Policy Research Working Paper 10551 A Climate-Fiscal Policy Mix to Achieve Türkiye’s Net-Zero Ambition under Feasibility Constraints Christian Schoder Remzi Baris Tercioglu Equitable Growth, Finance and Institutions Practice Group & Macroeconomics, Trade and Investment Global Practice August 2023 Policy Research Working Paper 10551 Abstract This paper employs an estimated dynamic stochastic of fossil fuel subsidies and public investment. Although open-economy macro framework to identify policy inter- the proposed policy package has only moderate effects on ventions that allow Türkiye to achieve net-zero emissions gross domestic product, transition risks involve declining by 2053 while respecting important feasibility constraints exports and fossil asset stranding. The paper highlights the such as fiscal consolidation and sovereign debt stability importance of transparent policy communication and a as well as compensation of low-income households. The credible commitment to the net-zero agenda to ensure an policy mix includes a carbon tax, a renewable energy sub- orderly transition. Improving the rule of law and access to sidy, transfer payments, public infrastructure investments, green finance considerably support the private sector-led a bad bank for stranded fossil fuel assets, and the phase-out low-carbon transition. This paper is a product of the Office of the Chief Economist, Equitable Growth, Finance and Institutions Practice Group and the Macroeconomics, Trade and Investment Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at cschoder@worldbank.org and rtercioglu@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team urkiye’s A Climate-Fiscal Policy Mix to Achieve T¨ Net-Zero Ambition under Feasibility Constraints∗ Christian Schoder† Remzi Baris Tercioglu‡ World Bank World Bank Keywords: Climate-fiscal policy, net-zero pathway, low-carbon transition, Bayesian urkiye estimation, T¨ JEL Classification: Q43, Q48 ∗ We would like to thank J´anos Varga, Charl Jooste, and Florent McIsaac for valuable comments and suggestions. † E-Mail: cschoder@worldbank.org ‡ E-Mail: rtercioglu@worldbank.org 1 Introduction T¨urkiye faces several climate-related risks. Recent extreme weather events, including floods and wildfires, have exposed the country’s vulnerability to climate change. T¨ urkiye is highly dependent on fossil fuel imports which amount to 87% of its total primary energy supply.1 This poses significant risks for energy security as the oil and gas market turmoils after the Russian invasion of Ukraine have forcefully illustrated. Moreover, the European Union’s ambitious low-carbon agenda will have considerable repercussions for the primary export market for Turkish manufacturing, which is more carbon-intensive than its European coun- terpart. In light of these climate-related risks, T¨urkiye has increased its climate ambition in recent years. T¨ urkiye’s renewable energy capacity grew by more than 50% since 2016 (IEA 2021). As a top-20 emitter, T¨ urkiye submitted its Intended Nationally Determined Contribution in 2015, setting the target to reduce greenhouse gas emissions by 21% relative to the business-as-usual (BAU) benchmark by 2030. In 2021, the country ratified the Paris Agreement and announced the net-zero target of 2053. T¨urkiye’s development progress since the early 2000s has been remarkable. Per-capita GDP has more than tripled within 20 years, making T¨ urkiye the 19th largest economy in the world. Since 2015 the growth dynamics have cooled down considerably. Unemployment has exceeded the 10% mark ever since 2015 despite a low labor force participation rate, especially among women, and a significant informal sector (World Bank 2023). While formally inde- pendent, the central bank’s monetary policy is subject to severe government interference, which let inflation explode to a two-decade-high value of 85.5% in October 2022 (TurkStat 2022). Addressing structural deficits is critical for the low-carbon transition as they constrain the effectiveness of climate policy in mobilizing private finance. Structural deficits of the Turkish economy include low credibility of policy commitments, high costs of contract enforcement combined with low business confidence in public institutions, and uneven access to bank fi- nance (OECD 2021, G¨ urkaynak et al. 2022). Dechezleprˆ etre et al. (2022) provide compelling evidence that transparency and commitment raise acceptance of climate policy. Regarding the business climate, a weak rule of law drives up equity risk premia making capital-intensive, productivity-enhancing technology less appealing than labor-intensive, low-value-added ac- tivities. Finance constraints further impede private investment in low-carbon technology and limit the power of market signals. The present paper aims to project the paths of climate-fiscal policy interventions that allow T¨urkiye to grow along the 2053 net-zero pathway while honoring significant feasibil- ity constraints. The policy instruments include a carbon tax, a renewable energy subsidy, transfer payments, public infrastructure investment, a bad bank for stranded assets, and the phase-out of fossil fuel subsidies and public investment. To study the role of supporting structural policies in the low-carbon transition, we explore how the macroeconomic effects 1 See release 055 of the GLORIA Global Environmentally-Extended Multi-Region Input-Output (MRIO) database constructed in the Global MRIO Lab (Lenzen et al. 2017, 2022). 