). A higher ${\varphi }_i$ indicates stronger inertia and slower adjustment in the dynamic adjustment process.
• ${\epsilon }_{it}$: Error term.

Regarding the dynamic specification, two main variants exist in the literature: the own-lag specification [23], which considers only the lagged share of each good on its own current demand, and the fully generalized (full-cross) specification [35], which allows lagged shares of all goods to affect each budget share. This study adopts the own-lag specification for both theoretical and practical reasons. Theoretically, our primary focus is on the demand inertia resulting from long-term supply contracts, which is most naturally captured by the own-lag term ($w_{i,t-1}$). Practically, given the limited sample size (55 quarterly observations), the full-cross model is highly susceptible to multicollinearity. Preliminary tests indicated a maximum Variance Inflation Factor (VIF) of 13.84 for the full-cross specification, whereas the own-lag model effectively reduced the VIF to 6.44, ensuring more stable and reliable parameter estimations.

To ensure the model is consistent with demand theory, the following constraints are imposed:

Adding-up: $\sum _i{\alpha }_i=1,\sum _j{\gamma }_{ij}=0,\sum _i{\beta }_i=0$;

Homogeneity: $\sum _j{\gamma }_{ij}=0$;

Symmetry: ${\gamma }_{ij}={\gamma }_{ji}$.

2.2.3. Calculation of Elasticities

Using the estimated coefficients from Equation (2), expenditure and price elasticities for each origin are calculated. In the dynamic model estimated here, we distinguish between short-run elasticities, which show immediate responses, and long-run elasticities, which incorporate the effects of the lag term.

(1)

Expenditure Elasticity:

This indicates the percentage change in the demand quantity for origin $i$ given a 1% change in real total expenditure.

• Short-run: ${\eta }_i^S=1+{\displaystyle \frac{{\beta }_i}{w_i}}.$
• Long-run: ${\eta }_i^L=1+{\displaystyle \frac{{\beta }_i}{w_i\left(1-{\varphi }_i\right)}}.$

(2)

Marshallian Own- and Cross-Price Elasticity:

This indicates the percentage change in the demand quantity for origin $i$ in response to a 1% change in the price of origin j. In this study, we focus on Marshallian (uncompensated) price elasticities rather than Hicksian (compensated) ones. This choice is motivated by the context of a market in the expansion phase, where both substitution and income effects are relevant for understanding changes in import quantities and budget shares. By using Marshallian elasticities, the analysis captures the overall response of Japanese pellet imports to price changes more realistically.

• Short-run: ${ϵ}_{ij}^S={\displaystyle \frac{{\gamma }_{ij}-{\beta }_i{w}_j}{w_i}}-{\delta }_{ij}$.
• Long-run: ${ϵ}_{ij}^L={\displaystyle \frac{{\gamma }_{ij}/\left(1-{\varphi }_i\right)-{\beta }_i{w}_j}{w_i}}-{\delta }_{ij}$.

Here, ${\delta }_{ij}$ is the Kronecker delta, which is 1 when $i=j$ and 0 otherwise.

2.3. Estimation Procedure

As the primary objective of this study is to characterize the demand structure of wood pellet imports, the analysis centers on the Dynamic LA-AIDS model, which accounts for the dynamic adjustment process. However, previous studies (e.g., [39,40]) have suggested that the Rotterdam model, which imposes fewer functional form restrictions, may outperform the LA-AIDS model in terms of goodness-of-fit. To ensure the robustness of the estimation results, four cases were estimated using Seemingly Unrelated Regression (SUR). These cases consist of two main specifications of the Dynamic LA-AIDS model—(i) the baseline specification and (ii) its reparameterized form based on relative prices—and two benchmark models: (iii) Static LA-AIDS and (iv) Static Rotterdam model. The Dynamic LA-AIDS model and its relative-price specification are algebraically equivalent representations of the same underlying model. The relative price specification utilizes import prices relative to the RoW price ($p_{i,t}/p_{RoW,t}$), thereby imposing the homogeneity restriction by construction. By normalizing prices against a common benchmark, this specification eliminates redundant scale-price variation and focuses on economic substitution patterns across origins, while maintaining full consistency with demand theory.

