## wpiea2021087-print-pdf

## Source details

**Canonical URL:** [wpiea2021087-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021087-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021087-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021087-print-pdf.pdf.json)

---

### I. Context and motivation
- In the first six months of 2020, 6.4 percent less carbon dioxide was emitted than in the same period in 2019 (Liu et al., 2021).
- The first half of 2020 saw an unprecedented decline in CO2 emissions—larger than during the financial crisis of 2008, the oil crisis of the 1979, or even World War II.
- By the end of 2020, the coronavirus pandemic had infected over 90 million people globally and killed almost 2 million, producing factory closures, massive job losses, and paralysis of large swathes of economic activity.
- To meet the 1.5°C target, global emissions would need to drop by the same exaggerated rate seen during the pandemic (7.6 percent per year) in every year for the next few decades (IPCC, 2019).
- World governments’ collective US$14 trillion fiscal response to the economic damage of the COVID-19 pandemic has concentrated on measures to address the health emergency and support households and firms stranded by the lockdowns (IMF, 2021).
- Advocacy for “build back better” stresses stewarding the global economy within limits set by nature (Rockström et al., 2017; Attenborough, 2020; Georgieva, 2020; Stiglitz, 2020; Gates, 2021; Carney, 2021).

### Contribution and methodological approach
- Contribution:
  - First study to directly estimate GDP effects of money spent to foster the transition to a zero-carbon, nature-friendly world across multiple green expenditure typologies.
  - Expands “green” expenditure beyond emissions-reduction to include nature-based negative emissions technologies (nature-based solutions, NBSs), biodiversity conservation and rewilding.
- Empirical design:
  - Investment multipliers estimated using factor-augmented panel vector-autoregressive (FAVAR) models with Bayesian estimation (Normal-Wishart conjugate priors, Minnesota-type priors).
  - Panel dimension exploits cross-country variation; factors extracted from many macroeconomic variables mitigate limited-information concerns.
  - IMF WEO one-year-ahead forecast of total investments used as an exogenous variable in energy specifications to mitigate shock-foresight issues.
  - Monte Carlo simulation: perform 20,000 parameter draws, discard first 10,000 burn-in, use 10,000 posterior draws; impulse responses computed for a time horizon of 5 years; median response and 16th/84th percentiles saved.
  - Identification via Cholesky scheme assuming spending reacts with a lag to GDP and GDP reacts contemporaneously to spending shocks.

### Key empirical findings — aggregate and comparative
- High-level finding:
  - Every dollar (private and public) spent on key carbon-neutral or carbon-sink activities can generate more than a dollar’s worth of economic activity: total increase in GDP is greater than the original increase in green spending.
  - Green spending multipliers appear significantly bigger and more long-lasting than non-eco-friendly spending.
- Policy implication:
  - In post-COVID-19 recovery design, investments in energy and land/sea use transitions are likely economically superior to supporting unsustainable energy and food production activities—enabling faster recovery while preserving ecological improvements achieved in 2020.

### Energy sector multipliers (exact cumulated multipliers preserved)
- Renewable (green) versus non-eco-friendly (fossil) energy investment multipliers (cumulated multipliers):
  - Impact:
    - Green (Renewable) Energy Investments Multiplier: 1.19*
    - Non-Eco-Friendly Energy Investments Multiplier: 0.65*
  - 1 Year:
    - Green: 1.20*
    - Non-Eco-Friendly: 0.64*
  - 2 Years:
    - Green: 1.19*
    - Non-Eco-Friendly: 0.62*
  - 3 Years:
    - Green: 1.17*
    - Non-Eco-Friendly: 0.59*
  - 4 Years:
    - Green: 1.14*
    - Non-Eco-Friendly: 0.55
  - 5 Years:
    - Green: 1.11
    - Non-Eco-Friendly: 0.52
  - Note: * denotes multipliers with credible intervals (16th–84th percentiles) that exclude zero.
- Merged-specification comparison (both investments included; exact figures preserved):
  - Impact: [A] Green 1.40*; [B] Non-Eco-Friendly 0.62*; Prob [A]>[B]: 0.92
  - 1 Year: Green 1.46*; Non 0.58*; Prob 0.94
  - 2 Years: Green 1.49*; Non 0.54*; Prob 0.94
  - 3 Years: Green 1.51*; Non 0.51; Prob 0.93
  - 4 Years: Green 1.53*; Non 0.48; Prob 0.92
  - 5 Years: Green 1.54; Non 0.47; Prob 0.91
  - At all horizons, more than 90 percent of the empirical distribution of the difference lies above zero, indicating the difference is positive with high probability.
- Interpretation and mechanisms:
  - A shock to renewable spending is more persistent; output response to renewable spending remains above zero beyond 5 years while the fossil-fuel response dies off by 5 years.
  - Renewables are more labor intensive and have higher domestic content than fossil fuels, generating larger direct and indirect employment and local ripple effects.
  - Example interpretation: a cumulated multiplier of 1.5 at three years indicates cumulative GDP increase is 1.5 times cumulative spending increase.
  - For non-eco-friendly spending, a dollar spent crowds out about 48 cents in the medium run (5-year multiplier 0.52); for renewables a dollar spent crowds in about 11 cents (5-year multiplier 1.11 implies crowding in of 11 cents relative to other components).

