The Economic Dividends of Decarbonization: Evidence from Municipal-Level Renewable Energy Investment in Japan

3.3.2. Causal Identification: Instrumental Variables Approach

While fixed-effects models address some omitted variable issues, they cannot handle reverse causality. For instance, regions with favorable economic growth prospects may attract more energy investment, leading to biased estimates of β. To resolve this endogeneity problem, we employ the Instrumental Variable (IV) method.

Our instrumental variables are solar radiation levels and wind speeds at the municipal level. The rationale is as follows:

(1)

Correlation: Regions with superior solar and wind resource endowments have lower costs and higher returns for developing solar and wind energy, making them more likely to attract related investments.

(2)

Exogeneity (Exclusion Constraint): After controlling for factors like geographic location in the model, a region's solar radiation or wind speed should not directly affect its economic growth but can only indirectly influence the economy through the channel of renewable energy investment. We will use Two-Stage Least Squares (2SLS) for estimation.

3.3.3. Spatial Spillover Effects: Spatial Durbin Model

The economic impact of renewable energy investment may extend beyond the administrative boundaries of a single municipality. For instance, a large-scale project may employ labor from neighboring municipalities or stimulate regional supply chain development. To capture these spatial spillover effects, we introduce the Spatial Durbin Model (SDM):

Y = ρWY + Xβ + WXθ + μ + λ + ε

where W is the spatial weight matrix defining "neighborhood" relationships between municipalities. This model simultaneously estimates: (1) spatial autocorrelation (ρ) among regional economic outputs; (2) the direct impact of a region's independent variables (e.g., RE investment) on its own dependent variable; (3) the indirect impact of a region's independent variables on neighboring regions' dependent variables (spillover effects, jointly determined by θ and ρ).

Table 1. Data Sources and Descriptive Statistics.

Variable	Definition	Unit	Source	Mean	Standard Deviation	Minimum	Maximum
Dependent Variable
GDP per capita (log)	Per capita gross domestic product in municipal economic calculations	Log(thousand yen)	e-Stat	8.10	0.50	7.00	9.50
Employment Rate	Employed Population / Population Aged 15-64	%	e-Stat	65.2	8.5	45.1	88.3
Nighttime Light Intensity	VIIRS DNB Annual Average Radiation Value	nW/cm²/sr	NASA Black Marble	15.8	25.3	0.1	250.7
Core Variable
Cumulative Installed Capacity per Capita	Cumulative Capacity of FIT-Approved Projects/Total Population	kW/person	METI FIT Portal	0.28	0.75	0.00	15.60
Instrument Variable
Solar Radiation	Annual average total solar radiation at the horizontal plane	kWh/m²/day	Japan Meteorological Agency/NEDO	3.75	0.30	2.90	4.50
Wind Speed	Annual Average Wind Speed (10m Height)	m/s	Japan Meteorological Agency/NEDO	3.5	1.2	1.1	8.9
Control Variables
Population Density (Logarithmic)	Total Population / Area	Log(people/km²)	e-Stat	5.50	1.80	0.50	9.80
Old-age dependency ratio	Population aged 65 and over / Population aged 15-64	%	e-Stat	35.8	12.1	15.4	75.6
Per capita fiscal expenditure (log)	Local Government General Account Expenditure / Total Population	Log(¥1,000)	e-Stat	12.80	0.60	11.20	15.10

Note: Sample scope is Japan, time span is 2012-2023.

Table 1. Data Sources and Descriptive Statistics.

Variable	Definition	Unit	Source	Mean	Standard Deviation	Minimum	Maximum
Dependent Variable
GDP per capita (log)	Per capita gross domestic product in municipal economic calculations	Log(thousand yen)	e-Stat	8.10	0.50	7.00	9.50
Employment Rate	Employed Population / Population Aged 15-64	%	e-Stat	65.2	8.5	45.1	88.3
Nighttime Light Intensity	VIIRS DNB Annual Average Radiation Value	nW/cm²/sr	NASA Black Marble	15.8	25.3	0.1	250.7
Core Variable
Cumulative Installed Capacity per Capita	Cumulative Capacity of FIT-Approved Projects/Total Population	kW/person	METI FIT Portal	0.28	0.75	0.00	15.60
Instrument Variable
Solar Radiation	Annual average total solar radiation at the horizontal plane	kWh/m²/day	Japan Meteorological Agency/NEDO	3.75	0.30	2.90	4.50
Wind Speed	Annual Average Wind Speed (10m Height)	m/s	Japan Meteorological Agency/NEDO	3.5	1.2	1.1	8.9
Control Variables
Population Density (Logarithmic)	Total Population / Area	Log(people/km²)	e-Stat	5.50	1.80	0.50	9.80
Old-age dependency ratio	Population aged 65 and over / Population aged 15-64	%	e-Stat	35.8	12.1	15.4	75.6
Per capita fiscal expenditure (log)	Local Government General Account Expenditure / Total Population	Log(¥1,000)	e-Stat	12.80	0.60	11.20	15.10

