Unequal Paths to Decarbonization in an Aging Society: A Multi-Scale Assessment of Japan’s Household Carbon Footprints

Abstract

Japan’s shift to a super-aged society is reshaping household carbon footprint (HCF) in ways that vary by age, income, and region. Drawing on a two-tier national–prefectural framework, we quantify the influence of demographic shifts on HCF and evaluate inequalities, and project prefectural HCF to 2050 under fixed 2005 technology and consumption baselines. Nationally, emissions follow an inverted-U age curve, peaking at the 50–54 s (2.16 tCO₂) and dropping at both the younger and older ends. Carbon inequality—the gap between high- and low-income households—displays the opposite U shape, being the widest below 30 and above 85. Regional HCF patterns add a further layer: while the inverted U persists, its peak shifts to the 60–64 s in high-income prefectures such as Tokyo—where senior emissions rise by 44% by 2050—and to the 45–49 s in low-income prefectures such as Akita, where younger age groups cut emissions by 58%. Although spatial carbon inequality narrows through midlife, it widens again in old age as eldercare and home energy needs grow. These findings suggest that a uniform mitigation trajectory overlooks key cohorts and regions. To meet the 2050 net-zero target, Japan should integrate age-, income-, and region-specific interventions—for example, targeted carbon pricing, green finance for middle-aged consumers, and less-urban low-carbon eldercare—into its decarbonization roadmap.

1. Introduction

1.1. Background

Global climate governance has increasingly prioritized carbon mitigation, with nations under the Paris Agreement committing to net-zero targets to limit global temperature rise to 1.5 °C [1]. In alignment with these commitments, Japan has committed to a 46% reduction in greenhouse gas emissions by 2030 relative to 2013 levels and to achieving net-zero emissions by 2050 [2]. Although supply-side decarbonization—including industrial efficiency gains and structural shifts toward low-carbon energy—has yielded substantial emissions savings, the proportional contribution of household consumption to national emissions has risen to approximately 80% [3,4]. In this context, technological interventions alone are subject to diminishing returns, rendering lifestyle and behavioral changes indispensable for closing residual emissions gaps. To operationalize this demand-side focus, Japan has introduced a suite of household-targeted policies. For example, the global warming countermeasure tax imposes a modest levy on fossil fuel consumption [5]; emissions trading systems in Tokyo and Saitama target large emitters and, through market mechanisms, exert downstream effects on residential energy prices [6,7]; and subsidies for zero-energy houses incentivize the adoption of highly efficient homes [8]. However, rapid demographic aging in Japan is not merely shifting the age composition but is also fundamentally transforming demand-side consumption profiles and, consequently, the determinants of household carbon footprint (HCF). To sustain both the efficacy and the distributive fairness of existing policy instruments such as carbon levies and retrofit subsidies, it is therefore imperative that demographic transitions be systematically integrated into policy design.

Disparities in CO₂ emissions among social groups have led to the concept of carbon inequality, which not only raises concerns of social justice but also serves as a critical precondition for achieving the Sustainable Development Goals. Incorporating carbon equity considerations into mitigation strategies can significantly improve both public acceptance and policy effectiveness [9]. Traditionally, income inequality has been identified as the primary driver of carbon inequality, with higher-income households contributing disproportionately to emissions due to their greater consumption, thereby skewing the distribution of mitigation burdens and social benefits [10]. However, an increasing body of research highlights the demographic age structure as another equally significant factor influencing emissions disparities [11,12]. As societies age, distinct cohorts display differing consumption patterns and energy-use behaviors, further exacerbating imbalances in CO₂ emissions.

Japan, one of the world’s most rapidly aging societies, is undergoing significant shifts in its population age structure, which are reshaping household consumption patterns and altering carbon inequality [13]. As the proportion of elderly households increases, the evolving HCF profiles present new challenges to long-term mitigation efforts [14]. The limited adoption of eco-friendly practices among the elderly exacerbates energy use for heating and cooking, leading to end-use emissions that eventually equal or surpass those of younger households, whose higher CO₂ emissions stem from transportation and discretionary consumption [15]. Moreover, income inequality exacerbates these disparities, creating a pronounced dual axis of carbon inequality in Japanese society, characterized by significant heterogeneity in HCF across both demographic and economic dimensions.

The impact of demographic aging on HCF varies considerably across national and regional scales [15,16]. National averages often obscure substantial spatial heterogeneity, as Japan’s prefectures differ in both aging intensity and economic structure, leading to significant disparities in HCF [17]. For instance, while Tokyo, with its relatively young population, contributes the highest total HCF, its per-household emissions are lower than those of Okayama, an economically disadvantaged prefecture with a higher aging profile [18]. These regional disparities highlight that macro-level aging effects do not necessarily translate uniformly at the local level. Although national mitigation targets provide a cohesive framework, local capacities and challenges differ, underscoring the need for region-specific climate strategies. Therefore, to meet its national emission reduction goals effectively, Japan must address the carbon inequality arising from the heterogeneous regional characteristics.

1.2. Literature Review

1.2.1. Methodological Approaches for HCF Estimation

HCF can be estimated by three main approaches: environmentally extended input–output (EEIO) analysis, life cycle assessment (LCA), and the consumer lifestyle approach (CLA). EEIO uses a top-down, economy-wide framework that integrates sectoral emission-intensity factors into standard IO tables, capturing both direct and upstream indirect emissions embodied in final demand. For example, Zhang et al. [11] employ an IO model to capture age cohort-specific heterogeneity in consumption structures and HCF. Its adaptability to multi-regional IO (MRIO) or subnational IO tables enables spatial disaggregation from national to prefectural levels; for instance, Huang et al. [19] built a monthly Japanese HCF dataset spanning national, regional, and city scales. In contrast, LCA is a bottom-up methodology for quantifying the environmental impacts of products or processes throughout their entire life cycle. It can be highly data-intensive, since energy and material flows must be quantified at each stage. For example, Campos et al. [20] employ an LCA approach—centered on the CF impact category—to quantify tourism-related emissions using comprehensive life cycle inventory data. The CLA—introduced by Bin and Dowlatabadi [21]—merges household expenditure data with carbon intensities and incorporates demographics, attitudes, and household characteristics. For example, Du et al. [22] used CLA on China’s CFPS data to estimate consumption-based emissions. However, CLA lacks a universally standardized protocol, which may hinder cross-study comparability. For our aim—analyzing HCF variations by age cohort and region—the EEIO approach is particularly appropriate, as it readily accommodates demographic and geographic disaggregation when matched with suitable consumption and population data.

1.2.2. Aging Impacts Across National Contexts

In Japan’s super-aged society, urban elderly households, whose greater wealth ac-cumulation drives significantly higher spending on energy-intensive categories like housing and heating, now exhibit the highest age-based HCF [23]. In contrast, studies from emerging economies, particularly in China, suggest that aging can reduce household emissions, as older cohorts tend to favor frugal consumption and engage less in energy-intensive activities, resulting in a net mitigation effect [24]. Additionally, demographic shifts toward smaller households can diminish economies of scale in shared public re-sources, thereby increasing per capita carbon burdens [25]. These contrasting findings highlight the considerable variation in how demographic changes influence HCF across different national contexts. Furthermore, recent studies have incorporated demographic projections into dynamic HCF assessment models, offering new insights into future emission trajectories. For instance, Wang et al. [14] examine the impact of demographic shifts on age-specific HCF in the U.S. and Japan, revealing that ignoring population dynamics can lead to overestimations of carbon policy effectiveness in rapidly aging societies.

1.2.3. Demographic Aging and HCF Pathways

As population aging accelerates, increasing attention is being given to how shifts in age structure impact HCF [11,12,26]. Systematic variations in consumption preferences and lifestyles across different life stages lead to distinct life cycle HCF profiles [27,28]. Notably, population aging can both mitigate and exacerbate household emissions through complex, multidimensional mechanisms. On one hand, seniors frequently adopt more frugal consumption habits, thereby reducing emissions in certain sectors [29]; on the other hand, the rising prevalence of smaller elderly households increases per capita energy consumption and infrastructure-related carbon intensity, while behavioral inertia further hinders the adoption of low-carbon technologies among older populations [30]. For instance, Wang & Liu [4] demonstrate that elderly households typically reduce spending on clothing, entertainment, and transportation, while spending more time indoors, which, in turn, leads to an increase in household energy consumption.

1.2.4. Multidimensional Inequalities in HCF

The growing focus on HCF has driven scholars to explore carbon inequalities across social groups [31,32]. HCF magnitude and composition vary significantly, influenced by economic stages [33], demographic structures [34], and regionality [35]. From an income perspective, global estimates indicate that in 2019, the wealthiest 10% were responsible for nearly half of all CO₂ emissions, while the poorest 50% contributed only 12% [36]. Regarding age, older cohorts now account for an increasing share of consumption-based emissions in developed countries, raising concerns about intergenerational equity [37]. Spatially, disparities in HCF emerge due to differences in development levels and consumption patterns [38]. Metropolitan areas, with higher incomes and younger populations, typically exhibit higher per capita HCF, while areas facing youth out-migration and accelerated aging tend to have lower emissions [39,40]. However, most studies focus on a single dimension of inequality [15,41,42] and fail to capture the complexity of carbon inequality arising from their interactions. This study addresses this gap by developing a unified framework that comprehensively examines HCF distributions across income, age, and regional dimensions, offering a more robust assessment of multidimensional carbon inequalities.