### Nuclear energy multipliers (sample-specific; exact figures preserved)
- Nuclear energy (non-renewable clean) cumulated multipliers (sample differs from other energy sets):
  - Impact: 4.11*
  - 1 Year: 3.97*
  - 2 Years: 3.88
  - 3 Years: 3.83
  - 4 Years: 3.80
  - 5 Years: 3.78
  - Note: * denotes multipliers with credible intervals (16th–84th percentiles) that exclude zero.
- Nuclear interpretation:
  - Nuclear spending is more persistent than fossil-fuel investment and tends to crowd in other investments.
  - Large near-term output effect—about six times larger than fossil fuel energy in the impact period—but loses statistical significance after two years.
  - Possible drivers: nuclear projects are frontloaded, intensive in skilled and unskilled labor during construction; average construction times for large recent reactors are 5.1 years; higher pay for nuclear workers relative to wind/solar may drive large near-term multipliers.
- Note: nuclear sample spans 1991 to 2017 for China, France, Japan, Korea, Canada and United States.

### Land use and ecosystem conservation multipliers (exact cumulated multipliers preserved)
- Impulse response dynamics:
  - A shock to conservation spending has a long-lasting effect, similar to green energy shocks, implying durable economic benefits plus mitigation and carbon-sink gains.
  - A spending shock supporting industrial farming activities is considerably shorter-lived and dissipates after 5 years.
- Cumulated multipliers (Table 5):
  - Impact: Green Land Use Multiplier -5.36 / Non-Eco-Friendly Land Use Multiplier 0.55*
  - 1 Year: Green -1.60 / Non-Eco-Friendly 0.85*
  - 2 Years: Green 1.45* / Non-Eco-Friendly 0.95*
  - 3 Years: Green 3.75* / Non-Eco-Friendly 0.96*
  - 4 Years: Green 5.45* / Non-Eco-Friendly 0.95
  - 5 Years: Green 6.67* / Non-Eco-Friendly 0.94
  - Note: * denotes multipliers with credible intervals (16th–84th percentiles) that exclude zero.
- Key quantitative finding:
  - "For every dollar spent in conservation, almost seven more are generated in the larger economy in the medium term."
- Why green land use multipliers are high:
  - Dataset composition: estimates rely on developing-country donor-financed programs that do not crowd out domestic resources.
  - Labor intensity and visitor-economy effects: conservation drives hospitality and tourism, especially in rural/coastal communities with higher propensities to spend.
  - Price effects and ecosystem services: limiting land for agricultural expansion lifts rural producer prices; protecting biodiversity underpins ecosystem services that create jobs and foster innovation.
- Contrast with industrial agriculture:
  - Multipliers for spending to support industrial agriculture are below one at every horizon—reflecting high mechanization, low domestic content (machinery, imported chemical inputs, fossil fuels, foreign-patented seeds), and concentrated global suppliers.
  - Policy implication: repurposing spending from unsustainable land uses toward labor-intensive, high-domestic-content sustainable land uses may yield important economic gains and aid a green recovery.

### Data coverage and construction (exact country lists and spans preserved)
- Energy investments (renewable and non-eco-friendly fossil): China, Japan, Korea, Canada, United States, Brazil, Indonesia, Mexico, Russia, Oceania group (Australia and New Zealand) and EA group (France, Germany and Italy); time span 2003 to 2019.
- Nuclear energy investments: China, France, Japan, Korea, Canada and United States; time span 1991 to 2017.
- Green land use spending: Burkina Faso; Burundi; Cambodia; Cameroon; Central African Republic; Chad; Ghana; Guatemala; Malawi; Mozambique; Niger; Senegal; Sierra Leone; Madagascar; Tanzania; Uganda; time span 1994 to 2008.
- Non-eco-friendly land use spending: China, Japan, Korea, Canada, United States, Australia, Chile, Indonesia, Mexico, New Zealand, Russia, South Africa, Colombia, Iceland, Israel, Kazakhstan, Norway, Switzerland, Turkey and Ukraine; time span 1997 to 2016.
- Other data construction:
  - All series transformed to real terms using the implicit GDP price deflator and normalized by dividing by real potential GDP.
  - Informational dataset: 12 series per country from IMF’s WEO and Thomson Reuters Datastream; four common factors extracted for FAVAR (uniformly used).

### Financing gaps and sectoral context (exact figures preserved)
- Clean energy context and 2019 investment shares:
  - Renewables (excluding large hydro) accounted for about one seventh of global generation (IRENA, 2019).
  - Nuclear energy accounted for about one-tenth (IEA, 2019a).
  - Nuclear capacity globally is estimated to have shrunk by a net 5GW in 2019.
  - Investment in nuclear energy in 2019: US$15 billion.
  - Investment in renewables in 2019: around US$282 billion (IEA, 2020a).
- Target gap to 2030 (BloombergNEF New Energy Outlook 2019 base-case):
  - Gross addition needed of some 2,836GW of new non-hydro renewable energy capacity by 2030.
  - Estimated cost: US$3.1 trillion over the decade.
- Ecosystem conservation finance gap:
  - Estimated current spending (2019) on biodiversity-associated goods: between US$124 and US$143 billion.
  - Estimated need: US$722-967 billion.
  - Paulson estimates represent 0.1-0.2 percent of 2019 global GDP.
  - Example cost to expand protected areas to 30 percent of Earth’s surface: between US$103 billion and US$178 billion per year; current investment in protected areas: US$24.3 billion (required cost represents 4-7 times current investment).
- Agri-food sector emissions:
  - The agri-food sector emits between 21-37 percent of greenhouse gases (IPCC, 2019); without policy action this share could raise to 50 percent of all global emissions by 2050 (IPCC, 2019; Willett et al., 2019).
- World Bank / agricultural support:
  - Countries producing two-thirds of the world's agricultural output spent US$600 billion per year in agricultural financial support on average from 2014 to 2016.
  - Out of US$300 billion in direct spending, only 9 percent explicitly supports conservation and 12 percent supports research and technical assistance.