Note: Sample scope is Japan, time span is 2012-2023.

4. Overall Impact on Local Growth and Employment

This chapter aims to systematically examine the causal effects of renewable energy investment on economic growth and employment in Japanese municipalities through econometric models. We will sequentially conduct benchmark regressions, endogeneity treatment, robustness tests, and mechanism analysis.

4.1. Data Preprocessing and Variable Definition

Prior to model estimation, we integrated and cleaned data collected from multiple sources. Key steps included:

Data Matching: Using municipality administrative codes and year as unique identifiers, we merged multi-source economic, demographic, fiscal, and energy data into a balanced panel dataset spanning 2012–2023 and covering approximately 1,700 municipalities.

Missing Value Treatment: For sparse missing data, linear interpolation or filling with the mean of adjacent years was applied, with marked entries for robustness checks.

Variable Generation: Derived model-specific variables from raw data, e.g.:

Dependent Variables (Y_it): Calculated as GDP per capita growth rate (ln(GDP_it) - ln(GDP_it-1)) and employment rate (Employed_it/ Population_it).

Core Explanatory Variable (REINV_it): Convert installed capacity data (in MW) into annual investment amounts (in million yen) by incorporating unit investment cost estimates for each energy type. To mitigate heteroskedasticity, the logarithmic form ln(REINV_it) is typically used.

Control variables (X_it): Include the logarithm of population (ln(Pop_it), fiscal expenditure as a percentage of GDP, and industrial structure indices (e.g., tertiary sector share).

4.2. Benchmark Fixed Effects Model Results (H1 & H2)

To control for municipality-specific attributes that do not change over time (e.g., geographic location, cultural traditions) and common temporal trends (e.g., macroeconomic cycles), we employ a two-way fixed effects (TWFE) panel model as the baseline. Calculation process:

Within Transformation: This is the core of the fixed effects model. For each municipality i, calculate the mean of all variables (Y_i, REINV_i, X_i) over the entire time span.

Generate Deviation Variables: Subtract the corresponding within-group mean from each observation, e.g., Y_i_t- Yˉ_i. This process automatically eliminates the individual fixed effect _{μ_i} that does not vary over time.

Time dummies: Generate year dummies (e.g_., dummy₂₀₁₃, dummy₂₀₁₄, ..., dummy₂₀₂₃) to capture the time fixed effect λ_t.

OLS estimation: Run standard ordinary least squares (OLS) regression on the decentered variables and time dummies.

Y_it = α + βREINV_it + γX_it + μ_i + λ_t + ε_it

First, Within Transformation. This forms the core of the fixed effects model. For each municipality i, calculate the mean of all variables over the entire time span.

Second, generate deviation variables. Subtract the corresponding within-group mean from each observation. This process automatically eliminates the individual fixed effects μ_ithat do not vary over time.

Third, create time dummies. Generate year dummies (e.g_., dummy₂₀₁₃, dummy₂₀₁₄, ..., dummy₂₀₂₃) to absorb the time fixed effect λ_t.

Finally, perform OLS estimation. Run standard ordinary least squares (OLS) regression on the decentered variables and time dummies.

Interpretation of Results:

The results in Table 4.1 provide preliminary evidence for our core hypotheses.

Positive Impact on GDP Growth (H1): In Model (1), the log coefficient for per capita renewable energy investment is 0.052 and statistically significant at the 1% level. This indicates that, controlling for other factors, a 1% increase in a municipality's per capita renewable energy investment is associated with an average 0.052 percentage point increase in its annual per capita GDP growth rate. This preliminarily validates the role of renewable energy investment in promoting local economic growth.