1.3. Research Purpose

Japan is undergoing an unprecedented acceleration in population aging, accompanied by pronounced disparities across prefectures. In this context, this study aims to clarify how deep-seated shifts in age structure and regional heterogeneity—both of which re-shape consumption patterns and amplify inequalities in HCF—drive the mechanisms underlying HCF formation and its spatiotemporal variability, thereby illuminating the principal determinants and evolutionary pathways of carbon inequality. To achieve this, we innovatively implement a dual-scale analytical framework that integrates age-structured demographics at both national and prefectural levels within an EEIO architecture (Figure 1). At the national tier, we disaggregate HCF by thirteen age cohorts and ten income deciles—thereby embedding age structure directly into the EEIO model—and apply the carbon footprint Gini (CF-Gini) approach to quantify within-cohort, income-based inequalities. At the prefectural tier, we integrate the same age cohort disaggregation into an MRIO-augmented EEIO framework to evaluate HCF from 2005 and project prefectural HCF trajectories from 2020 to 2050 based on demographic forecasts, then deploy the Dagum–Gini approach to isolate spatial sources of age-driven carbon disparities. By targeting cohort-specific disparities and capturing regional variation across multiple scales within Japan’s rapidly aging population, this innovative dual-scale framework systematically reveals how demographic aging reshapes HCF both within and between age cohorts and prefectures. It uncovers the cumulative effects of long-term mitigation policies alongside hidden equity trade-offs, and disentangles the combined roles of age, income, and spatial heterogeneity to yield actionable, equity-focused insights. These findings support the development of differentiated, precision emission reduction strategies and provide empirical guidance for designing nuanced carbon-neutrality instruments.

The remainder of the paper is organized as follows. Section 2 describes the methodology and data used in this work, and Section 3 presents the results and discussion. Section 4 concludes the paper with policy implications.

2. Materials and Methods

Using household consumption data, we first assess HCF at both the national and the prefectural level. We subsequently analyze carbon inequalities from both income and spatial perspectives. Finally, leveraging prefectural demographic data, we project the long-term trends of HCF across prefectures.

2.1. HCF Assessment Framework

2.1.1. National-Scale HCF

Initially proposed by Leontief [43], the IO model quantifies sectoral outputs by tracing intersectoral dependencies. Moreover, two methodological branches predominate: the single-region IO (SRIO) model and the MRIO model. Extending this top-down macroeconomic framework, contemporary applications integrate environmental data—specifically energy consumption and emission intensity coefficients—yielding the EEIO model [44]. Consistent with Japan’s national IO table, this study employs the SRIO model to estimate national HCF.

The SRIO model is formally defined by the following equations:

$$X_{\mathrm{N}}=(I-A{)}^{-1}{F}_{\mathrm{N}}$$

(1)

where $X_{\mathrm{N}}$ is the total national output, $I$ denotes the identity matrix, $A$ refers to the technical coefficient matrix, and $F_{\mathrm{N}}$ is the national final demand.

National HCF is calculated using CO₂ emission intensity (i.e., CO₂ emissions per unit of economic output) as follows:

$$C_{\mathrm{N}}=K_{\mathrm{N}}(I-A{)}^{-1}{H}_{\mathrm{N}}$$

(2)

where $C_{\mathrm{N}}$ is the national HCF driven by household consumption; $K_{\mathrm{N}}$ is the national carbon intensity vector, referring to sectoral CO₂ emissions per monetary unit; and $H_{\mathrm{N}}$ is the national household consumption.

Households underwent a two-stage classification: by age into thirteen cohorts (0–29 s, 30–34 s, 35–39 s, 40–44 s, 45–49 s, 50–54 s, 55–59 s, 60–64 s, 65–69 s, 70–74 s, 75–79 s, 80–84 s, 85 s–) and, within each cohort, by income into ten brackets (0–200, 200–300, 300–400, 400–500, 500–600, 600–700, 700–800, 800–1000, 1000–1500, 1500–; unit: JPY 10,000). Age intervals were defined by stratifying household heads’ ages in the National Survey of Family Income and Expenditure (NSFIE) into thirteen quantile-based groups. Likewise, income intervals were derived from the NSFIE’s monthly income categories for household heads (converted to annual values for this study) and organized into ten quantile-based groups. We assumed that household consumption gradually increases with improvements in income, resulting in greater HCF. Additionally, we estimated the national-level average household size for each age cohort (see

Table A1. Household characteristics at the national level in Japan (2020).

Age	The Number of Households by Income Group										Average Household Size (Person)	Average Annual Housheold Income
	0– 200	200– 300	300– 400	400– 500	500– 600	600– 700	700– 800	800– 1000	1000– 1500	1500–
0–29 s	416,388	803,869	1,143,757	911,629	463,847	154,418	91,262	165,307	21,602	0	1.69	386.7
30–34 s	178,895	255,612	473,428	534,216	483,269	374,470	206,290	193,847	134,098	15,974	2.34	544.4
35–39 s	115,678	318,692	427,661	529,869	503,397	478,034	320,104	369,383	277,905	37,207	2.79	601
40–44 s	229,379	274,005	393,289	459,337	554,260	567,284	443,517	508,149	492,093	76,593	2.88	652
44–49 s	347,913	328,701	432,610	531,148	496,155	559,105	523,026	692,103	579,109	173,072	2.67	681.1
50–54 s	385,887	373,559	354,285	376,207	386,643	458,727	382,128	672,941	725,131	304,298	2.55	749
55–59 s	341,745	295,340	297,250	340,756	386,260	408,003	339,205	659,127	695,942	285,197	2.31	746.4
60–64 s	512,847	516,144	525,754	572,912	435,892	339,108	309,861	399,300	425,545	202,006	2.49	629.6
65–69 s	717,189	732,400	848,910	729,036	516,584	381,416	279,763	309,731	268,508	125,881	2.48	523
70–74 s	770,436	898,066	957,374	677,523	503,564	291,955	193,383	262,332	260,338	89,540	2.54	473.8
75–79 s	867,380	985,646	850,586	471,640	257,861	161,626	97,474	150,354	109,785	57,307	2.63	418.8
80–84 s	934,955	776,894	657,186	326,314	184,050	89,748	61,662	87,570	111,743	30,440	2.42	374.5
85 s–	536,197	662,786	497,521	248,855	132,784	68,124	71,828	44,448	60,224	27,352	2.57	361.4

Note: Household income is divided into ten deciles (units of JPY 10,000), and age structure is divided into thirteen cohorts. Here, ${Num}_{HH}$ denotes the number of households. The units for both the income quantile groups and the annual average income are in JPY 10,000.

), which exhibited a narrow range and strong convergence across cohorts. Accordingly, adopting a per-household HCF metric neutralized these marginal compositional differences, ensuring unbiased comparisons both within and between age groups. Moreover, given the relative stability of household composition in Japan over the study period, we aligned each prefectural age cohort’s average household size with the national mean. This simplifying assumption reinforced the robustness and comparability of the HCF assessments across regions and cohorts.

Accordingly, $H_{\mathrm{N}}$ can be reformulated as:

$$H_{\mathrm{N}}=\sum _{k=1}^{13}\sum _{\mathcal{l}=1}^{10}{H}_{\mathrm{N}}^{k,\mathcal{l}}$$

(3)

where $H_{\mathrm{N}}^{k,\mathcal{l}}$ denotes the household expenditure of the $k$th age cohort and $\mathcal{l}$th income bracket.

Accordingly, Equation (2) can be rewritten as follows:

$$C_{\mathrm{N}}^{k,\mathcal{l}}=K_{\mathrm{N}}(I-A{)}^{-1}\sum _{k=1}^{13}\sum _{\mathcal{l}=1}^{10}{H}_{\mathrm{N}}^{k,\mathcal{l}}$$

(4)

where $C_{\mathrm{N}}^{k,\mathcal{l}}$ refers to the national HCF of the $k$th age cohort and $\mathcal{l}$th income bracket.