### Robustness checks (exact alternative specification results preserved)
- Alternative lag structure (2 years) — Green Energy Investments Multiplier / Non-Eco-Friendly Energy Investments Multiplier:
  - Impact: 1.28* / 0.66*
  - 1 Year: 1.43* / 0.66*
  - 2 Years: 1.48* / 0.65*
  - 3 Years: 1.46* / 0.62
  - 4 Years: 1.41 / 0.60
  - 5 Years: 1.35 / 0.58
- Alternative measure of potential GDP — Energy:
  - Impact: 1.73* / 0.65*
  - 1 Year: 1.68* / 0.65*
  - 2 Years: 1.61* / 0.64*
  - 3 Years: 1.53* / 0.61
  - 4 Years: 1.45 / 0.58
  - 5 Years: 1.39 / 0.55
- Nuclear energy robustness:
  - Lag 2 years: Impact 4.38*; 1 Year 4.23*; 2 Years 4.10; 3 Years 4.01; 4 Years 3.95; 5 Years 3.92
  - Alternative potential GDP: Impact 3.26*; 1 Year 3.17*; 2 Years 3.12; 3 Years 3.09; 4 Years 3.07; 5 Years 3.06
- Land use robustness:
  - Lag 2 years:
    - Impact: -3.38 / 0.19*
    - 1 Year: -1.77 / 0.22
    - 2 Years: 0.93 / 0.25
    - 3 Years: 3.59* / 0.27
    - 4 Years: 5.60* / 0.28
    - 5 Years: 6.85* / 0.30
  - Alternative potential GDP:
    - Impact: -5.18 / 0.42*
    - 1 Year: -1.97 / 0.55*
    - 2 Years: 0.81 / 0.62*
    - 3 Years: 3.02* / 0.66*
    - 4 Years: 4.70* / 0.67
    - 5 Years: 5.98* / 0.67
- Merged specification and alternative ordering (selected rows):
  - Impact: [A] 1.12* / [B] 0.68* ; Prob [A]>[B] = 0.79
  - 1 Year: 1.23* / 0.65* ; Prob [A]>[B] = 0.84
  - 2 Years: 1.31* / 0.62* ; Prob [A]>[B] = 0.87
  - 3 Years: 1.36* / 0.59 ; Prob [A]>[B] = 0.87
  - 4 Years: 1.40* / 0.57 ; Prob [A]>[B] = 0.86
  - 5 Years: 1.44 / 0.55 ; Prob [A]>[B] = 0.86
- Robustness summary:
  - All baseline conclusions survive alternative lag structures, alternative potential GDP measures, and reordering; green energy multipliers generally remain higher or slightly higher; non-eco-friendly multipliers are virtually unaffected or lose significance in some land-use specifications.

### Sectoral comparisons and supporting evidence (selected exact findings)
- Job creation:
  - Clean energy spending may beat job creation from fossil fuels by a ratio of 3:1 (Pollin et al., 2009; Garrett-Peltier, 2017; WRI, 2020b).
  - IRENA (2016): doubling the share of renewables in the global energy mix by 2030 would increase global GDP by up to 1.1 percent or US$ 1.3 trillion compared to business as usual.
  - McKinsey (2020b): every €1 spent in clean energy could generate some €2 to €3 of GVA; 1.1 million to 3.0 million new “job years” of employment from a stimulus package.
- Nuclear normalized estimates (NEI, 2014):
  - Every dollar spent by the average reactor results in the creation of US$1.04 in the local community, US$1.18 in the state economy and US$1.87 in the U.S. economy.
- Conservation benefits:
  - Waldron et al. (2020): protecting 30 percent of the world’s land and ocean provides greater benefits than the status quo; revenues associated with protected areas outweighed the costs by a factor of at least 5:1.

### Conclusions and policy recommendations
- Primary conclusions:
  - Investing in clean energy (solar, wind, nuclear) produces more GDP than the initial expenditure; non-eco-friendly energy spending tends to crowd out other domestic spending.
  - Conservation spending is associated with large economic gains; spending to support unsustainable land uses (industrial crop and animal agriculture) returns less than the initial expenditure.
  - Estimates are robust to different econometric specifications and may underestimate the economic gains from green investments because they do not account for GDP impacts of climate change, biodiversity loss, or public health implications of non-eco-friendly spending.
- Policy implication and recommendation:
  - Gearing post-COVID economic stimulus toward investments that favor decarbonization and carbon-capture through nature-based solutions is presented as the cheapest and shortest route back to a prosperous global economy.
  - Redirect subsidies and support from unsustainable land-use practices toward sustainable, labor-intensive, high-domestic-content conservation and green land-use programs.

*Content extracted from wpiea2021087-print-pdf - References and Section VI/III excerpts.*

### References .............................................................................................................