Given the adding-up restriction inherent in demand systems, only two equations were directly estimated in the three-good specification, and the parameters of the omitted equation were recovered using the theoretical constraints. Homogeneity and symmetry restrictions were imposed and tested where appropriate.

The performance of these specification cases is evaluated based on information criteria (AIC and BIC), goodness-of-fit measures, and consistency with theoretical restrictions. The empirical results of this model specification evaluation are presented in Section 3.

3. Results and Discussion

3.1. Model Selection and Estimation Results

To identify the most appropriate specification, four alternative estimation cases were estimated. The results are summarized in

Table 2. Comparison of demand system specifications and benchmark models.

	Main Specification (Dynamic LA-AIDS)	Main Specification (Dynamic LA-AIDS, Relative-Price Reparameterization)	Benchmark: Static LA-AIDS	Benchmark: Static Rotterdam
McElroy’s R2	0.91780	0.91780	0.84117	0.91798
AIC	−256.19	−256.19	−222.44	−199.98
BIC	−224.00	−224.00	−195.62	−173.35
Endogeneity (DWH test)	Not significant	Not significant	Not significant	Not significant
Durbin–Watson statistic (by equation)	1.945, 2.428	1.950, 2.448	0.888, 0.908	2.665, 2.766
Variance Inflation Factor (max)	6.520	6.444	4.951	1.390
Homogeneity & Symmetry (LR Test)	Accepted (LR = 2.218, p = 0.528)	Accepted (LR = 2.218, p = 0.528)	Accepted (LR = 2.486, p = 0.478)	Accepted (LR = 4.222, p = 0.239)
Sign of Hicksian own-price elasticity (North America, Vietnam, RoW)	Long-run: +, −, + Short-run: ≈0, −, −	Long-run: +, −, − Short-run: ≈0, −, −	+, −, +	+, +, +
Theoretical consistency	Partial	Partial (relatively stronger)	Partial	Partial

Note: Endogeneity was assessed using the Durbin–Wu–Hausman test, where ln(Y/P) was instrumented with its lag in the LA-AIDS models, and the lagged quantity index dlnQ_t−1 was utilized in the Rotterdam model. Identical McElroy’s R², AIC and BIC values between the dynamic LA-AIDS model and its relative price reparameterization confirm that the two formulations are algebraically equivalent representations of the same underlying demand system, differing only by price normalization. For expositional clarity, both specifications are reported separately. In the row “Sign of Hicksian own-price elasticity,” “+”, “−”, and “≈0” denote positive, negative, and approximately zero estimated elasticities, respectively. All estimations were performed using R version 4.5.2.

.

As shown in Table 2, the Dynamic LA-AIDS model under alternative parameterizations was evaluated in terms of information criteria and goodness-of-fit, together with benchmark static models. While the static Rotterdam model showed marginally higher McElroy’s R² (0.91798 compared to 0.91780 for the Dynamic LA-AIDS), the Dynamic LA-AIDS model yielded the lowest AIC (−256.19) and BIC (−224.00) across all specifications. Notably, its AIC was approximately 33.75 points lower than the static LA-AIDS model and substantially lower than that of the static Rotterdam model, providing strong empirical support for the selection of the Dynamic LA-AIDS specification.

The introduction of the dynamic structure resulted in highly significant lag terms (p < 0.01) in all estimated equations, providing empirical evidence of demand inertia. The Durbin–Watson statistics, which indicated serial correlation in the static LA-AIDS model, improved substantially (ranging from 1.950 to 2.448), suggesting that residual autocorrelation was largely mitigated. Furthermore, the relative price specification slightly improved the maximum Variance Inflation Factor (VIF) to 6.444 compared to the non-relative dynamic model. Since this value is well below the conventional threshold of 10, concerns regarding multicollinearity are considered negligible. Finally, the Durbin–Wu–Hausman (DWH) test showed no evidence of endogeneity, confirming the statistical robustness of the selected specification. The detailed coefficient estimates for the Dynamic LA-AIDS model (relative price specification), along with standard errors and significance levels, are reported in Appendix C.