2.1.2. Prefectural-Scale HCF

We employed an EEIO model, derived from Japan’s subnational MRIO table, to ex-amine how interactions between age structure and regional economic characteristics shape HCF across prefectures. The core framework of the MRIO model can be formulated as follows [45]:

$$X_{\mathrm{P}}=(I-A{)}^{-1}{F}_{\mathrm{P}}$$

(5)

$$X_{\mathrm{P}}=\left[\begin{array}{c}{x}^1\\ x^2\\ ⋮\\ x^n\end{array}\right],A=\left[\begin{array}{cccc}{A}^{11}& A^{12}& \dots & A^{1n}\\ A^{21}& A^{22}& \dots & A^{2n}\\ ⋮& ⋮& \ddots & ⋮\\ A^{n1}& A^{n2}& \dots & A^{nn}\end{array}\right],F_{\mathrm{P}}=\left[\begin{array}{cccc}{F}^{11}& F^{12}& \dots & F^{1n}\\ F^{21}& F^{22}& \dots & F^{2n}\\ ⋮& ⋮& \ddots & ⋮\\ F^{n1}& F^{n2}& \dots & F^{nn}\end{array}\right]$$

(6)

where $X_{\mathrm{P}}$ denotes the vector of total output by prefecture. The technical coefficient submatrix, $A=\left(a_{ij}^{gs}\right)$, is defined by $a_{ij}^{gs}={\displaystyle \frac{z_{ij}^{gs}}{x_j^s}}$, where $z_{ij}^{gs}$ represents the financial transfer from sector $i$ in prefecture $g$ to sector $j$ in prefecture $s$, and $x_j^s$ denotes the total output of sector $j$ within prefecture $s$. The entry $f_i^{gs}$ of the final demand matrix $F_{\mathrm{P}}$ captures the demand of prefecture $s$ for goods of sector $i$ imported from prefecture $g$. The subscripts $i$ and $j$ run from 1 to 80, representing the full set of economic sectors, while $g$ and $s$ run from 1 to 47, with each corresponding to one prefecture.

Accordingly, the prefectural HCF is calculated thus:

$$C_{\mathrm{P}}^k=K_{\mathrm{P}}(I-A{)}^{-1}{H}_{\mathrm{P}}^k$$

(7)

where $C_{\mathrm{P}}^k$ denotes prefectural HCF derived from household consumption of the $k$th age cohort, $K_{\mathrm{P}}$ is the vector of carbon intensities for each prefecture, and $H_{\mathrm{P}}$ represents prefectural household consumption. Meanwhile, at the prefectural level, to preserve comparability with the national age distribution, we stratified household heads’ ages into the same thirteen quantile-based intervals.

2.2. Approach to Carbon Inequality Assessment

2.2.1. Income-Based HCF Inequality

Originating in Corrado Gini’s work [46], the Gini coefficient is a standard measure of regional income inequality, assigning values from 0 to 1, where higher numbers reflect more pronounced inequality. Applying the standard Gini coefficient formula to the national HCF estimates by age cohort and income bracket (Section 2.1.1), we derived the CF-Gini coefficient as follows:

$$G_{\mathrm{N}}=\sum _{\mathcal{l}=1}^{10}{w}_{\mathcal{l}}{s}_{\mathcal{l}}+2\sum _{\mathcal{l}=1}^{10}{w}_{\mathcal{l}}(1-T_{\mathcal{l}})-1$$

(8)

where $G_{\mathrm{N}}$ denotes the CF-Gini coefficient, indicating the dispersion of HCF across income levels. For each income bracket $\mathcal{l}$, $w_{\mathcal{l}}$ captures the fraction of all households falling into that bracket, $s_{\mathcal{l}}$ indicates the share of total HCF contributed by those same households, and $T_b$ then accumulates these emission shares up through bracket $\mathcal{l}$.

2.2.2. Spatial-Based HCF Inequality

Standard indices such as the Gini coefficient assume normality and homoscedasticity, thereby excluding overlapping group observations and impeding meaningful economic decomposition of total inequality. To overcome these limitations, Dagum [47] introduced a decomposition framework that partitions overall disparity into intragroup disparity, intergroup disparity, and overlap disparity—a methodology now widely adopted in empirical inequality research [48,49]. Figure 2 illustrates the analytical framework applied in this study to examine spatial heterogeneity in HCF inequality.

The Gini coefficient quantifying inequality across age cohorts is defined as follows:

$$G_{ab}^{age}={\displaystyle \frac{\sum _{m=1}^{n_a} \sum _{n=1}^{n_b} \left|C_{am}^{\mathrm{H}\mathrm{C}\mathrm{F}}-C_{bn}^{\mathrm{H}\mathrm{C}\mathrm{F}}\right|}{n_a{n}_b\left(\overline{C_a^{\mathrm{H}\mathrm{C}\mathrm{F}}}+\overline{C_b^{\mathrm{H}\mathrm{C}\mathrm{F}}}\right)}}$$

(9)

where $a$ and $b$ represent the group identifiers, which denote any two of the income categories—high, middle, or low income. The variables $n_a$ and $n_b$ indicate the total number of prefectures in each group. The values $C_{am}^{\mathrm{H}\mathrm{C}\mathrm{F}}$ and $C_{bn}^{\mathrm{H}\mathrm{C}\mathrm{F}}$ refer to the HCF of prefecture $m$ in group $a$ and prefecture $n$ in group $b$, respectively. Additionally, $\overline{C_a^{\mathrm{H}\mathrm{C}\mathrm{F}}}$ and $\overline{C_b^{\mathrm{H}\mathrm{C}\mathrm{F}}}$ represent the average HCF values for all prefectures within each respective group.

When all prefectures are treated as a single group, the intragroup Gini coefficient coincides with the overall Dagum–Gini index for HCF (denoted $G^{age}$). Accordingly, Equation (9) can be reformulated as follows:

$$\begin{gathered} G^{age}=\sum _{a=1}^k G_{aa}^{age}{r}_a{t}_a+\sum _{a=1}^k \sum _{b\ne a} G_{ab}^{age}{r}_a{t}_b{D}_{ab} \\ +\sum _{a=1}^k \sum _{b\ne a} G_{ab}^{age}{r}_a{t}_b{(1-D}_{ab}) \end{gathered}$$

(10)

$$G^{age}=G_{intra}+G_{inter}+G_{ultra}$$

(11)

where $G^{age}$ signifies the overall Dagum–Gini coefficient. The term $G_{intra}$ refers to the disparities within groups, while $G_{inter}$ captures the differences between groups, and $G_{ultra}$ reflects overlap disparity. The variable $r_a$ represents the ratio of prefectures in group $a$ relative to the total sample size $n$, and $t_a\equiv {\displaystyle \raisebox{1ex}{$n_a\overline{C_a^{\mathrm{H}\mathrm{C}\mathrm{F}}}$}\!\left/ \!\raisebox{-1ex}{$n\overline{C^{\mathrm{H}\mathrm{C}\mathrm{F}}}$}\right.}$ and $t_b\equiv {\displaystyle \raisebox{1ex}{$n_b\overline{C_b^{\mathrm{H}\mathrm{C}\mathrm{F}}}$}\!\left/ \!\raisebox{-1ex}{$n\overline{C^{\mathrm{H}\mathrm{C}\mathrm{F}}}$}\right.}$ express the average HCF values for groups $a$ and $b$, respectively, compared to the overall average HCF across all prefectures. The condition ${\sum }_a{\sum }_b{r}_a{t}_b=1$ holds, meaning the overall age-based Gini coefficient $G^{age}$ is the weighted average of the Gini coefficients $G_{ab}^{age}$ for all combinations of groups, with the weights determined by $r_a{t}_b$. The variable $D_{ab}$ denotes the relative influence between groups $a$ and $b$.

2.3. Evaluation of the Influence of Population Dynamics on HCF

This study incorporates prefectural demographic data to elucidate how interregional demographic shifts amid Japan’s rapid population aging drive spatial heterogeneity in HCF. To this end, we applied the following equation to quantify the effect of projected demographic shifts on prefectural HCF across age cohorts [13,14]:

$$Q{\left(t\right)}^g=\sum _i\sum _b{C}_{\mathrm{P}}^{g,k}H{\left(t\right)}^{g,k}$$

(12)

where $C_{\mathrm{P}}^{g,k}$ denotes the HCF for age group $k$ in prefecture $g$. $H{\left(t\right)}^{g,k}$ is the number of households in that age group and prefecture in year $t$, where $t^{\prime }=1,\dots 7$ correspond to 2020, 2025, 2030, 2035, 2040, 2045, and 2050, respectively. $Q{\left(t\right)}^g$ is derived from changes in $H{\left(t\right)}^{g,k}$, with carbon intensity, production technology, supply chains, commodity prices, and age-specific consumption patterns held constant at their 2005 base-year levels. Consequently, this study isolated the effect of Japan’s demographic dynamics, particularly population aging, on HCF.

2.4. Data

The energy statistics used to calculate the national-level carbon intensity for 2020 ($K_{\mathrm{N}}$) and the prefectural-level carbon intensity for 2005 ($K_{\mathrm{P}}$) were obtained from Japan’s Agency for Natural Resources and Energy—strictly employing the 2020 national energy consumption statistics for $K_{\mathrm{N}}$ and the 2005 prefectural energy consumption statistics for $K_{\mathrm{P}}$. As the agency periodically revises its historical series and the original, unamended data are no longer accessible, we therefore utilized the latest release—comprising the 2020 national energy consumption statistics (updated 25 April 2025) and the 2005 prefectural energy consumption statistics (updated 27 December 2024)—to ensure data integrity and reproducibility [50,51]. Prefecture-level demographic data were sourced from the National Institute of Population and Social Security Research [52]. Household consumption data were obtained from Japan’s national IO table and the NSFIE compiled by the Ministry of Internal Affairs and Communications [53,54], as well as the subnational MRIO table [55]. Because the NSFIE records expenditures at purchaser prices, while the IO and MRIO tables use producer prices, we applied an optimization method to reconcile both price bases and sectoral classifications [13,56]. Specifically, we first obtained sector-specific retail and transportation margins from the IO structural survey—margin survey [57], then applied these margins to each sector’s household expenditure to calculate their contribution to total consumption. For each sector, margin values were divided by consumer price expenditure to derive margin ratios, which were then applied to estimate allocated retail and transport margins. Subtracting these margins from sectoral consumer expenditures yielded annual household consumption at producer prices—directly comparable to emission intensities for consumption-based accounting. Because margin data are available only at the national level, we assumed that, under a broadly stable economic structure, each prefecture’s sector-specific margin ratios coincided with the corresponding national averages.