### wpiea2021087-print-pdf - References

### I. Introduction — context and motivation
- In the first six months of 2020, 6.4 percent less carbon dioxide was emitted than in the same period in 2019 (Liu et al., 2021).
- The first half of 2020 saw an unprecedented decline in CO2 emissions—larger than during the financial crisis of 2008, the oil crisis of the 1979, or even World War II.
- By the end of 2020, the coronavirus pandemic had infected over 90 million people globally and killed almost 2 million, producing factory closures, massive job losses, and paralysis of large swathes of economic activity.
- To meet the 1.5°C target, global emissions would need to drop by the same exaggerated rate seen during the pandemic (7.6 percent per year) in every year for the next few decades (IPCC, 2019).
- The global policy response to the pandemic included trillion-dollar stimulus packages, some of which risk supporting non-eco-friendly industries that need drastic reform to meet climate goals (Hepburn et al. 2020; Carney, 2021).
- Measures taken to sustain activity have generally not been directly targeted at mitigating emissions via fostering conservation, which is necessary for meeting 1.5°C and for reducing zoonotic risks (OECD, 2020).
- Advocacy for “build back better” stresses stewarding the global economy within limits set by nature (Rockström et al., 2017; Attenborough, 2020; Georgieva, 2020; Stiglitz, 2020; Gates, 2021; Carney, 2021).

### Contribution of this paper
- The paper is the first study to directly estimate the effect on GDP of money spent to foster the transition to a zero-carbon, nature-friendly world across a variety of green expenditure typologies.
- “Green” expenditure is expanded beyond emissions-reduction spending to include nature-based negative emissions technologies (“nature-based solutions” or NBSs), including biodiversity conservation and rewilding.
- NBSs are treated as supporting natural carbon sequestration and as vital complements for climate and global temperature stabilization strategies (IPCC 2019; IPBES, 2019; Foley et al., 2020; Dasgupta et al., 2021).

### Key empirical findings
- Using a new international dataset (partly assembled for this analysis), the paper finds that every dollar (private and public) spent on key carbon-neutral or carbon-sink activities—from zero-emission power plants to protection of wildlife and ecosystems—can generate more than a dollar’s worth of economic activity: the total increase in GDP is greater than the original increase in green spending.
- Green spending multipliers appear significantly bigger and more long-lasting than non-eco-friendly spending in alternative energy technologies or land/sea uses.
- For the renewable versus fossil fuel energy investment comparison (where samples are homogeneous and allow formal statistical comparison), estimated point multipliers are:
  - Renewable energy investment: 1.1-1.5 (depending on horizon and specification).
  - Fossil fuel energy investment: 0.5-0.6 (depending on horizon and specification).
- The difference between associated multipliers for renewables and fossil fuels is non-zero with very high probability.
- These findings survive several robustness checks and support bottom-up analyses that stabilizing climate and reversing biodiversity loss are compatible with continuing economic advances.
- Policy implication: in post-COVID-19 recovery design, investments in energy and land/sea use transitions are likely economically superior to supporting unsustainable energy and food production activities—enabling faster recovery while preserving ecological improvements achieved in 2020.

### Methodology and empirical design
- The empirical analysis quantifies investment multipliers using estimates from factor-augmented panel vector-autoregressive (FAVAR) models.
- Motivation for the methodology:
  - Panel dimension exploits cross-country variation where green and non-eco-friendly spending estimates are available.
  - Augmenting with factors extracted from many macroeconomic variables mitigates limited-information concerns by incorporating information likely used by economic agents but not explicitly in the model.
  - Including forecasts of investments formed over the past year as an exogenous variable purges green and non-eco-friendly spending shocks from their anticipated component and mitigates shock-foresight issues highlighted in macro-fiscal literature.
- The methodological choices aim to address the problem of “non-fundamentalness,” a source of bias from misalignment between information sets of economic agents and the econometrician.
- The empirical approach follows similar lines to Fragetta and Gasteiger (2014); Caggiano et al. (2015; 2017); Amendola et al. (2020) and draws on literature about limited information and shocks (Bernanke et al., 2005; Fragetta and Gasteiger, 2014; Stock and Watson, 2005; Forni and Gambetti, 2010).

*Italic source attribution: Content extracted from wpiea2021087-print-pdf - References.*

### Section VI draws policy implications and concludes.

### wpiea2021087-print-pdf - Section VI draws policy implications and concludes

### Why more is needed on clean energy and conservation
- World governments’ collective US$14 trillion fiscal response to the economic damage of the COVID-19 pandemic has concentrated on measures to address the health emergency and support households and firms stranded by the lockdowns (IMF, 2021).
- As the immediate health crisis recedes, attention and funding will turn toward economic recovery, creating opportunities to “build back better” via green stimulus; however, few governments have yet heeded this advice (VividEconomics, 2020).
- The paper quantifies the GDP impact of green stimulus measures using standard investment-multiplier estimation methods, focusing on:
  - (i) reducing emissions by increasing use of clean energy; and
  - (ii) supporting nature’s carbon sinks by enhancing biodiversity conservation.

### Clean energy: context and financing gaps
- Energy consumption contributes to around 3/4 of all anthropogenic greenhouse gas emissions (WRI, 2020a).
- Two categories of clean energy discussed: renewable (solar, wind, hydropower, geothermal, marine; biomass debated) and non-renewable clean (nuclear).
- Nuclear energy: classified as “non-renewable” because uranium is not renewable on a human timescale; yet ranks among the lowest carbon forms of energy generation when lifecycle impacts are considered (IPCC, 2018).
- Technical and system constraints noted:
  - Renewables are intermittent, variable and unpredictable; storage technology is expensive and developing (Lazard Asset Management, 2020; Goldstein and Qvist, 2019).
  - Hydrogen energy from renewables is not an immediate scalable option; hydrogen production today involves fossil fuels with emissions equivalent to the CO2 emissions of the United Kingdom and Indonesia combined (IEA, 2019).
- Shares and investment in 2019:
  - Renewables (excluding large hydro) accounted for about one seventh of global generation (IRENA, 2019).
  - Nuclear energy accounted for about one-tenth (IEA, 2019a).
  - Nuclear capacity globally is estimated to have shrunk by a net 5GW in 2019.
  - Investment in nuclear energy in 2019: US$15 billion.
  - Investment in renewables in 2019: around US$282 billion (IEA, 2020a).
- Target gap to 2030 (BloombergNEF New Energy Outlook 2019 base-case):
  - Gross addition needed of some 2,836GW of new non-hydro renewable energy capacity by 2030.
  - Estimated cost: US$3.1 trillion over the decade.