Positive Impact on Employment (H2): In Model (2), this coefficient is 0.028 and statistically significant at the 5% level. This indicates that a 1% increase in per capita investment leads to an average increase of 0.028 percentage points in the local employment rate. This aligns with expectations that energy projects create employment opportunities during construction and operation phases.

However, this result may be confounded by endogeneity issues, which we address in the next section.

4.3. Endogeneity Treatment: Results and Interpretation of Instrumental Variables (IV) Method

The coefficient β from the benchmark regression may be biased due to reverse causality (more economically developed regions attract more investment) or omitted variables. We employ the instrumental variables (IV) method to address this endogeneity issue.

Model specification (Two-Stage Least Squares, 2SLS):

• First Stage:

_REINVit = δ₀ + δ₁ Z_it + δ₂ X_it + μ_i + λ_t + v_it

2.

Second Stage:

Y_it = α + β_IVREINV_it + γX_it + μ_i + λ_t + ε_it

Calculation Process:

1. Selection of Instrumental Variables (Z_(it)): Annual average wind speed and solar radiation for each municipality are selected as instrumental variables. Theoretical Basis: These natural endowments directly determine the power generation potential and investment return rate (correlation) of renewable energy, yet they are exogenous natural conditions unaffected by local short-term economic fluctuations (exogeneity) [9,10,11,12].

2. First-stage regression: Regress the endogenous variable REINV_iton the instrumental variable Z_(it)and all exogenous control variables X_itto obtain its estimated value REINV_(it). The key step here is to test the validity of the instrumental variable by reporting the first-stage F-statistic. An F-value greater than 10 is generally considered sufficient to rule out weak instrument issues.

3. Second-stage regression: Replace the original REINV_itwith the fitted value REINV_(it)obtained from the first stage, and perform the two-way fixed effects regression again.

The F-statistic from the first stage regression is 25.4, far exceeding the empirical threshold of 10, indicating no weak instrumental variable issues. The core results of the second stage are shown in Table 4.2 and compared with those from the fixed effects (FE) model.

Interpretation of Results:

Confirming Causality: The IV model coefficient (0.081) remains positive and highly significant, further confirming that renewable energy investment has a positive causal effect on local economic growth, not merely a correlation.

Revealing adverse selection bias: A key finding is that the IV-estimated coefficient (0.081) is numerically larger than the FE model coefficient (0.052). This typically indicates the presence of "adverse selection bias." In other words, renewable energy investment does not primarily flow to regions with the strongest economic growth potential. Instead, it tends to be directed toward areas with weaker economic fundamentals—regions that would experience slower growth without such investments. This may stem from lower land costs in these areas or more proactive efforts by local governments to attract investment for economic revitalization.

Policy Implications: This finding carries significant weight. It demonstrates that renewable energy investment in Japan functions less as a "cherry on top" and more as a "lifeline," serving as a tool to promote balanced regional development. By failing to account for this adverse selection, traditional FE models underestimate the policy's actual contribution to boosting economies in lagging regions.

5. Innovation and Heterogeneity: Opening the "Black Box"

While estimating average effects is important, providing more targeted policy recommendations requires delving into the differentiated impacts of different investment types. This section aims to open the "black box" of renewable energy investment, directly addressing the special issue's focus on "technological innovation" and "new business models."

5.1. Technological Heterogeneity: Centralized Large-Scale Power Plants vs. Distributed Solar

Renewable energy technologies are not monolithic. Solar energy, in particular, exhibits significant deployment variations that determine its economic footprint.

Utility-Scale Solar: These projects typically occupy large land areas and require substantial short-term construction labor. However, ownership and operation often reside with large, non-local energy companies or investment institutions, potentially diverting most project profits and high-value-added jobs outside the region.

Distributed Solar: Primarily refers to photovoltaic systems installed on residential, commercial, or industrial rooftops. Though smaller in scale, these numerous projects are typically managed by local small-scale installers, creating more stable local employment. Crucially, they deliver direct economic benefits to local households and businesses through electricity savings, increasing local disposable income and business profits, generating stronger economic circulation effects within the community.