The NSFIE survey is conducted every five years, with the most recent data from 2019 used in this study. At the national scale, we employed the quinquennially updated 2020 national IO tables together with NSFIE consumption patterns to estimate the HCF. At the subnational level, although we intended to use MRIO data from a year close to that of the national IO table for consistency, the latest publicly available prefectural MRIO data were from 2005. Accordingly, considering data limitations, we were compelled to combine the NSFIE survey with the prefectural MRIO table to estimate HCF by prefecture in 2005. Additionally, the constraint arising from the discrepancy in data timing between the national- and subnational-level analyses is explicitly addressed in the research limitations (Section 3.3).

3. Results and Discussion

3.1. National-Level HCF

Prior to analyzing national HCF patterns, we summarize national-level household characteristics in Table A1—detailing household counts across ten income deciles and thirteen age cohorts—which reveals two fundamental dimensions of heterogeneity at the national scale that underpin our subsequent HCF analysis. The age cohort distribution exhibits clear trends: younger groups (0–29 s and 30–34 s) concentrate almost entirely in the lower income deciles (0–400); middle-aged cohorts (35–59 s) spread broadly across mid-to-high-income brackets (200–1000), with a peak in the 45–49 s; and household counts fall sharply in the oldest cohorts (85 s–), especially at higher incomes. Moreover, cohort-level affluence varies markedly: mean annual household income rises from JPY 386.7 × 10⁴ in the 0–29 s to JPY 749 × 10⁴ in the 50–54 s before tapering among retirees. Focusing on income variation within each age cohort is essential to disentangling how economic capacity modulates consumption patterns at different life stages and, in turn, shapes HCF, thereby enabling more precise, cohort- and income-targeted mitigation policies.

3.1.1. HCF Patterns by Age and Income Strata

Overall, in 2020, Japan’s per-household HCF exhibits an inverted U-shaped trajectory: it increases from the youngest group (0–29 s) to the middle-aged group (30–64 s) before declining among older households (65 s–) (Figure 3), in agreement with prior research [9]. Two interrelated mechanisms drive this pattern: life cycle-related shifts in consumption structure and changes in mobility behavior [13,58,59]. First, household consumption patterns evolve markedly with age. Middle-aged householders typically enjoy peak earnings while shouldering greater family responsibilities—especially child-rearing—which together boost both the demand for and the ability to purchase carbon-intensive goods and services [16,60]. Upon retirement, however, a reduction in disposable income and a contraction in overall expenditure translate into lower energy use and product consumption, thereby reducing HCF [4,61]. Second, mobility behavior exhibits a parallel life cycle shift. Middle-aged adults often undertake regular, sometimes long-distance commutes by private car or public transit, which elevates transport-related emissions [15]. In contrast, older adults generally limit their activities to local neighborhoods—favoring nearby shopping, healthcare visits, and social engagements—and participate less frequently in long-distance travel or high-intensity leisure, further curtailing transport and service-sector emissions in later life [14,62]. Notably, when we disaggregate HCF by both age and income, a deviation from the usual positive income–emissions relationship emerges among those aged 70 years and above: beyond this age threshold, higher income no longer corresponds to proportionally higher HCF, suggesting distinct behavioral or structural constraints in the oldest cohort.

In the 0–29 s and 30–34 s, average HCF stands at 1.27 and 1.61 tCO₂, respectively, with no representation from high-income households (incomes ≥ 800). This pattern reflects the early career stage of these cohorts, whose constrained resources concentrate expenditures on housing rentals, daily commuting, and essential services, resulting in lower emissions. In contrast, HCF in the 35–59 s rises steadily with income, peaking at 2.16 tCO₂ in the 50–54 s; at this stage, mid-to-late career incomes and larger household sizes drive increased consumption of energy-intensive goods and services, such as housing and private vehicles. Notably, high-income households in the 50–54 s exhibit still higher HCF, with the 1500– income group reaching 3.91 tCO₂, underscoring the amplification of consumption with income accumulation. For the 60–64 s and 65–69 s, despite retirement or semi-retirement, households maintain elevated living standards and consumption patterns, yielding HCF averages of 2.00 and 1.87 tCO₂, respectively. Among the 70–84 s, however, per-household HCF declines markedly in both low-income and high-income groups, whereas the middle-income bracket sustains or slightly increases its HCF levels. For example, in the 80–84 s, per-household HCF declines to 0.86 tCO₂ for the 0–200 income group and to 2.15 tCO₂ for the 1500– income group, corresponding to 78% and 55% of the levels observed in the 50–54 s at the same income categories. Conversely, within the 80–84 s, households in the 300–400 and 500–600 income groups register a per-household HCF of 1.57 tCO₂ and 2.16 tCO₂—values that match or exceed the 1.59 tCO₂ and 1.81 tCO₂ recorded for the 50–54 s. Low-income elderly, constrained by minimal pensions, limited savings, and scarce secondary employment opportunities, sustain lower consumption levels and consequently record reduced HCF. Meanwhile, high-income late-stage elderly allocate a larger proportion of their expenditures to service-based consumption—primarily healthcare—which attenuates their HCF. However, in middle-income elderly, augmented financial resources foster a more balanced consumption structure, whereas the rising prevalence of the nuclear family among the elderly intensifies per-household energy use, jointly elevating their HCF.

3.1.2. Income-Related Carbon Inequality Across Age Cohorts

Analysis of CF-Gini coefficient trends reveals a U-shaped pattern across age cohorts: carbon inequality declines from the youngest group (e.g., 0–29 s; Gini = 0.44) through middle-aged populations (e.g., 45–49 s; Gini = 0.29) before rising again among the elderly (e.g., 85 s–; Gini = 0.48). Both the youngest and oldest cohorts exhibit elevated inequality, driven chiefly by pronounced HCF polarization within low-income households. For example, low-income households constitute 29.2% of the 0–29 s yet contribute 23% of its total HCF; this polarization further intensifies among older cohorts: in the 80–84 s, which comprise 52.8% of the population, low-income households contribute 38% of HCF; in the 85 s–, which represent 51.3% of the population, they contribute 37.1% of HCF. By contrast, middle-aged cohorts display only modest fluctuations in carbon inequality, owing to the predominance of the middle-income subgroup, whose stable earnings and balanced demographic representation buffer HCF distribution. Specifically, the middle-income group comprises 54.5% of the 45–49 s and contributes 49.9% of its HCF. Amid Japan’s aging population, carbon mitigation strategies must address the needs of the elderly within the broader demographic framework. Younger and older cohorts experience significant disparities in social resource allocation during carbon-reduction initiatives, largely due to income instability and social security uncertainties. In contrast, the middle-aged group exhibits relatively lower carbon inequality, suggesting that their stable economic conditions and established household structures provide a promising foundation for reducing consumption-based emissions and alleviating income-related social conflicts.

3.2. Prefecture-Level HCF

To contextualize prefectural HCF patterns, we have compiled intra-prefectural household characteristics in

Table A3. Household characteristics at the prefectural level in Japan (2005).