Regarding consistency with economic theory, Likelihood Ratio (LR) tests confirmed that all four models satisfy the required theoretical constraints. In the case of the Dynamic LA-AIDS model with the relative price reparameterization, homogeneity is inherently imposed by the model, so the LR test was employed solely to examine symmetry. For the Dynamic LA-AIDS model estimated without the relative price specification, both homogeneity and symmetry constraints were accepted in LR tests. These results indicate that imposing homogeneity via the relative price specification is statistically justified, supporting the adoption of this specification.

In the estimation of Hicksian own-price elasticities, very small or sign-ambiguous values are observed for some origins (reported in

Table 3. Short-run and long-run Hicksian price elasticities.

	North America	Vietnam	RoW
North America (SR)	0.007	0.106	−0.113
Vietnam (SR)	0.203	−1.080 ***	0.828 ***
RoW (SR)	−0.740	2.973	−2.244
North America (LR)	0.454	−0.118	−0.336
Vietnam (LR)	0.182	−1.101 ***	0.920 ***
RoW (LR)	−1.260	4.038	−2.778

Notes: Statistical significance at the 1% levels is indicated by ***. SR, LR, and RoW denote short-run, long-run, and the Rest of the World, respectively.

). The Dynamic LA-AIDS model (relative price reparameterization), which is adopted as the preferred specification, provides results that are relatively more consistent with theoretical expectations compared to alternative specifications. Nevertheless, for North American pellets, the estimated Hicksian own-price elasticities remain inconsistent with the standard requirement of negative compensated price responses, taking values of 0.007 in the short run and 0.454 in the long run. These results therefore warrant careful interpretation.

Empirical studies on wood pellet demand have frequently reported extremely weak or near-zero Hicksian price responses rather than clearly negative elasticities. For example, Sun and Niquidet [16] report a Hicksian own-price elasticity of 0.000 for U.S. wood pellets in the EU market, while Atasoy and Zhang [17] obtain similarly near-zero values for Rest-of-World pellets. These findings suggest that, in wood pellet markets, price may not act as a primary driver of demand, particularly under institutional settings characterized by long-term procurement arrangements and limited short-run price flexibility.

More broadly, the demand system literature has documented cases in which Hicksian own-price elasticities do not satisfy the standard sign restrictions. For instance, Ito et al. [41] report positive Hicksian own-price elasticities for several consumer goods in Japan, suggesting that empirical deviations from theoretical sign restrictions are not uncommon in applied demand analysis, especially under conditions of limited price variation.

In the present context, the observed pattern for North American pellets is likely related to structural features of the Japanese wood pellet import market, including long-term procurement contracts under the FIT scheme and import quantities that are fixed or only gradually adjusted. Under such institutional conditions, the estimated Hicksian elasticities may reflect contract-constrained procurement behavior rather than the instantaneous substitution effects assumed in compensated demand theory.

In addition, the Hicksian own-price elasticities for North American pellets are not statistically significant in both the short and long run, suggesting that price responsiveness in this segment is weak and not precisely identified. This result is consistent with previous findings in the EU wood pellet markets [16,17] and supports the view that price plays a limited role under contract-based procurement structures.

Importantly, these findings do not undermine the overall economic interpretation of demand behavior. The Marshallian own-price elasticities, which incorporate income effects and reflect observed market responses, are consistently negative across all origins and time horizons, in line with standard demand theory.

Taken together, although the theoretical restrictions associated with compensated demand are not fully satisfied in all cases, the estimated demand system still provides economically coherent and behaviorally meaningful results. Accordingly, the Dynamic LA-AIDS model (relative price reparameterization) is retained as the preferred specification based on its overall statistical performance and its ability to generate interpretable demand patterns.