### Ecosystem conservation: scale of the shortfall
- The agri-food sector emits between 21-37 percent of greenhouse gases (IPCC, 2019); without policy action this share could raise to 50 percent of all global emissions by 2050 (IPCC, 2019; Willett et al., 2019).
- Industrialized agri-food practices are primary drivers of biodiversity loss and deforestation (IPCC, 2019; IPBES, 2019; Willett et al., 2019).
- World Bank / agricultural support estimates:
  - Countries producing two-thirds of the world's agricultural output spent US$600 billion per year in agricultural financial support on average from 2014 to 2016 (Searchinger et al., 2020; UNEP-UNDP-FAO, 2021).
  - Out of US$300 billion in direct spending, only 9 percent explicitly supports conservation and 12 percent supports research and technical assistance.
- Global biodiversity conservation finance gap:
  - Estimated current spending (2019) on a wide range of biodiversity-associated goods: between US$124 and US$143 billion (Paulson Institute-The Nature Conservancy-Cornell University, 2020).
  - Estimated need: US$722-967 billion.
  - Paulson estimates represent 0.1-0.2 percent of 2019 global GDP but may be high-side due to heterogeneous reporting.
- Example cost to expand protected areas to 30 percent of Earth’s surface (Waldron et al., 2020):
  - Estimated cost: between US$103 billion and US$178 billion per year.
  - Current level of investment in protected areas: US$24.3 billion.
  - Required cost represents 4-7 times current investment.

### Data on green and non-eco-friendly spending (coverage and sources)
- Green energy spending data:
  - IEA estimates of capital spending on power generation using renewable sources for 11 countries/groups for 2000-2020 (reported coverage in Table 1 for 2003-2019).
  - Includes investment in electricity networks (transmission, distribution, digital equipment) and repowering; excludes financing and operational costs.
- Nuclear energy (clean non-renewable) capital expenditure:
  - Dataset extends Lovering et al. (2016) to 2017 and includes China; assembled by OECD Nuclear Energy Agency with World Nuclear Association and IAEA.
  - Metric: real Overnight Construction Cost (OCC), converted to constant 2010 US dollars, adjusted using GDP deflators; OCC excludes interest during construction.
- Non-eco-friendly energy spending:
  - IEA capital spending on fossil fuels for the same set of 11 countries/groups for 2000-2020; covers upstream oil and gas, refining, midstream, and power generation; excludes financing and operational costs.
- Green land use spending:
  - Focus on “strict” biodiversity conservation spending (Miller et al., 2012; Waldron et al., 2013, 2017) to avoid heterogeneous reporting.
  - Final sample for biodiversity-spending multiplier analysis: 16 countries (Burkina Faso; Burundi; Cambodia; Cameroon; Central African Republic; Chad; Ghana; Guatemala; Malawi; Mozambique; Niger; Senegal; Sierra Leone; Madagascar; Tanzania; Uganda).
  - Time period compiled: 1994-2008.
- Non-eco-friendly land use spending:
  - OECD producer support estimates (PSE) elaborated by Searchinger et al. (2020); focus on agricultural subsidies that increase quantity/productivity via chemical inputs, mechanization, fossil fuels.
  - Data in current prices converted to US$ using OECD average yearly exchange rates; coverage: 20 countries responsible for 2/3 of global agricultural production; period 1995-2016.
- Summary coverage (from Table 1):
  - Renewable energy: 2003-2019; 9 + 2 groups; countries listed; sources IEA, IMF’s WEO, Thomson Reuters Datastream.
  - Nuclear energy: 1991-2017; 6 countries (China, France, Japan, Korea, Canada, Usa); sources OECD-NEA, IMF’s WEO, Thomson Reuters Datastream.
  - Fossil fuel energy: 2003-2019; 9 + 2 groups; same country/groups as renewables; sources IEA, IMF’s WEO, Thomson Reuters Datastream.
  - Green land use: 1994-2008; 16 countries (listed above); sources Waldron et al. 2018, IMF’s WEO, Thomson Reuters Datastream.
  - Non-eco-friendly land use: 1997-2016; 20 countries (listed in source); sources Searchinger et al. 2020, IMF’s WEO, Thomson Reuters Datastream.

### Methodology (empirical model and identification)
- Econometric approach:
  - Panel vector-autoregressive (VAR) models estimated with Bayesian methods using Normal-Wishart conjugate priors and Minnesota-type priors.
  - Reduced form: yi,t = ρi + γt + A1 yi,t−1 + … + Ap yi,t−p + Bi xi,t + εi,t.
  - Prior for β: multivariate normal with β0 values around 1 for own first lag coefficients and 0 for further lags/cross-variable/exogenous coefficients.
  - Prior for Σc: inverse Wishart: Σc ~ IW(S0, α0).
- Identification:
  - Cholesky identification scheme assuming spending variables react with a lag to GDP, and GDP reacts contemporaneously to spending shocks (spending treated as more exogenous than GDP).
- Simulation and inference:
  - Monte Carlo simulation: perform 20,000 parameter draws and discard first 10,000 draws as burn-in; utilize 10,000 draws from the posterior to derive impulse responses.
  - Impulse responses computed for a time horizon of 5 years; save median response and 16th and 84th percentiles as credible bands.
- Variable normalization and controls:
  - Endogenous variables divided by real potential GDP computed using the HP filter (baseline); robustness checks with alternative filters reported elsewhere.
  - Vector of endogenous variables: y i,t = [Si,t, Ii,t, GDPi,t, Fi,t], except green and non-eco-friendly land use specifications which exclude Ii,t.
  - Four common factors (principal components) extracted uniformly across countries to proxy unobserved macroeconomic factors; Bai and Ng ICp2 selects 2 to 4 factors; the paper uses 4 factors uniformly.
  - For energy specifications, add IMF WEO one-year-ahead forecast of total investments as an exogenous variable to mitigate shock foresight.