	Age	0–29 s	30–34 s	35–39 s	40–44 s	44–49 s	50–54 s	55–59 s	60–64 s	65–69 s	70–74 s	75–79 s	80–84 s	85 s–
Prefecture
1	${Num}_{HH}$	180,754	135,358	147,690	186,799	217,802	196,744	198,742	195,224	238,834	229,271	181,182	161,815	116,240
	${Ave}_{HI}$	334.8	456.2	554.3	536.6	579.1	603.9	630	502.5	436	361	275.9	283.5	266.9
2	${Num}_{HH}$	21,301	17,281	23,466	33,181	37,969	41,233	42,367	54,829	62,304	55,687	43,099	39,733	23,515
	${Ave}_{HI}$	420.5	459.6	529.2	599.2	607.5	665.9	585.2	584.1	473.9	411.6	360.9	312.6	314.8
3	${Num}_{HH}$	24,964	18,715	23,065	31,071	35,267	40,424	39,027	51,617	57,163	51,326	40,081	34,369	22,187
	${Ave}_{HI}$	389.5	475.9	573.4	617.5	579.5	708.6	612.3	580.5	576.3	477.5	392.8	373.5	414.6
4	${Num}_{HH}$	71,885	44,582	54,217	68,689	74,070	74,278	72,986	91,889	99,391	86,135	61,228	54,709	37,570
	${Ave}_{HI}$	367.8	569.7	521.5	609.3	642.8	737.1	794	607	563.8	589.8	370.9	386	437.4
5	${Num}_{HH}$	13,238	11,736	16,737	20,662	24,756	28,587	31,943	43,566	49,596	43,471	35,460	34,776	24,552
	${Ave}_{HI}$	320.6	550	480.9	821.1	626.1	645.7	606.1	633.7	521.4	480.5	395.4	333.5	436.9
6	${Num}_{HH}$	17,433	14,827	18,750	22,010	24,139	30,046	31,725	43,978	48,102	43,040	32,376	30,599	22,661
	${Ave}_{HI}$	426.5	499.4	636.5	573	668.1	752.5	774.2	639.7	552.9	559.7	460.9	462.9	461.3
7	${Num}_{HH}$	37,135	27,898	36,428	49,840	55,288	61,737	63,665	80,812	85,382	74,820	52,026	49,456	35,086
	${Ave}_{HI}$	375.5	465.6	546.1	571.9	679.3	770	635.8	628.5	503.6	470	390.3	403.3	345.5
8	${Num}_{HH}$	72,307	57,572	67,452	83,366	94,858	95,837	101,336	104,036	110,222	109,403	89,600	64,918	44,989
	${Ave}_{HI}$	433.4	554.6	635.2	681.8	683.9	816.3	788.3	658.6	547.5	491.5	440.3	432.9	370.1
9	${Num}_{HH}$	51,471	42,502	46,777	57,893	62,831	64,796	68,996	71,431	73,183	68,834	56,114	35,649	23,702
	${Ave}_{HI}$	388.7	512.1	556.6	594.7	757.6	772.5	799.2	626.1	588.3	529.1	460.6	303.1	340.3
10	${Num}_{HH}$	50,608	35,239	46,777	57,207	65,836	62,085	64,879	66,388	74,125	71,021	60,627	48,846	35,584
	${Ave}_{HI}$	441.1	544.7	597.2	586.9	641.4	743.6	792.5	625.8	559	441.5	436.3	330.6	348.5
11	${Num}_{HH}$	222,764	171,757	193,817	246,660	284,367	271,388	236,226	236,751	269,650	290,436	232,826	162,744	93,397
	${Ave}_{HI}$	403.8	546.7	644.3	651.8	677.9	752	881.3	668.3	508.8	510.2	396.9	394	392.6
12	${Num}_{HH}$	199,138	140,275	159,189	193,801	231,391	214,221	192,601	188,225	222,733	230,617	190,588	133,825	77,745
	${Ave}_{HI}$	410.3	557.1	564.7	698.9	770.6	793.6	858.8	674	539.5	502.6	400.5	342.6	417.2
13	${Num}_{HH}$	793,446	514,154	499,859	522,312	600,076	549,980	435,356	368,671	431,567	440,678	395,874	336,692	260,731
	${Ave}_{HI}$	415.8	665.3	705.3	774	690.8	845.8	752.2	768	580.6	515	578.7	434.6	385.1
14	${Num}_{HH}$	367,982	256,176	283,589	337,285	407,634	370,894	303,392	272,077	318,997	332,535	285,921	222,622	155,673
	${Ave}_{HI}$	415.5	577.1	634.3	698.7	752.2	867.2	832.1	708	512.6	494	437.3	477.3	389.4
15	${Num}_{HH}$	56,248	39,043	36,960	57,236	60,526	68,301	68,544	86,887	88,911	86,112	68,320	61,264	46,626
	${Ave}_{HI}$	408	557	552.7	597	583.4	677.3	727.3	643.1	502.8	539.5	504.9	394.7	396.5
16	${Num}_{HH}$	23,840	17,715	18,031	28,918	30,172	31,046	28,248	35,886	44,125	44,476	30,260	27,796	20,574
	${Ave}_{HI}$	443.1	540.9	603.8	663.4	636.9	808.5	817.5	749.9	580.6	574.6	440.9	513.5	493
17	${Num}_{HH}$	35,746	22,384	22,672	37,758	36,307	37,668	32,676	39,487	46,863	46,355	31,066	26,122	18,943
	${Ave}_{HI}$	385.2	562	574.8	682.6	659.3	741.6	733.2	655.1	565	477.3	367.9	395.1	486.6
18	${Num}_{HH}$	17,668	12,934	12,643	16,415	20,551	23,365	22,312	26,658	29,677	29,549	21,355	17,213	14,030
	${Ave}_{HI}$	427.6	568.2	708.9	668	635.1	794.3	821.4	693.1	577.9	551.4	520.9	475.7	388
19	${Num}_{HH}$	17,954	13,720	17,020	20,344	25,636	29,393	29,078	27,067	29,117	30,855	27,160	23,570	19,539
	${Ave}_{HI}$	350.9	526	561.3	583.3	642.9	700.9	779.8	573.1	487.3	510.8	390.9	427.7	316.3
20	${Num}_{HH}$	44,725	35,144	44,018	54,274	62,976	64,717	66,823	65,830	75,303	78,238	67,726	61,109	51,488
	${Ave}_{HI}$	374.1	576.9	590	599.8	717.3	701.3	732.8	589.2	589	543.1	439.5	386	348.5
21	${Num}_{HH}$	39,505	33,111	44,035	56,065	62,624	64,237	57,148	66,356	78,186	79,882	65,138	46,594	33,577
	${Ave}_{HI}$	402.9	530.4	650.8	617	721.8	732.7	917.6	688.7	564.3	499.7	471.1	405.7	532.1
22	${Num}_{HH}$	80,606	68,477	84,850	104,021	124,797	124,589	112,326	124,186	139,277	145,323	117,480	90,901	63,489
	${Ave}_{HI}$	409	544.3	641.3	643.2	684.7	747.7	744.4	650.3	597.6	552.9	421.1	395.1	453
23	${Num}_{HH}$	258,997	192,876	222,900	273,259	302,865	282,484	220,450	225,755	262,795	281,035	222,482	165,199	104,415
	${Ave}_{HI}$	378.8	545.4	653	718.7	787.8	791.2	733.1	773.8	608.8	495.9	431.1	416.3	406.3
24	${Num}_{HH}$	41,558	34,272	44,272	57,010	61,898	62,588	55,853	56,860	68,780	71,711	60,451	48,626	35,120
	${Ave}_{HI}$	426.2	624.9	622.6	621.4	715.7	921.4	801.3	567.8	557.1	470.6	454.5	402.1	337.4
25	${Num}_{HH}$	47,799	25,696	40,482	41,346	52,768	45,680	45,575	42,703	48,537	43,676	35,157	28,666	20,506
	${Ave}_{HI}$	411	596.5	590.1	617.6	729.8	816.6	850.3	632.5	573.1	525.5	424.3	464	452.9
26	${Num}_{HH}$	101,182	44,320	71,063	77,000	101,372	89,498	84,806	80,626	105,964	106,237	89,729	69,497	52,041
	${Ave}_{HI}$	319.8	516.8	507	698.4	759.9	715.5	662.6	623.1	494.2	453.5	429.1	365.4	308.1
27	${Num}_{HH}$	346,565	195,250	273,528	294,481	394,810	333,847	291,525	277,607	347,900	380,638	333,373	238,931	159,529
	${Ave}_{HI}$	377	457.8	544.5	589.7	611.2	699.8	655.5	586.3	465.3	421	381.7	298.1	345.4
28	${Num}_{HH}$	168,425	98,157	147,950	158,528	221,792	192,149	187,743	177,963	214,242	220,677	189,107	147,309	107,172
	${Ave}_{HI}$	396	477.9	569.5	653.9	735.3	733.9	823	599.6	525.2	438	402	361.3	377.7
29	${Num}_{HH}$	28,802	19,091	30,487	34,888	50,012	44,210	45,208	44,481	53,914	57,799	49,660	35,172	24,270
	${Ave}_{HI}$	308.4	485.4	539.9	569.2	655.9	798.1	759.1	571.9	538.7	440.9	383.5	428.2	352.7
30	${Num}_{HH}$	14,604	12,978	20,360	22,075	28,797	29,807	32,503	31,817	37,631	40,332	36,720	30,833	24,532
	${Ave}_{HI}$	338.2	507.1	488.1	593.8	606.6	598.5	724.9	484.9	426.6	497.8	377.1	272	300.6
31	${Num}_{HH}$	12,112	10,072	10,812	14,756	14,928	17,082	16,483	22,886	22,524	22,086	17,237	15,411	12,466
	${Ave}_{HI}$	334.3	526.4	490.3	569.8	710.9	726.1	754.4	603	518.4	524.9	442.2	366.4	363.6
32	${Num}_{HH}$	15,329	11,916	12,269	16,861	16,575	19,191	19,032	27,378	28,911	28,230	21,450	20,370	7860
	${Ave}_{HI}$	357.9	572.3	510.5	679.8	715.2	773.4	655.6	628.2	611.7	516.5	510.7	372.2	377.3
33	${Num}_{HH}$	61,559	40,823	41,341	58,914	51,212	57,756	49,872	66,162	75,285	82,289	59,103	52,291	40,629
	${Ave}_{HI}$	398.7	547.1	544.4	576.8	659.8	844	932	576.6	681.4	499.9	401.6	390.8	330.1
34	${Num}_{HH}$	102,145	68,360	68,856	96,659	96,649	90,909	78,463	97,390	114,618	122,720	87,479	73,943	63,710
	${Ave}_{HI}$	389.1	485.7	612.3	575.8	706.3	646.8	756.1	592.9	574.6	429	404.2	329.9	349.3
35	${Num}_{HH}$	41,061	24,780	27,859	39,723	39,364	41,753	39,172	54,609	62,728	66,508	52,001	48,670	38,339
	${Ave}_{HI}$	364.1	518.5	572.2	594	656.5	706.3	779	531.5	474.8	412.7	416	338.6	287.7
36	${Num}_{HH}$	17,457	11,866	13,177	19,702	22,149	23,353	23,973	30,625	30,947	33,265	23,709	23,395	16,764
	${Ave}_{HI}$	298.9	510.5	587.9	665.7	687.9	677	685.6	511.8	468.5	398.8	385.6	370.9	319.5
37	${Num}_{HH}$	22,218	14,545	19,369	30,542	30,402	29,930	29,596	36,911	40,302	43,617	30,137	28,590	22,653
	${Ave}_{HI}$	357.9	534.5	584.2	652.1	686.7	736.9	826.8	579	505.4	456.9	324.5	411.1	318
38	${Num}_{HH}$	33,539	29,080	26,229	40,801	41,434	44,811	44,328	56,458	56,277	64,508	46,200	45,005	39,358
	${Ave}_{HI}$	295.3	469.1	662.3	572.9	611.8	585.8	561.1	588.2	521.5	473.7	381.3	314.1	271.9
39	${Num}_{HH}$	16,297	13,482	14,058	20,949	20,703	22,920	23,635	29,582	31,958	36,281	26,118	27,071	23,155
	${Ave}_{HI}$	288.3	454.2	551	608.3	655.7	531.9	596.8	596.3	401.8	372.4	308.5	278.7	277.9
40	${Num}_{HH}$	182,030	110,465	158,484	169,226	188,309	167,528	156,192	185,707	226,705	200,533	153,251	127,946	93,966
	${Ave}_{HI}$	315.8	445.5	524	622.1	589.4	629.7	650.1	581.3	443.6	413	356.3	354.8	330.2
41	${Num}_{HH}$	17,998	11,011	18,216	18,583	22,568	22,554	24,645	30,250	34,678	30,153	24,811	19,872	15,350
	${Ave}_{HI}$	366.2	484.5	533.3	588.7	715.4	780	726.1	627.3	511.2	486	419.7	437	374.6
42	${Num}_{HH}$	31,695	19,867	31,463	35,361	42,782	43,115	45,617	54,674	62,435	53,703	46,161	42,694	34,481
	${Ave}_{HI}$	318.1	441	577.9	632.7	552.2	710.4	658.9	470.8	424.5	411.2	383.3	310	346.7
43	${Num}_{HH}$	45,906	30,910	43,532	43,500	50,711	50,850	55,452	64,968	73,269	62,933	56,063	48,933	41,322
	${Ave}_{HI}$	315.1	484.6	546.9	584.1	620.2	584.8	706.9	581.6	465.5	516.5	462.8	299	330.8
44	${Num}_{HH}$	33,447	19,404	29,354	30,032	34,440	32,301	36,845	42,946	54,105	48,188	39,085	33,596	25,862
	${Ave}_{HI}$	345.2	468.6	558	611.5	604.7	632.9	640	515.6	443.7	398.8	332.7	319.1	302.8
45	${Num}_{HH}$	26,744	18,580	29,332	29,892	31,950	31,701	36,838	43,002	50,383	42,884	36,340	33,496	26,979
	${Ave}_{HI}$	312.2	450.6	505.2	570.7	596.1	665.3	648	512.3	467.2	370.2	317	329.5	235
46	${Num}_{HH}$	48,248	30,222	42,405	43,981	52,033	53,577	57,740	68,080	72,965	64,279	57,994	59,669	54,432
	${Ave}_{HI}$	386.5	402.9	507	517.2	610	586.8	617.8	461.1	380	366.5	371.4	246.8	226
47	${Num}_{HH}$	42,380	31,472	42,094	44,028	52,544	44,642	46,882	48,077	58,858	32,168	30,335	30,054	23,314
	${Ave}_{HI}$	373.3	393.8	497.7	480.8	494.6	532.3	497.1	350.2	454.5	399.4	273.8	337.9	241.1