3.2. Expenditure Elasticities and Demand Adjustment

Table 4. Short-run and long-run expenditure elasticities.

Origin	Expenditure Elasticity (SR)		Expenditure Elasticity (LR)
North America	0.912	***	0.815	***
Vietnam	1.263	*	1.277	*
RoW	0.681	***	0.553	***

***, and * denote significance at the 1% and 10% levels, respectively. Parameters for RoW are recovered using the adding-up restrictions; significance tests for these recovered parameters are omitted.

presents the short-run (SR) and long-run (LR) expenditure elasticities derived from the dynamic LA-AIDS model (relative price specification). Vietnam exhibits expenditure elasticities greater than unity in both the short and long run, indicating that Vietnamese pellets behave as a luxury (or growth) good, with demand increasing faster than income. In contrast, pellets from North America and the RoW are characterized as necessities, with elasticities below unity. Notably, the budget share of Vietnamese pellets has risen substantially alongside overall market expansion, suggesting that Vietnam has been the main beneficiary of the growing Japanese wood pellet import market. The expenditure elasticities for Vietnam are statistically significant at the 10% level; their relatively lower significance compared with other origins may reflect high volatility due to rapid market entry during the FIT-induced expansion. For North America, the elasticities remain below unity (SR: 0.912, LR: 0.815; Table 4) but are relatively high, indicating that demand quantities have grown despite a declining budget share in favor of Vietnam during the market expansion. Conversely, RoW exhibits lower elasticities (SR: 0.681, LR: 0.553; Table 4), reflecting a limited impact from overall market growth.

Focusing on the dynamic adjustment process, the lag coefficients for both North America and Vietnam are significant at the 1% level (

Table A4. Coefficient estimates along with standard errors and significance levels.

Coefficient	North America (NA) Equation	Vietnam (V) Equation
Intercept $\left({\alpha }_i\right)$	0.5195 *** (0.1372)	−0.1059 * (0.0622)
${\gamma }_{i,NA}$	0.2449 * (0.1424)	−0.1227 (0.1080)
${\gamma }_{i,V}$	−0.1227 (0.1080)	0.0821 (0.0979)
${\beta }_i$	−0.0527 *** (0.0149)	0.0508 *** (0.0130)
${\varphi }_i$	0.5209 *** (0.0974)	0.5556 *** (0.0942)

Notes: Standard errors are reported in parentheses, and statistical significance is indicated as *** p < 0.01 and * p < 0.10. Estimation was performed using R version 4.5.2.

in Appendix C). This statistically confirms the presence of “inertia,” where budget shares are influenced by the previous quarter’s values, reflecting the reality that Japanese power generators engage in continuous and stable fuel procurement based on long-term contracts. The estimated lag coefficients ${\varphi }_i$ are comparable between North America and Vietnam (North America: 0.5209; Vietnam: 0.5556; Table A4), suggesting comparable degrees of persistence in the adjustment process. Accordingly, there is no clear evidence of differential adjustment speeds between the two regions based on the lag structure itself.

It should be noted that these lag coefficients capture persistence in budget shares and should be interpreted separately from differences between short-run and long-run elasticities, which may provide complementary insights into adjustment patterns following expenditure and price shocks.

Turning to expenditure elasticities as a complementary perspective on adjustment dynamics, North America’s expenditure elasticity decreases from the short-run (0.912) to the long-run (0.815), indicating a moderate change in responsiveness across time horizons. While North American pellet demand appears to respond temporarily to overall expenditure shocks, the presence of long-term contracts may constrain deviations, resulting in a gradual return toward a predetermined procurement trajectory.

In contrast, Vietnam appears to maintain relatively high expenditure elasticity in both the short- and long-run, with only minimal differences observed between the two horizons. This pattern may be descriptively interpreted as reflecting relatively stable responsiveness across time horizons. Compared with North America, Vietnamese pellets may benefit from greater flexibility in spot procurement and contract adjustments, potentially allowing them to accommodate growth alongside the expanding market. The observed contrast between Vietnam’s relatively stable high elasticity and North America’s near less responsive adjustment pattern provides a tentative explanation for the dynamic structure of Japan’s two-tiered pellet market.