### Results — energy sector multipliers (green renewables, nuclear, and non-eco-friendly fossil)
- Renewable (green) versus non-eco-friendly (fossil) energy investment multipliers (Table 2; values are cumulated multipliers, exact figures preserved):
  - Impact:
    - Green (Renewable) Energy Investments Multiplier: 1.19*
    - Non-Eco-Friendly Energy Investments Multiplier: 0.65*
  - 1 Year:
    - Green: 1.20*
    - Non-Eco-Friendly: 0.64*
  - 2 Years:
    - Green: 1.19*
    - Non-Eco-Friendly: 0.62*
  - 3 Years:
    - Green: 1.17*
    - Non-Eco-Friendly: 0.59*
  - 4 Years:
    - Green: 1.14*
    - Non-Eco-Friendly: 0.55
  - 5 Years:
    - Green: 1.11
    - Non-Eco-Friendly: 0.52
  - Note: * denotes multipliers with credible intervals, delimited by the 16th and the 84th percentiles, that exclude zero.
- Key findings on renewables vs fossil fuels:
  - A shock to spending on green renewable energy is more persistent than an equal-sized shock to fossil-fuel spending; output response to renewable spending remains above zero beyond 5 years while the fossil-fuel response dies off by 5 years.
  - Renewable spending multipliers are systematically higher than non-eco-friendly multipliers at short and longer horizons.
  - Interpretation example: a cumulated multiplier of 1.5 at three years indicates cumulative GDP increase is 1.5 times cumulative spending increase.
  - For non-eco-friendly spending, a dollar spent crowds out about 48 cents in the medium run (5-year multiplier 0.52); for renewables a dollar spent crowds in about 11 cents (5-year multiplier 1.11 implies crowding in of 11 cents relative to other components).
  - Statistical comparison (merged specification containing both investments; Table 3):
    - Merged specification multipliers (exact figures preserved):
      - Impact: [A] Green 1.40*; [B] Non-Eco-Friendly 0.62*; Prob [A]>[B]: 0.92
      - 1 Year: Green 1.46*; Non 0.58*; Prob 0.94
      - 2 Years: Green 1.49*; Non 0.54*; Prob 0.94
      - 3 Years: Green 1.51*; Non 0.51; Prob 0.93
      - 4 Years: Green 1.53*; Non 0.48; Prob 0.92
      - 5 Years: Green 1.54; Non 0.47; Prob 0.91
    - At all horizons, more than 90 percent of the empirical distribution of the difference lies above zero, indicating the difference is positive with high probability.
- Mechanisms explaining higher green multipliers:
  - Clean energy spending is more labor intensive than fossil-fuel spending and has higher domestic content, generating larger direct and indirect employment and local ripple effects.
  - Clean-energy investments produce more jobs at all pay levels and more entry-level jobs per dollar than fossil fuel investments.
- Nuclear (non-renewable clean) energy multipliers (Table 4; not strictly comparable to other sets due to sample differences):
  - Impact: 4.11*
  - 1 Year: 3.97*
  - 2 Years: 3.88
  - 3 Years: 3.83
  - 4 Years: 3.80
  - 5 Years: 3.78
  - Note: * denotes multipliers with credible intervals, delimited by the 16th and the 84th percentiles, that exclude zero.
- Nuclear findings:
  - Spending on nuclear energy is more persistent than fossil-fuel investment and tends to crowd in other investments.
  - Nuclear spending shows a large near-term output effect—about six times larger than fossil fuel energy in the impact period—but loses statistical significance after two years.
  - Possible explanations: nuclear projects are more frontloaded and intensive in skilled and unskilled labor during construction; nuclear construction times (average 5.1 years for large recent reactors) and higher pay for nuclear workers relative to wind/solar may drive large near-term multipliers.

### Results — land use multipliers (overview)
- The document begins reporting comparisons of ecosystem conservation (green land use) versus subsidies to conventional agriculture (non-eco-friendly land use) impulse responses and multipliers; interpretation requires caution because:
  - The IRFs and associated multipliers were estimated over different country and time samples and in separate specifications due to data availability constraints.
- (Detailed land-use multiplier numeric results and further interpretation are presented beyond the provided excerpt.)

*Source: wpiea2021087-print-pdf - Section VI draws policy implications and concludes.*

### Section 3. This is also the reason why a statistical test on their difference cannot be

### wpiea2021087-print-pdf - Section 3. This is also the reason why a statistical test on their difference cannot be constructed.

### Green versus non-eco-friendly land use spending: dynamics and interpretation
- Impulse responses:
  - A shock to conservation spending has a long-lasting effect, similarly to the shock to green spending in the energy sector, implying durable economic benefits plus mitigation and carbon-sink gains.
  - A spending shock to support industrial farming activities is considerably shorter-lived and completely dissipates after 5 years.
- Economic interpretation:
  - Conservation spending mixes public consumption (wages, education, training, recreational programming) and some public investment, whereas spending on conventional agriculture primarily reflects public transfers and subsidies to crop and animal producers in industrial farm systems.
  - Coarse comparisons remain informative; consensus is emerging that subsidies to unsustainable land use and conventional agriculture should be quickly redirected toward sustainable uses.