Note: Age structure is divided into thirteen cohorts. Prefectural codes 1–47 correspond to Japan’s 47 prefectures (see Table A2 for details). Here, ${Num}_{HH}$ denotes the number of households, and ${Ave}_{HI}$ indicates the average annual household income (units of JPY 10,000).

—detailing age cohort household counts and corresponding average annual incomes—to provide household-level background for our ensuing discussion of HCF’s regional heterogeneity. The spatial distribution of age cohorts exhibits marked variability: for example, Tokyo contains 793,446 households in the 0–29 s and still over 549,980 in the 50–54 s, whereas Tottori records just 12,112 and 17,082 households in those same cohorts, respectively. These disparities reflect divergent local population structures and highlight potential scale effects on regional HCF. Moreover, cohort-level affluence also differs substantially across prefectures: Tokyo’s average household incomes span JPY 385.1–774.0 × 10⁴ across the thirteen cohorts, compared with Kagoshima, which spans only JPY 226.8–617.8 × 10⁴, and Okinawa, JPY 241.1–532.3 × 10⁴. Such income heterogeneity implies varying consumption capacities and energy-use profiles, which can materially influence HCF. Together, these demographic and economic differentials demonstrate the necessity of subnational HCF research: effective carbon mitigation strategies must be tailored to local cohort structures and income dynamics rather than predicated solely on national aggregates.

3.2.1. Overall HCF by Prefecture

Japan’s total HCF exhibits pronounced spatial heterogeneity (Figure 4). A small number of metropolitan prefectures account for disproportionately high emissions, while most less advanced areas remain low-emission and dispersed. Specifically, the Tokyo Metropolitan (e.g., Tokyo and Kanagawa) and the Kansai regions (e.g., Osaka and Hyogo) form the two primary high-HCF hubs, reflecting the synergistic coupling of high population density and affluent consumption. Likewise, the Chubu region (e.g., Aichi and Mie) constitute an industrialized zone whose HCF markedly exceeds that of similarly populated regions, underscoring the joint impact of durable goods demand and embedded supply chain transportation. Notably, Hokkaido—despite its sparse population—ranks among the highest-emission prefectures, driven predominantly by elevated heating energy requirements in cold climates.

Japan’s prefectural HCF comprises two core components: essential and discretionary expenditures (Figure 5). Essential expenditures—food, utilities (electricity and gas), healthcare and social security, and education—define the HCF’s lower bound. Food-related emissions, the most frequent and wealth-sensitive category, exhibit clear regional clustering: Tokyo’s food emissions are about 1.25 times those of Okinawa. Utility emissions—dominated by electricity consumption and influenced by both climatic extremes and housing insulation quality—vary markedly across prefectures, with Hokkaido at 1.12 tCO₂ per household and Okinawa at 1.23 tCO₂. Although healthcare and education together contribute only 0.06–0.27 tCO₂, their inelastic demand implies that mitigation must target the supply side—expanding telemedicine, online education, and digital public-service delivery can uphold service quality while curbing emissions. In the discretionary spending category, private transport offers the greatest mitigation potential: Tokyo’s extensive public transit network limits household transport emissions to 0.56 tCO₂, whereas car-dependent prefectures in Tohoku (e.g., Akita) and Kyushu (e.g., Oita) approach 1 tCO₂. Durable goods and household chemicals drive elevated emissions across the Chubu manufacturing corridor (e.g., Aichi, Shizuoka), reflecting supply-chain spillover effects on consumption upgrading. Moreover, personal services emissions demonstrate an inverse income–emission relationship: Tokyo records 0.32 tCO₂, whereas low-income Aomori records 0.42 tCO₂, underscoring the carbon-saving impact of digitalized services and dematerialized consumption in metropolitan areas.