3.3. Price Elasticities and the Two-Tiered Structure

Table 5. Estimated short-run and long-run Marshallian price elasticities.

	North America	Vietnam	RoW
North America (SR)	−0.536	−0.179	−0.197
Vietnam (SR)	−0.549	−1.475 ***	0.761 ***
RoW (SR)	−1.135	2.760	−2.306
North America (LR)	−0.031 *	−0.373	−0.411
Vietnam (LR)	−0.579	−1.500 ***	0.802 ***
RoW (LR)	−1.589 *	3.865	−2.829

Notes: Statistical significance at the 1% and 10% levels is indicated by *** and *, respectively. SR, LR, and RoW denote short-run, long-run, and the Rest of the World, respectively.

displays the Marshallian price elasticities (hereafter referred to simply as “price elasticities”).

The own-price and cross-price elasticities are broadly consistent with the bifurcation of the demand structure based on regional attributes, highlighting the emergence of a two-tiered demand structure in Japan’s wood pellet import market. Specifically, North American demand is highly price-inelastic, whereas Vietnam and RoW exhibit price-elastic demand that responds strongly to price changes.

North America’s own-price elasticity is relatively small in both the short run (−0.536; Table 5) and long run (−0.031; Table 5). Given that these values are statistically insignificant in the short run and only marginally significant at the 10% level in the long run, the impact of price on demand quantity appears to be negligible. This contrasts sharply with its expenditure elasticity, which was significant at the 1% level. Notably, the own-price elasticity of North American wood pellets is nearly zero in the long run, which may indicate adjustment toward a stable long-run procurement pattern. While demand for North American pellets exhibits only a minimal temporary response to price shocks, it appears to be consistent with a relatively stable long-run procurement pattern, reflecting a relatively price-insensitive state shaped by long-term contracts. This pattern suggests that price may not be the primary determinant when Japanese power generators select North American pellets. Rather, North American pellets are likely treated as essential fuels for stable power plant operation, supported by quality, adherence to international sustainability standards, and low country risk.

In contrast, Vietnam’s own-price elasticity exceeds unity in absolute terms (−1.475 in the short run and −1.500 in the long run; Table 5) and is statistically significant in both the short- and long-run, with only a small difference between the two horizons, similar to expenditure elasticity. This suggests a sustained responsiveness to price. The pattern may indicate that Vietnamese wood pellets are primarily procured as a price-sensitive “commodity,” potentially aimed at reducing fuel costs. While RoW pellets also exhibit high own-price elasticity, their low statistical significance implies that price responses may be more uncertain.

On the other hand, the short- and long-run cross-price elasticities of Vietnam with respect to RoW prices (0.761 and 0.802, respectively; Table 5) are positive and statistically significant, suggesting that the two operate within a similar price segment and may exhibit substitutive relationships, particularly from the perspective of Vietnam’s demand response. Given that Vietnamese pellets account for a substantial market share while RoW pellets hold only a limited share, Japanese power generators appear to rely on Vietnamese pellets as the core of a cost-reduction procurement strategy, with RoW serving as a peripheral adjustment margin. The cross-price elasticities of RoW with respect to Vietnamese prices (short-run: 2.760, long-run: 3.865; Table 5), although above unity, are not statistically significant, and therefore do not provide strong evidence of substitution in the reverse direction. By contrast, such substitutability is not observed between North American and Vietnamese pellets, or between North American and RoW pellets. In these cases, cross-price elasticities are generally not statistically significant, but they appear to indicate complementary rather than substitutable relationships.