### Cumulated multipliers for land use spending (Table 5)
- Horizon / Green Land Use Multiplier / Non-Eco-Friendly Land Use Multiplier
  - Impact: -5.36 / 0.55*
  - 1 Year: -1.60 / 0.85*
  - 2 Years: 1.45* / 0.95*
  - 3 Years: 3.75* / 0.96*
  - 4 Years: 5.45* / 0.95
  - 5 Years: 6.67* / 0.94
- Note: * denotes multipliers with credible intervals, delimited by the 16th and the 84th percentiles, that exclude zero.
- Key quantitative finding: "for every dollar spent in conservation, almost seven more are generated in the larger economy in the medium term."

### Why green land use multipliers are high
- Three main determinants:
  - Dataset composition: estimates are conducted on data from developing countries, which typically capture spending programs financed by donors that do not crowd out or absorb domestic resources, leading to high multipliers.
  - Labor intensity: conservation activity has strong labor intensity and drives visitor-economy effects (hospitality and tourism), especially in rural and coastal communities with below-average income and higher propensities to spend.
  - Price effects and ecosystem services: limiting land available for agricultural expansion lifts prices paid to rural producers; protecting biodiversity underpins ecosystem services (food production, fresh water, natural resources, protection from extreme weather events), creating jobs and fostering innovation (biomimicry).
- Contrast with industrial agriculture:
  - Multipliers of spending to support industrial agricultural production are below one at every horizon.
  - Reflects high mechanization, low value added, high costs of machinery, fossil fuel energy, imported chemical inputs and foreign-patented GMO seeds with low domestic content and high global market concentration of suppliers.
  - Implication: repurposing spending from unsustainable land uses toward more labor-intensive and high-domestic-content sustainable land uses may yield important economic gains and aid a green recovery.

### Comparison to sectoral impact studies (selected findings)
- Clean (renewable) energy:
  - Job creation: spending in clean energy may beat job creations from fossil fuels by a ratio of 3:1 (Pollin et al., 2009; Garrett-Peltier, 2017; WRI, 2020b).
  - IRENA (2016): doubling the share of renewables in the global energy mix by 2030 would increase global GDP by up to 1.1 percent or US$ 1.3 trillion compared to business as usual.
  - McKinsey (2020b): every €1 spent in clean energy could generate some €2 to €3 of GVA; 1.1 million to 3.0 million new “job years” of employment from a stimulus package.
- Clean (non-renewable) energy (nuclear):
  - NEI (2014) normalized estimates: every dollar spent by the average reactor results in the creation of US$1.04 in the local community, US$1.18 in the state economy and US$1.87 in the U.S. economy.
- Green land use:
  - Waldron et al. (2020): protecting 30 percent of the world’s land and ocean provides greater benefits than the status quo; revenues associated with protected areas outweighed the costs by a factor of at least 5:1, a multiplier close to the aggregate top-down estimates reported here.

### Robustness analysis: specification checks and alternative measures
- Main robustness checks performed:
  - Alternative lag structure: uniform lag of two years (versus baseline one-year lag).
  - Alternative measure of potential GDP: Mohr (2005) filter versus conventional HP filter.
  - Reordering of endogenous variables in VAR when comparing green and non-eco-friendly investments.
- Robustness results summary:
  - All conclusions from baseline estimates survive the changes.
  - Green (renewable) energy investment multipliers are slightly higher under alternative specifications; non-eco-friendly energy investment multipliers are virtually unaffected (Table 6).
  - Nuclear energy multipliers are somewhat higher with two-year lag and slightly lower with alternative potential GDP measure; overall dynamics and significance similar (Table 7).
  - Green land use multipliers under alternative specifications are comparable to baseline; non-eco-friendly land use multipliers become smaller and quickly lose statistical significance with a two-year lag (Table 8).
- Selected robustness tables (exact figures preserved):
  - Table 6 — Lag structure of 2 years (Green Energy Investments Multiplier / Non-Eco-Friendly Energy Investments Multiplier)
    - Impact: 1.28* / 0.66*
    - 1 Year: 1.43* / 0.66*
    - 2 Years: 1.48* / 0.65*
    - 3 Years: 1.46* / 0.62
    - 4 Years: 1.41 / 0.60
    - 5 Years: 1.35 / 0.58
  - Table 6 — Alternative measure of potential GDP
    - Impact: 1.73* / 0.65*
    - 1 Year: 1.68* / 0.65*
    - 2 Years: 1.61* / 0.64*
    - 3 Years: 1.53* / 0.61
    - 4 Years: 1.45 / 0.58
    - 5 Years: 1.39 / 0.55
  - Table 7 — Nuclear Energy Investments (Lag structure of 2 years)
    - Impact: 4.38*
    - 1 Year: 4.23*
    - 2 Years: 4.10
    - 3 Years: 4.01
    - 4 Years: 3.95
    - 5 Years: 3.92
  - Table 7 — Nuclear Energy Investments (Alternative measure of potential GDP)
    - Impact: 3.26*
    - 1 Year: 3.17*
    - 2 Years: 3.12
    - 3 Years: 3.09
    - 4 Years: 3.07
    - 5 Years: 3.06
  - Table 8 — Land use (Lag structure of 2 years)
    - Impact: -3.38 / 0.19*
    - 1 Year: -1.77 / 0.22
    - 2 Years: 0.93 / 0.25
    - 3 Years: 3.59* / 0.27
    - 4 Years: 5.60* / 0.28
    - 5 Years: 6.85* / 0.30
  - Table 8 — Land use (Alternative measure of potential GDP)
    - Impact: -5.18 / 0.42*
    - 1 Year: -1.97 / 0.55*
    - 2 Years: 0.81 / 0.62*
    - 3 Years: 3.02* / 0.66*
    - 4 Years: 4.70* / 0.67
    - 5 Years: 5.98* / 0.67
  - Table 9 — Merged specification and alternative ordering (selected rows)
    - Impact: [A] 1.12* / [B] 0.68* ; Prob [A]>[B] = 0.79
    - 1 Year: 1.23* / 0.65* ; Prob [A]>[B] = 0.84
    - 2 Years: 1.31* / 0.62* ; Prob [A]>[B] = 0.87
    - 3 Years: 1.36* / 0.59 ; Prob [A]>[B] = 0.87
    - 4 Years: 1.40* / 0.57 ; Prob [A]>[B] = 0.86
    - 5 Years: 1.44 / 0.55 ; Prob [A]>[B] = 0.86