3.2.2. Age-Specific HCF Across Prefectures

Although the inverted U-shaped per-household HCF profile persists at the prefectural level, significant regional divergences emerge (Figure 6). First, the age at which prefectural HCF peaks varies with local affluence: in high-gross-regional-product (GRP) prefectures (i.e., prefectures with a high GRP per household (GRP per household is reported in Table A1)) (e.g., Tokyo and Aichi), the maximum shifts to the 60–64 s, whereas in low-GRP regions (e.g., Aomori, Tottori) it occurs much earlier, in the 45–49 s. Long-term economic prosperity in high-GRP prefectures enables households to maintain elevated consumption of energy-intensive goods and services (e.g., tourism, healthcare, entertainment) into later life, thereby delaying the HCF peak. In contrast, in low-GRP prefectures, rising incomes among middle-aged cohorts, alongside the increased demand for improved living standards due to expanding household sizes, drive rapid consumption growth, precipitating an earlier apex. Second, certain prefectures exhibit a late-life rebound in HCF, which falls into two distinct typologies. In high-aging, low-GRP locales (e.g., Akita, Yamagata), sluggish economic growth, low energy efficiency, and a high proportion of single-elderly households drive up spending on essential services—such as healthcare and transportation—boosting per-household emissions. Conversely, in low-aging, high-GRP regions (e.g., Chiba, Saitama), robust economies underpin energy-intensive eldercare infrastructures and high-end residential services, along with increased tourism and cultural consumption, similarly precipitating a resurgence in HCF.

Spatial hot spots of HCF by age structure exhibit pronounced heterogeneity. In low-GRP prefectures, younger cohorts record elevated HCF; for example, among the 0–29 s, per-household HCF in Aomori and Wakayama reaches 3.66 tCO₂ and 3.85 tCO₂, respectively, compared with 1.63 tCO₂ in Tokyo and 2.01 tCO₂ in Aichi. This disparity likely reflects constrained transport infrastructure, limited consumer choice, and underdeveloped logistics in low-GRP regions, which together drive more carbon-intensive consumption. Although elevated HCF persists into middle age, the gap with high-GRP prefectures narrows over the life course: in the 40–44 s, Akita’s per-household HCF averages 5.60 tCO₂—1.50 times that of Chiba (3.73 tCO₂)—but by the 55–59 s, this ratio declines to 1.19. As households age, spending shifts from discretionary consumption to home comfort and energy efficiency investments, thereby reducing interregional HCF disparities. However, among the oldest old cohort, HCF in high-GRP prefectures rises markedly and the gap with low-GRP regions widens again: Shiga’s per-household HCF reaches 4.45 tCO₂—1.77 times Saga’s 2.52 tCO₂—whereas in the 45–49 s, the ratio is only 0.88. This resurgence reflects the super-aged population’s greater demand for energy-intensive eldercare services—such as personalized health management and specialized leisure—in affluent prefectures, thereby driving higher HCF.

3.2.3. Projected Trends in Prefecture-Level HCF

We used prefecture-level demographic projections in Japan to estimate HCF for all 47 prefectures over 2020–2050. Prefectures were ranked by their 2020 population aging rates and divided into tertiles. From each tertile, we selected one representative prefecture—Akita (higher aging rate), Okayama (mid-tertile), and Tokyo (lower aging rate)—for detailed case studies (Figure 7). We then analyzed HCF trajectories across four age cohorts (0–29 s, 30–59 s, 60–84 s, and 85 s–). Trends for all prefectures are depicted in Figure A2.

While population aging is generally exerting a mitigating effect on HCF, the magnitude of this effect varies substantially across prefectures. Between 2020 and 2050, total HCF is projected to decline by 45% in Akita (from 1.89 to 1.04 MtCO₂) and by 18% in Okayama (from 3.12 to 2.55 MtCO₂), yet to rise by 5% in Tokyo (from 26.06 to 27.53 MtCO₂). Notably, the carbon reduction effect of aging is concentrated among the young and middle-aged cohorts—and becomes more pronounced as aging advances. For instance, by 2050, the 0–29 s HCF is projected to decline by 9% in Tokyo, 33% in Okayama, and 58% in Akita, relative to 2020 levels. These divergent trajectories reflect the interaction between demographic shifts and regional economic structure. In low-GRP prefectures such as Akita and Okayama, accelerated aging amplifies labor shortages and erodes economic dynamism, leading to lower incomes and suppressed consumption demand among younger households. Conversely, Tokyo’s high-GRP economy—characterized by a diversified labor market, advanced educational infrastructure, and abundant urban amenities—continues to attract and retain younger cohorts, thereby attenuating the aging-induced decline in HCF among its young and middle-aged households.

Even within the elderly population, the HCF trajectories of the 60–69 s and 85 s– cohorts diverge substantially. In Akita, HCF for the 60–69 s declines steadily from 0.26 MtCO₂ in 2020 to 0.14 MtCO₂ by 2050. In contrast, Tokyo’s 60–69 s cohort exhibits an initial increase from 3.86 MtCO₂ in 2020 to 5.59 MtCO₂ in 2035, before decreasing to 4.71 MtCO₂ in 2050. For the 85 s–, Akita follows a similar rise–fall pattern—0.09 MtCO₂ in 2020, 0.12 MtCO₂ in 2035, and 0.11 MtCO₂ in 2050—whereas Tokyo’s HCF for the same cohort grows continuously from 1.14 MtCO₂ to 1.87 MtCO₂ over the same period. These contrasting trajectories reflect differences in demographic composition and consumption structures. Tokyo’s robust economy and comprehensive social welfare systems attract both long-term retirees and in-migrants, thereby expanding its elderly consumer base. Moreover, for the super-aged 85 s– cohort, high-GRP prefectures such as Tokyo provide extensive medical infrastructure and advanced eldercare services, which not only enhance health management but also drive expenditures across multiple sectors. For example, utility-related emissions increase from 0.49 MtCO₂ to 0.80 MtCO₂ between 2020 and 2050, sustaining the upward trend in HCF among the oldest age group.

3.2.4. Decomposition of Spatial HCF Inequalities

Applying the Dagum–Gini approach, we quantified spatial disparities in HCF across Japan’s prefectures by age cohort (Figure 8). Aggregate spatial inequality exhibits a pronounced U-shaped trajectory over the life course: the overall Gini coefficient peaks at 0.144 in the 0–29 s, declines steadily to a nadir of 0.088 in the 65–69 s, and then rises again to 0.117 in the 85 s–. Significantly, the balance between intragroup and intergroup inequality shifts with age: intragroup disparity contracts from 0.047 in the 0–29 s to 0.026 in the 80–84 s, whereas intergroup disparity expands from 0.014 to 0.040 over the same age span.

Intragroup disparities in HCF exhibit a distinct age-dependent trajectory. Among high-income households, inequality peaks at midlife—reaching 0.169 for the 40–44 s—while both the youngest (0–29 s) and the oldest (85 s–) cohorts display substantially lower heterogeneity. This pattern likely reflects divergent midlife expenditure demands driven by life-stage factors such as children’s education and multigenerational caregiving. By contrast, middle-income households display a steady decline in disparity—from 0.091 in the 30–34 s to 0.058 in the 60–64 s—indicative of converging consumption patterns and lifestyle preferences over the life course; however, this convergence reverses in the 85 s–, where disparity rises again to 0.115. Low-income households show the greatest disparities at both age extremes: a coefficient of 0.145 in the 0–29 s—comparable to middle-income group and likely driven by heterogeneous living arrangements and early career income instability—and an increase to 0.123 in the 85 s–, reflecting amplified consumption heterogeneity among the elderly poor due to variations in regional healthcare policies and local socio-economic support structures.

Intergroup disparities elucidate the nuanced trajectory of HCF across income pairings. The high–middle disparity is most pronounced in early life—registering 0.141 in the 0–29 s—and persists into mid-adulthood before converging thereafter. Between the 50–54 s and 75–79 s, this coefficient declines to approximately 0.10, underscoring a pronounced convergence in consumption behaviors between these income strata during later life stag-es. The high–low gap remains pronounced—around 0.142 in both the 0–29 s and 30–34 s—before narrowing modestly during middle age, albeit with intermittent fluctuations. The middle–low disparity reaches its apex at 0.155 in the 0–29 s, even exceeding the high–low gap. This likely reflects high-income households’ focus on savings and investment rather than on immediate consumption in early life stages, in line with Franco Modigliani’s life cycle hypothesis [63], which suggests individuals save during high-income periods to smooth consumption during lower-income phases. In contrast, middle-income households, once they achieve early income stability, tend to increase discretionary spending on durable goods (e.g., vehicles, housing upgrades), thus widening the consumption gap with low-income households due to differences in financial flexibility and access to credit. In advanced age, the middle–low gap diminishes but resurges to 0.126 in the 85 s–, suggesting that heterogeneous eldercare arrangements and unequal access to medical resources among middle- and low-income elderly households drive renewed differentiation in HCF.