3.4. Potential Factors Behind Japan’s Two-Tiered Wood Pellet Market

The analysis of Japanese wood pellet imports under the FIT system reveals patterns that differ from those observed in the EU and Korean markets. Previous studies generally report that import demand in both the EU and Korea is uniformly price-inelastic. In the EU, Sun and Niquidet [16] and Atasoy and Zhang [17] suggest that non-price factors—such as pellet quality, supply reliability, and sustainability—likely influence procurement decisions. Similarly, in Korea, where power generation under the RPS is the main driver of demand, Oh and Suh [18] suggest that increases in the mandated renewable quotas may exert a stronger influence on import demand than price fluctuations alone.

In contrast, the Japanese wood pellet market has expanded under the FIT scheme, with imports primarily used for large-scale power generation. Under this scheme, the electricity purchase price is fixed in nominal terms over the long term, limiting power generators’ ability to increase revenue through higher prices and thereby placing greater strategic emphasis on cost containment, particularly fuel costs. This cost-reduction pressure has likely been reinforced by the recent tightening of the FIT framework. Concerns over rapid biomass power generation capacity expansion and the associated financial burden on electricity consumers led, for example, to the introduction of a competitive bidding system in FY2018 for large-scale (≥10,000 kW) wood biomass power projects [42]. As a result, the feed-in tariff for “general wood,” which was uniformly set at 24 JPY/kWh in FY2012, was reduced to 18.5 JPY/kWh in FY2021 for projects above 10,000 kW based on competitive bidding outcomes [7] (See

Table A5. FIT levels by fiscal year (JPY/kWh).

Fiscal Year	Woody Biomass from Thinning, etc.		General Woody Biomass
	≥2000 kW	<2000 kW	Solid Fuel ≥ 10,000 kW	Solid Fuel < 10,000 kW
2012	32		24	24
2013	32		24	24
2014	32		24	24
2015	32	40	24	24
2016	32	40	24	24
2017	32	40	21 (24 until Sep) ≥20,000 kW	24 <20,000 kW
2018	32	40	Bid system	24
2019	32	40	Bid system	24
2020	32	40	Bid system	24
2021	32	40	Bid system	24
2022	32	40	Bid system	24
2023	32	40	Bid system	24
2024	32	40	Bid system	24
2025	32	40	Bid system	24
2026	32	40	Not eligible	24

Notes: Compiled by the author based on Agency for Natural Resources and Energy [7]. Wood pellet imports for power generation fall under the “General Woody Biomass” category. In FY2017, the tariff for projects ≥ 20,000 kW was reduced from 24 to 21 JPY/kWh (effective September). A competitive bidding system was introduced for projects ≥ 10,000 kW from FY2018 onward. Across eight bidding rounds, two projects were awarded: 19.6 JPY/kWh in Round 1 (FY2018) and 18.5 JPY/kWh in Round 4 (FY2021).

in Appendix D for the trajectory of FIT purchase prices since FY2012).

Despite these economic pressures, power generators are also required to comply with the FIT’s criteria, including long-term fuel supply stability and adherence to legality and sustainability standards. These institutional requirements constrain procurement decisions beyond purely price-based considerations.

Within this context, Japanese power generators appear to adopt bifurcated procurement strategies. North American pellets, valued for their quality, supply stability, and sustainability, serve primarily as base-load fuel. Consequently, their demand is relatively price-inelastic. In contrast, imports from Vietnam and the RoW tend to be more price-responsive, reflecting institutional pressures to reduce total fuel costs.

Taken together, these observations suggest a potential link between the FIT framework and the two-tiered demand structure observed in Japan’s imported wood pellet market. This two-tiered structure contrasts with the more uniformly price-inelastic patterns documented in the EU and Korean markets, indicating that institutional variations may underlie these differences in demand structures across countries. Overall, this provides suggestive evidence that differences in policy design may influence procurement behavior in biomass fuel markets, thereby offering a useful perspective for understanding procurement behavior within bio-based value chains.

4. Conclusions

This study analyzed the demand structure of Japan’s imported wood pellet market within the context of the country’s renewable energy transition from 2012Q1 to 2025Q3 using quarterly data and a dynamic LA-AIDS model. The dynamic specification, which accounts for long-term supply contracts, provides a superior fit to static models, as suggested by the statistical significance of lagged coefficients and the mitigation of autocorrelation in the residuals, as indicated by improvements in the Durbin–Watson statistics.