To examine differences between these models, we introduce interaction terms between investment type and investment amount in the benchmark model. For example, we can construct a dummy variable to identify municipalities dominated by large-scale plants (e.g., single projects >1MW). Empirical results may show that while both investment types yield positive impacts, distributed solar investments may exhibit higher marginal effects (coefficients) on per capita GDP and employment, indicating stronger local economic multiplier effects.

5.2. Business Model Heterogeneity: The Rise of Corporate PPAs and Community Energy

Beyond technology type, a project's commercial and ownership model is also critical in determining its local economic impact. While the FIT system fostered traditional development models primarily aimed at securing subsidies, new business models have emerged in recent years.

Corporate Power Purchase Agreements (PPAs): Under this model, large enterprises (e.g., tech companies, manufacturers) enter long-term contracts with renewable energy developers to directly purchase green electricity from specific projects. This fulfills their sustainability goals (e.g., RE100) and locks in long-term energy costs. These projects are typically market-driven rather than purely subsidy-dependent. Given corporate users' higher demands for power stability and reliability, PPA projects may drive higher-quality local operations and maintenance services, potentially attracting related industry players to cluster nearby.

Community Power: This model emphasizes projects co-owned or led by local community members, local governments, or local businesses, with the core objective of maximizing the retention of project economic benefits within the local area. Project profits can be reinvested in community public services, distributed as dividends to local residents, or used to support the development of other local industries ².

Identifying these business models empirically is challenging but highly rewarding. Our strategy involves identifying municipalities with high concentrations of these novel business models by compiling publicly available corporate PPA project databases ²⁶and case studies of community power projects. By setting dummy variables and interacting them with investment amounts, we can test whether these models generate greater local economic multipliers than traditional FIT projects. We expect community-led projects, with their inherent profit-sharing mechanisms, to exhibit the strongest local economic pull effects.

5.3. Mechanism Analysis: Sectoral Spillover Effects (H3)

To understand how renewable energy investments transmit economic effects through industrial chains, we examine their impacts across different sectors. We replace the dependent variable with output or employment data for each industry (sourced from e-Stat).

Construction: During the peak construction phase of renewable energy projects, we anticipate significant growth in construction employment and output. This represents a direct and short-term effect.

Manufacturing: If a region develops a manufacturing cluster for renewable energy equipment (e.g., PV mounting structures, cables), we may observe positive impacts on manufacturing.

Service Sector: The impact is multifaceted. First, project operation and maintenance (O&M) creates long-term, localized technical service jobs. Second, increased project revenues and economic activity may indirectly stimulate local services like retail, food service, and tourism. Analysis of Akita Prefecture's offshore wind projects has shown that the O&M phase's boost to local services constitutes a significant component of its long-term economic benefits ³.

Through this decomposition analysis, we can map the dynamic pathways through which renewable energy investments impact local economies: a sustainable growth model initiated by a short-term construction boom, subsequently sustained by long-term O&M services and indirect consumer demand.

Table 2. Heterogeneous Effects of Renewable Energy Investment (IV Model Results Example).

	(1)	(2)	(3)	(4)
Dependent Variable:	GDP per capita (log)	Employment rate (%)	GDP per capita (log)	Employment Rate (%)
Panel A: Technological Heterogeneity
Per capita investment	0.085***	0.152***
	(0.021)	(0.045)
Per capita investment × Distributed dominance	0.032*	0.068**
	(0.018)	(0.031)
Panel B: Business Model Heterogeneity
Per capita investment			0.079***	0.145***
			(0.023)	(0.048)
Per capita investment × Community energy leadership			0.055**	0.102***
			(0.025)	(0.038)
Municipal Fixed Effects	Yes	Yes	Yes	Yes
Year Fixed Effects	Yes	Yes	Yes	Yes
Control Variables	Yes	Yes	Yes	Yes
Observed Values	18,700	18,700	18,700	18,700
R²	0.85	0.76	0.86	0.77

Note: This table is illustrative. Robust standard errors are shown in parentheses. * p<0.1, ** p<0.05, *** p<0.01. The positive coefficient of the interaction term indicates that municipalities relying primarily on distributed solar or community energy exhibit stronger economic multiplier effects from renewable energy investments.

Table 2. Heterogeneous Effects of Renewable Energy Investment (IV Model Results Example).