3.3. Research Limitations

National-level HCF is derived using the 2020 IO table, whereas prefectural estimates are based on the 2005 MRIO table, the most recent publicly available dataset. This discrepancy in data timing introduces a temporal inconsistency that may introduce bias in cross-scale comparisons, particularly in light of substantial changes in economic structure and consumption patterns. To mitigate this concern, we emphasize relative magnitudes and temporal patterns across age cohorts rather than absolute emission levels. When updated prefectural MRIO data become available, we will recalibrate our model to enhance comparability between national and subnational estimates. Moreover, we acknowledge that assuming household consumption patterns, production technologies, and carbon intensities remain fixed at their 2005 levels represents a simplifying assumption that limits the validity of our long-term emissions forecasts. In reality, these parameters evolve over time in response to technological innovation, policy interventions, and shifts in consumer behavior—through energy source shifts, efficiency gains, supply chain restructuring, and changing consumption preferences—all of which could materially alter future emission pathways. By holding them constant, this study intentionally isolates the demographic drivers of HCF, treating our projections as counterfactual scenarios under unchanging technological and behavioral conditions rather than precise forecasts. Nonetheless, it is important to acknowledge that this modeling framework may exert a latent influence on our principal findings, including HCF estimations and regional heterogeneity assessments. In line with standard economic development trajectories, the carbon intensity of Japan’s final demand will gradually decline, driven by enhanced energy efficiency and a growing share of renewable energy. However, using 2005 carbon intensities to estimate prefectural-level HCF is liable to overstate absolute emission levels, since all prefectures likely followed a similar downward trajectory. Moreover, technological progress and industrial upgrading exhibit marked spatial asynchrony: highly service-oriented or early decarbonizing regions (e.g., Tokyo) tend to achieve larger reductions in carbon intensity than manufacturing-dominated areas, and thus, under a common application of the 2005 carbon intensities, interregional emission differences are objectively compressed, rendering our measure of regional heterogeneity conservative. We have therefore clarified that, by fixing carbon intensities, our model excludes broader decarbonization trends and serves as a theoretical exercise to highlight the independent effect of population dynamics. In subsequent studies, we will further commit to employing dynamic scenario analyses to more accurately capture evolving decarbonization pathways and concurrent shifts in household lifestyles.

4. Conclusions and Policy Implication

This study employs a two-tier framework—integrating the EEIO model with national IO and prefectural MRIO tables—to quantify HCF across age cohorts and income deciles in Japan. Meanwhile, CF-Gini and Dagum–Gini reveal substantial carbon inequality within life stages, driven by evolving demographic shifts and regional economic contexts. The main findings of this study are as follows:

• Japan’s per-household HCF exhibits an inverted U-shaped age profile, peaking at 50–54 s (2.16 tCO₂). Disaggregating HCF by age and income reveals that, although HCF generally increases with income, it falls at both lower and upper ends of the income spectrum within the 70–84 s. Carbon inequality reveals a U-shaped pattern, with higher CF-Gini coefficients in the younger and elderly groups.
• While the inverted U-shape in per-household HCF holds across prefectures, its peak shifts: 60–64 s in high-GRP prefectures (e.g., Tokyo) versus 45–49 s in low-GRP prefectures (e.g., Aomori). Moreover, late-life HCF rebounds in two prefecture types: high-aging, low-GRP prefectures (e.g., Akita) via essential expenditures, and low-aging, high-GRP prefectures (e.g., Chiba) via energy-intensive eldercare services.
• Population aging’s long-term impact on HCF exhibits marked regional heterogeneity. In low-GRP prefectures, HCF among young and middle-aged cohorts declines obviously from 2020 to 2050 (e.g., a 58% reduction in Akita’s 0–29 s). In high-GRP prefectures, these declines are marginal, while HCF among the elderly rises markedly (e.g., 44% growth in Tokyo’s 60–69 s).
• Aggregate spatial inequality in HCF exhibits a U-shaped pattern. Intragroup disparities decline with age. In contrast, intergroup disparities increase over the life course: the high–low and middle–low disparities, initially 0.142 and 0.155, narrow to approximately 0.10 in midlife, then the middle–low disparities rebound to 0.126 in the 85 s–, reflecting unequal access to medical and eldercare services.

Our quantitative analysis indicates that a uniform mitigation strategy cannot simultaneously achieve Japan’s 2050 net-zero goal and alleviate the carbon inequality pressures intensified by population aging. Instead, Japan must urgently implement differentiated decarbonization policies—calibrated by age cohort, income bracket, and regional context—across both national and subnational levels. Specifically, the main results suggest the following policy implications:

First, integrating age and income dimensions into climate mitigation strategies is essential. The Japanese government should incorporate a life cycle perspective and stratified income policies into its Nationally Determined Contributions (NDCs (national efforts to curb emissions and mitigate climate change are vital for achieving the Paris Agreement’s long-term GHG emission reduction objectives)), thereby establishing a multidimensional governance framework that horizontally addresses all age cohorts and vertically spans income tiers. At the statistical and monitoring level, a comprehensive HCF database should be developed by integrating census data and NSFIE household expenditure surveys, with emission baselines updated across five-year age bands and detailed income brackets. Additionally, structural decarbonization initiatives should prioritize the mid-life high-emission cohort (30–64 s, incomes 600–) by deploying green finance incentives and differentiated carbon pricing to shift consumption toward low-carbon alternatives. One option is a targeted carbon tax rebate scheme, in which a portion of paid carbon taxes is refunded to those who meet predefined low-carbon investment or consumption criteria; rebates are strictly earmarked for acquiring or retrofitting high-impact substitutes (e.g., electric vehicles, energy-efficient appliances). This preserves fiscal revenue while directly incentivizing emission reduction at the individual level. Simultaneously, a green consumption tax deduction would permit middle- and high-income consumers to offset personal income or value-added taxes on qualifying energy-saving products, thereby lowering the net cost of low-carbon purchases. Together, these instruments curb carbon-intensive spending among high-income, middle-aged cohorts; reduce HCF; and uphold equity through a balanced combination of rebates and tax deductions. For younger and low-income elderly groups, the government should adopt a welfare synergy approach that links access to green public services—such as low-carbon rental housing and community energy networks—with social security subsidies, thereby mitigating the regressive impacts of carbon reduction policies on vulnerable populations.

Second, the Japanese government must implement a geographically differentiated decarbonization framework to address regional and life cycle disparities. In high-GRP prefectures like Tokyo, household consumption is concentrated in energy-intensive high-end services. As population aging progresses, emissions may not decline as quickly as in low-GRP prefectures but may instead rise due to increased demand from the elderly for personalized services. In high-GRP prefectures, efforts should concentrate on key sectors—eldercare and cultural services—by pursuing an age-friendly low-carbon transition that accelerates the development of zero-emission medical facilities and smart elderly communities, thereby driving the low-carbon upgrading of the service industry. By contrast, in low-GRP prefectures such as Akita, constrained infrastructure and sparse service availability exacerbate HCF. Hence, decarbonization efforts must prioritize both the expansion of essential services and the realignment of consumption patterns to support low-carbon living. First, sustainable transport should be strengthened through dedicated investments in public transit and targeted subsidies for electric shared mobility schemes, thereby reducing reliance on private vehicles. Second, energy efficiency retrofits—especially in long-occupied, elderly-dominated households—must be accelerated to achieve direct HCF reductions. Complementing these supply-side measures, welfare-oriented subsidies for low-income elderly cohorts can lower barriers to decarbonized consumption: under such schemes, eligible seniors would receive carbon vouchers redeemable for public transit fares, energy efficient appliances, or eco-friendly foods, and benefit from a senior mobility subsidy offering discounted access to public transit, shared mobility, and low-carbon healthcare services. Together, these interventions decrease living costs, foster sustainable behaviors, and enhance health and well-being—thereby protecting vulnerable populations while advancing overarching carbon reduction targets.

Lastly, a comprehensive, multi-scalar governance framework is imperative to redress spatial inequalities in Japan’s HCF. At the intragroup level, policy interventions must be both age- and income-sensitive: high-income households in the 40–44 s should face targeted measures to curb emissions associated with children’s education, intensive eldercare, and high-return investments; middle-income households in the 85 s– require enhanced long-term care insurance and subsidized community services to offset late-life healthcare shocks; and low-income households demand expanded youth rental housing, public transport subsidies, in-home energy assistance for seniors, and equitable primary healthcare to mitigate extremes of demographic polarization. Intergroup reforms should prioritize cohorts with the greatest disparities—particularly the 0–29 s—by deploying income-indexed green consumption vouchers and educational subsidies to instill low-carbon behaviors early. Meanwhile, to address the emissions rebound associated with end-of-life care among the 85 s–, we propose that Japan’s central government pool long-term care insurance funds nationally and impose binding carbon baselines on elderly healthcare. This approach mirrors China’s interprovincial pairing and ecological compensation mechanism, whereby wealthier, technologically advanced provinces transfer expertise and fiscal resources to less developed regions through a unified emissions-trading and compensation framework, thereby equalizing abatement capacities and coordinating pollution control [64]. Similarly, integrated electricity–hydrogen markets have shown that harmonized cross-regional standards, uniform subsidy schemes, and shared infrastructure can remove policy and technical barriers, align regulatory frameworks, and improve overall market efficiency [65]. Under central oversight, aggregated long-term care funds across prefectures and uniform carbon baselines would harmonize fiscal support and advance decarbonization in end-of-life care. Subnational authorities could then implement these guidelines by defining differentiated emission targets tailored to local demographics, industrial structures, and energy endowments, and enforcing mandatory mitigation measures with accompanying fiscal incentives. In summary, by establishing a cross-regional platform for sharing eldercare infrastructure, Japan can streamline facility use, eliminate redundant construction and its associated energy demand, ensure equitable, high-quality eldercare, and thereby strengthen the coherence and efficiency of its decarbonization strategy.