The analysis indicates several notable features of the Japanese market. First, the overall structure of import demand appears to be two-tiered, differentiated by origin into “strategic necessity” and “price-sensitive commodity.” North American pellets, with expenditure elasticity (0.815 in the long-run) below unity and near-zero own-price elasticity (−0.031 in the long-run), seem to function as a base-load fuel that supports the stable operation of power generation facilities. In contrast, Vietnamese pellets exhibit expenditure elasticity exceeding unity and more price-elastic responsiveness, suggesting they serve as a commodity that facilitates cost containment while capturing the benefits of market expansion. These findings imply that Vietnam has emerged as the primary beneficiary of Japan’s market growth, with its supply likely to continue increasing alongside further market expansion.

Second, the dynamic adjustment process reveals distinctive demand behavior for North American pellets. While demand may respond temporarily to external shocks, such as price fluctuations, these effects do not appear to persist; instead, demand tends to revert quickly to a price-insensitive state based on long-term contracts. Such strong mean-reverting behavior was not observed for Vietnamese pellets. This contrast supports the interpretation that, for North American pellets, Japanese power generators prioritize supply stability and sustainability, and energy security under the FIT scheme.

Third, a comparison with the EU and South Korean markets reveals a unique characteristic of the Japanese market. Prior studies suggest that demand in these regions is uniformly price-inelastic across origins. In contrast, the Japanese market is characterized by the coexistence of a segment dominated by non-price factors (North America) and a segment in which price competition is more active (Vietnam and the RoW). Under the FIT scheme, in which electricity purchase prices are fixed, this two-tiered portfolio management may be interpreted as a rational strategy to simultaneously ensure supply stability and encourage cost optimization. These findings suggest that institutional differences in energy policy, within the broader context of carbon neutrality initiatives, may underlie the observed variations in fuel demand structures across countries.

However, several limitations remain. Although this study provides a systematic and dynamic characterization of the demand structure of the Japanese wood pellet import market, it does not provide a strict causal identification of whether the observed two-tiered demand structure is attributable solely to the FIT scheme. Establishing such causality would require a more explicit identification strategy, possibly incorporating policy variation or counterfactual analysis. Furthermore, the cross-regional comparisons based on the elasticities should be interpreted with caution, as they are descriptive in nature and intended to provide indicative rather than definitive interpretations.

In addition, due to data constraints, Canadian and U.S. wood pellets were aggregated into a single “North American” category. Future research could disaggregate these two major suppliers to investigate potential heterogeneity in their substitution relationships. This would likely require the collection of additional observations or the use of higher-frequency (e.g., monthly) data to secure sufficient degrees of freedom.

Another important extension concerns competing biomass fuels. In Japan’s renewable energy mix, Palm Kernel Shells (PKS) and wood chips hold substantial market shares alongside wood pellets. Since the present study focuses exclusively on the wood pellet sector, cross-price elasticities between pellets and other biomass fuels remain unexplored. In addition, potential substitution between biomass fuels and fossil fuels is not explicitly incorporated in the current model. Moreover, exchange rate fluctuations may not be fully captured despite the use of real price adjustments, and could influence import price dynamics. Expanding the analytical framework to incorporate these competing fuels into a broader demand system would enable a more comprehensive assessment of substitution patterns within the overall biomass fuel market and across energy sources.

Furthermore, imports from emerging suppliers such as Malaysia and Indonesia are increasing, suggesting a gradual multipolarization of Japan’s biomass import structure. Future research should examine how structural changes in supplier composition may influence the two-tiered market framework. Such analysis will be crucial for anticipating future developments in the global wood pellet market.

Continued academic investigation along these lines will contribute to a more robust understanding of international biomass trade dynamics and provide a stronger empirical foundation for designing effective renewable energy and energy transition policies under increasingly tight global biomass supply conditions.