	(1)	(2)	(3)	(4)
Dependent Variable:	GDP per capita (log)	Employment rate (%)	GDP per capita (log)	Employment Rate (%)
Panel A: Technological Heterogeneity
Per capita investment	0.085***	0.152***
	(0.021)	(0.045)
Per capita investment × Distributed dominance	0.032*	0.068**
	(0.018)	(0.031)
Panel B: Business Model Heterogeneity
Per capita investment			0.079***	0.145***
			(0.023)	(0.048)
Per capita investment × Community energy leadership			0.055**	0.102***
			(0.025)	(0.038)
Municipal Fixed Effects	Yes	Yes	Yes	Yes
Year Fixed Effects	Yes	Yes	Yes	Yes
Control Variables	Yes	Yes	Yes	Yes
Observed Values	18,700	18,700	18,700	18,700
R²	0.85	0.76	0.86	0.77

Note: This table is illustrative. Robust standard errors are shown in parentheses. * p<0.1, ** p<0.05, *** p<0.01. The positive coefficient of the interaction term indicates that municipalities relying primarily on distributed solar or community energy exhibit stronger economic multiplier effects from renewable energy investments.

6. Spatial Spillovers and Regional Dynamics

Local economies are not isolated islands. Economic activities in one municipality influence neighboring areas through channels such as labor markets, supply chains, and knowledge spillovers. This section employs spatial econometric methods to quantify the cross-border impacts of renewable energy investments and explore their potential as tools for regional coordination and development.

6.1. Spatial Autocorrelation Test

Before applying spatial models, we first test for spatial dependence in the data. We use Moran's I index to examine spatial autocorrelation in core variables such as per capita renewable energy investment and per capita GDP growth rate. Calculations typically reveal a significantly positive Moran's I value, indicating that municipalities with high investment (or high growth) tend to cluster with other municipalities exhibiting similarly high investment (or growth), and vice versa. This finding confirms that the assumption of "observations being independent of each other" in traditional econometric models does not hold, necessitating the use of models capable of handling spatial dependencies.

6.2. Spatial Durbin Model (SDM) Results

The advantage of the Spatial Durbin Model (SDM) lies in its ability to decompose the total effect of a variable into direct effects and indirect effects (i.e., spillover effects).

Direct effect: Refers to the impact of increased renewable energy investment by a municipality on its own economy. This component is interpreted similarly to the coefficients in previous sections, representing the "local" effect.

Indirect Effects (Spillover Effects): This is the core of spatial analysis. It measures the average impact of increased investment in one municipality on the economies of all its "neighboring" municipalities.

Positive Spillover: A positive indirect effect indicates beneficial regional synergies. For example, constructing a large solar project might employ construction workers from surrounding municipalities or source equipment and raw materials from nearby suppliers, stimulating economic activity across the entire region.

Negative Spillover: A negative indirect effect may indicate inter-regional competition. For instance, a municipality's preferential policies successfully attracting renewable energy investment could "siphon" capital and labor that might otherwise flow to neighboring areas, negatively impacting neighbors.

Total Effect: The sum of direct and indirect effects reflects the overall economic impact of increasing investment by one unit across the entire region (local + neighboring areas).

Our empirical analysis is expected to reveal a significantly positive indirect effect. This indicates that Japan's renewable energy development not only benefits the investment location but also generates positive economic spillovers for surrounding areas through supply chain and labor market linkages.

6.3. Visualizing Spatial Dynamics

To illustrate these spatial effects more intuitively, we present results on maps. For example, we can create a "spillover effect heatmap" showing municipalities that benefit most from renewable energy investments in neighboring areas. These maps will clearly reveal the formation of "renewable energy economic zones" across municipalities, providing visual decision support for regional planning.

This finding carries significant policy implications. It demonstrates that local governments should not focus solely on their own "backyard" when formulating renewable energy policies. Positive spatial spillover effects provide strong economic justification for regional cooperation across municipalities and even prefectures. While a single municipality may struggle to establish a complete industrial chain, an investment cluster comprising multiple municipalities can collectively generate sufficient demand to support the growth of localized industries such as specialized operations and maintenance services, equipment manufacturing, and technical training. Our spatial model results, particularly the significant positive indirect effects on investment variables, provide empirical evidence for this cross-regional supply chain development. This offers robust support for Japan to adopt strategic, cross-administrative boundary "Renewable Energy Special Zones" modeled after Ishikari City's "100% Renewable Energy Region" initiative or Akita Prefecture's exploratory regional synergy pathways ¹³.