### Conclusions and policy implications (Section VII)
- Primary conclusions:
  - Investing in clean energy (solar, wind, nuclear) produces more GDP than the initial expenditure; non-eco-friendly energy spending tends to crowd out other domestic spending.
  - Conservation spending is associated with large economic gains; spending to support unsustainable land uses (industrial crop and animal agriculture) returns less than the initial expenditure.
  - All estimates are robust to different econometric specifications and may underestimate the economic gains from investing in green energy and land use because they do not account for the GDP impact of climate change and biodiversity loss or public health implications of non-ecofriendly spending.
- Policy implication:
  - Gearing post-COVID economic stimulus toward investments that favor decarbonization and carbon-capture through nature-based solutions is presented as the cheapest and shortest route back to a prosperous global economy.

*Source: wpiea2021087-print-pdf - Section 3. This is also the reason why a statistical test on their difference cannot be constructed.*

### REFERENCES

### wpiea2021087-print-pdf - REFERENCES

### Appendix A — Data: Endogenous variables and series construction
- Variables of interest: gross domestic product; total investments; renewable energy investments; non-eco-friendly energy investments; nuclear energy investments; green land use spending; non-eco-friendly land use spending.
- Data sources:
  - Clean renewable energy and non-eco-friendly energy investments: International Energy Agency.
  - Nuclear energy investments: assembled specifically for this project by the OECD’s Nuclear Energy Agency in collaboration with the World Nuclear Association and the International Atomic Energy Agency.
  - Green land use spending: updated starting from Waldron et al (2013, 2017).
  - Non-eco-friendly land use spending: elaboration of OECD producer support estimates (PSE) assembled by Searchinger et al. in 2020 for the World Bank Group.
  - Gross domestic product and total investments: IMF’s World Economic Outlook database.
- Data transformation:
  - All series are transformed in real terms using the implicit GDP price deflator.
  - Series are then normalized by dividing by real potential GDP.
- Country samples and time spans by specification:
  - Clean renewable energy and non-eco-friendly energy investments: China, Japan, Korea, Canada, United States, Brazil, Indonesia, Mexico, Russia, Oceania group (Australia and New Zealand) and EA group (France, Germany and Italy); time span 2003 to 2019.
  - Nuclear energy investments: China, France, Japan, Korea, Canada and United States; time span 1991 to 2017.
  - Green land use spending: Burkina Faso, Burundi, Cambodia, Cameroon, Central African Republic, Chad, Ghana, Guatemala, Malawi, Mozambique, Niger, Senegal, Sierra Leone, Madagascar, Tanzania and Uganda; time span 1994 to 2008.
  - Non-eco-friendly land use spending: China, Japan, Korea, Canada, United States, Australia, Chile, Indonesia, Mexico, New Zealand, Russia, South Africa, Colombia, Iceland, Israel, Kazakhstan, Norway, Switzerland, Turkey and Ukraine; time span 1997 to 2016.

### Appendix A.2 — Exogenous variables
- For specifications that include clean renewable energy investments, non-eco-friendly energy investments and nuclear energy investments:
  - Exogenous variable used: the forecast of the total investments made at time t-1 for time t, provided by IMF’s World Economic Outlook.

### Appendix A.3 — Informational dataset and factor extraction
- Informational dataset composition:
  - Consists of 12 series for each country downloaded from IMF’s World Economic Outlook and Thomson Reuters Datastream Economics databases.
  - Choice of series driven by availability for all countries and all periods included in the analysis.
- Variables downloaded for each country:
  - National Account: Government Consumption Expenditure; Total Government Revenue; Export of Goods and Services; Imports of Goods and Services; Final Consumption Expenditure of Households; Gross National Saving.
  - Output: Industrial Production Index (not available for Green Land Use Dataset); Change in Inventories.
  - Employment: Employees Domestic Concept.
  - Exchange rates: Real Effective Exchange Rates (not available for Green Land Use Dataset).
  - Money and credit quantity aggregates: Broad Money or Money Supply M0, M1, M2, M3 (depending on the availability).
  - Price indexes: Consumer Price Index.
- Stationarity treatment:
  - Variables are transformed where appropriate to guarantee stationarity.
  - Stationarity tested by the Phillips and Perron (1986) and Kwiatkowski et al. (1992) tests.

*Source: wpiea2021087-print-pdf - REFERENCES*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021087-print-pdf.pdf_