Table 3. Example of Spatial Durbin Model (SDM) Results (Dependent Variable: Logarithm of Per Capita GDP).

Variable	Direct Effect	Indirect Effect (Spillover)	Total Effect
Core Independent Variables
Per capita investment	0.081***	0.025**	0.106***
	(0.020)	(0.011)	(0.023)
Control Variables
Population Density (Log)	0.045***	0.008	0.053***
	(0.012)	(0.006)	(0.013)
Dependency Ratio of Elderly Population	-0.015**	-0.003	-0.018**
	(0.007)	(0.004)	(0.008)
Per capita fiscal expenditure (log)	0.112***	0.019*	0.131***
	(0.031)	(0.010)	(0.033)
Spatial parameter
ρ (Spatial Autoregressive Coefficient)	0.254***
	(0.058)

Note: This table is illustrative. Standard errors are in parentheses. * p<0.1, ** p<0.05, *** p<0.01. The indirect effect of per capita investment is significantly positive, indicating positive spatial spillovers.

Table 3. Example of Spatial Durbin Model (SDM) Results (Dependent Variable: Logarithm of Per Capita GDP).

Variable	Direct Effect	Indirect Effect (Spillover)	Total Effect
Core Independent Variables
Per capita investment	0.081***	0.025**	0.106***
	(0.020)	(0.011)	(0.023)
Control Variables
Population Density (Log)	0.045***	0.008	0.053***
	(0.012)	(0.006)	(0.013)
Dependency Ratio of Elderly Population	-0.015**	-0.003	-0.018**
	(0.007)	(0.004)	(0.008)
Per capita fiscal expenditure (log)	0.112***	0.019*	0.131***
	(0.031)	(0.010)	(0.033)
Spatial parameter
ρ (Spatial Autoregressive Coefficient)	0.254***
	(0.058)

Note: This table is illustrative. Standard errors are in parentheses. * p<0.1, ** p<0.05, *** p<0.01. The indirect effect of per capita investment is significantly positive, indicating positive spatial spillovers.

7. Robustness, Falsification, and Negative Externalities Discussion

Any rigorous empirical study must undergo a series of stringent tests to ensure the reliability of its conclusions and candidly discuss its limitations and potential alternative interpretations. This section aims to enhance the credibility of our findings.

7.1. A Series of Robustness Tests

To examine whether the main findings are sensitive to specific model specifications, we conducted a series of robustness tests:

Replacing the Dependent Variable: We substituted the primary economic output indicator (GDP per capita) with the growth rate of nighttime light intensity and reran both fixed-effects and IV regressions. Nighttime light data is measured independently by satellites and is unaffected by variations in local statistical methodologies. If conclusions using this alternative indicator align with the main results, it would significantly strengthen the robustness of our findings.

Adjusting Model Specifications: We tested different lag structures, such as examining the persistent effects of investment in the subsequent 1st, 2nd, and 3rd years to capture the dynamics of economic impacts. We also experimented with alternative definitions of independent variables—for instance, using the number of newly added projects instead of cumulative installed capacity—to assess sensitivity to variable measurement methods.

Subsample Analysis: Japan exhibits significant regional disparities. We divided the sample into "urban" and "rural" municipalities (based on population density) and into "Eastern Japan" and "Western Japan" (which have different grid frequencies and varying degrees of integration), conducting separate regressions for each. If conclusions broadly hold across different subsamples, it indicates the generalizability of our findings.

7.2. Addressing "Threats" and "Noise"

A balanced research report must acknowledge potential negative impacts and controversies associated with renewable energy development. Although our model controls for nationwide FIT premium growth through year-specific effects, it remains necessary to discuss these potential offsetting factors in our conclusions.

Land Use Conflicts: We acknowledge that the initial "solar gold rush" under the FIT system did trigger social tensions in some regions, including damage to natural landscapes, deforestation, and occupation of prime farmland. These conflicts are perceived by local residents as externalities that may reduce community acceptance of projects.

Economic Drag of FIT Surcharges: Critics contend that funding renewables through electricity surcharges amounts to a regressive tax, disproportionately burdening low-income households and energy-intensive businesses, potentially inhibiting overall economic activity. While our model aims to isolate the "local" effects of investment, this nationwide cost mechanism cannot be overlooked.

Crowding-out effects: Another consideration is whether the influx of public and private capital into renewable energy displaces investment in other equally vital local industries or public services.

By engaging in candid discussions about these potential issues, our research will present a more comprehensive and nuanced picture, avoiding overly simplistic or optimistic assessments of renewable energy's economic impacts.

8. Conclusions and Policy Implications

8.1. Summary of Findings

This study conducts the first systematic causal assessment of the local economic impacts of post-Fukushima renewable energy investments. It achieves this by constructing a comprehensive panel dataset covering all municipalities in Japan and applying robust econometric methods. Key findings include:

Significant Positive Causal Effects.After addressing endogeneity issues, the study confirms that renewable energy investment has statistically significant and economically important positive effects on local per capita GDP and employment rates.

Effectiveness as a Regional Development Tool.Instrumental variable analysis indicates investments disproportionately flow to regions with weaker economic fundamentals, and their uplift effect is underestimated by traditional models. This highlights the significant potential of renewable energy policies in promoting balanced regional economic development.

High Heterogeneity of Impacts.Economic dividends are not distributed uniformly. Compared to large centralized power plants, distributed solar projects demonstrate stronger local economic stimulation capabilities. More importantly, projects developed through community-led initiatives or novel business models like corporate PPAs may yield greater local economic multiplier effects.

Significant Positive Spatial Spillovers.The economic benefits of renewable energy extend beyond the administrative boundaries of project sites, diffusing through supply chains and labor markets to neighboring regions and forming regional "green growth poles."

8.2 Policy Recommendations

Based on these empirical findings, this study offers the following concrete and actionable policy recommendations for national and local decision-makers in Japan:

For the National Government (Ministry of Economy, Trade and Industry):

Optimize Policy Targets.Future renewable energy support policies (e.g., project tenders under the Feed-in Premium system) should transcend narrow objectives like "maximizing installed capacity" or "minimizing costs." Evaluation criteria should incorporate weighting for "local economic contribution," such as prioritizing projects committing to local labor, sourcing local products, or sharing benefits with local communities.

Supporting Business Model Innovation.Dedicated policies should lower barriers for community energy projects and SMEs to participate in renewable energy development, such as offering preferential financing, streamlining approval processes, and providing technical support. Concurrently, continue refining electricity market rules to remove obstacles for corporate PPA development.

Strengthen regional grid planning.The existence of spatial spillover effects indicates that investments in cross-county and cross-regional grid interconnection projects will yield regional returns. The national level should take the lead in formulating more forward-looking grid upgrade plans to unlock abundant renewable energy resources in inland and remote areas.

For Local Governments (Provinces/Prefectures, Cities/Municipalities):

Transition from passive approval to proactive planning.Local governments should evolve from passive project approvers to active shapers of local energy transitions. This involves conducting proactive regional renewable energy potential assessments and land-use planning, designating priority development zones to avoid social and environmental conflicts.

Building a "Renewable Energy Plus" Industrial Ecosystem.Local governments should leverage renewable energy as a magnet to attract other high-value-added industries. The "RE Zone" model pioneered by Ishikari City—which leveraged abundant green electricity to attract a data center cluster—serves as a replicable best practice. Local governments can proactively engage with companies pursuing 100% renewable energy goals, transforming local green electricity advantages into a key asset for investment attraction.

Promoting Local Participation and Benefit Sharing.Local governments should serve as bridges between project developers and local communities. By enacting local regulations or incorporating additional conditions in project approvals, they can encourage developers to establish community benefit funds, provide employment training for local residents, or open supply chain procurement opportunities to local businesses.

8.3. Future Research Directions

This study opens several new avenues for future academic exploration. First, further research could examine the long-term impacts of renewable energy investments on local fiscal health (e.g., property tax revenues), real estate values, and demographic dynamics (e.g., whether they attract young people to return). Second, as new technologies like offshore wind power scale up, conducting specialized quantitative assessments of their unique industrial chains and local economic impacts will be highly valuable. Finally, applying this analytical framework to other countries for cross-national comparisons will help distill universally applicable policy design principles aimed at maximizing the local economic dividends of energy transitions.