2 of the policy interventions differ in counterfactual scenarios of a fully credible commitment to net zero, an improved rule of law, and facilitated access to green bank finance. We address these research questions using an empirical dynamic stochastic open-economy macro framework similar to the European Commission’s E-QUEST model presented by Varga et al. (2022) but shifts the focus on an emerging-market context.2 The model is designed for data-driven policy analysis and the simulation of low-carbon transitions. Re- newable and fossil public investment can go into energy production or infrastructure. Infras- tructure improves the marginal product of private capital but only up to a satiation point. To allow the model to capture deep transformations such as the low-carbon transition, it includes two critical features: a learning-by-doing externality in the generation of renewable energy (Mercure 2012, Rubin et al. 2015); and the distinction between short- and long-term elasticities of substitution between renewable and fossil energy (Guerrieri et al. 2008). On the labor market, the model allows for the possibility of dis-equilibrium unemployment at the steady state, which provides flexibility regarding model closures and minimizes the impact of theory on the results for the long run.3 The model features various financial instruments, bank lending constraints, and a financial accelerator mechanism along the lines of Kiyotaki and Moore (1997). The interested reader is referred to Appendix A, which presents the details of the model. We fit the model to Turkish data using standard Bayesian techniques as outlined in Herbst and Schorfheide (2015). The dataset covers the period from 2000Q2 to 2022Q3 and includes 31 time series in mixed frequency. Appendix B reports the details of the model estimation. We run all estimations and simulations in Dynare 5.3 (Adjemian et al. 2022). The simulation results suggest that T¨ urkiye needs to employ a dynamic policy mix to achieve the steady decline of 12 million (mn) tonnes of CO2 emissions per year required for net zero by 2053 while observing the following constraints: (a) keeping sovereign debt at 34% of trend GDP, (b) fully compensating low-income households for consumption lost relative to BAU, (c) public financing of necessary infrastructure investments which amounts to an average of 2 billion (bn) of 2015 USD ($) per year (World Bank 2022), and (d) buying up private stranded assets. Under these constraints and the phase-out of fossil subsidies and public investment, our results suggest that T¨urkiye should kick off the decarbonization of its economy with a carbon or fuel tax reaching $70 per tonne of CO2 by 2027. By then, public debt will be stabilized at the desired level. The carbon tax will then generate excess revenues that can be recycled as renewable energy subsidies to minimize the rise in energy prices. Renewable subsidies of up $1400 per tonne of oil equivalent (toe) trigger a renewable energy boom reinforced by endogenous productivity gains from scaling up renewable generation. As fossil energy 2 Our modeling framwork also connects us to the literature on optimal environmental policy in a general equilibrium setting (Fischer and Springborn 2011, Annicchiarico and Di Dio 2015, Dissou and Karnizova 2016, van der Ploeg and Rezai 2021). 3 Dis-equilibrium unemployment is known, among many others, from Dynamic Stochastic Dis-Equilibrium models (Schoder 2017, 2020) as well as dis-equilibrium theory (Barro and Grossman 1971, Chiarella et al. 2005). 3 consumption is reduced, the carbon tax will find it increasingly challenging to generate fis- cal revenues. By 2037, debt stability will require the government to phase out renewable subsidies completely and adjust another instrument to keep public debt stable. In our sim- ulation, government consumption declines by 1.8%-points of trend GDP between 2037 and 2053. Over the projection period, the government will also increase the required spending on renewable infrastructure and phase out investment in fossil energy production and infras- tructure. Finally, transfer payments to low-income households increase by 1%-point of trend GDP over the projection period. This policy package is projected to mobilize a considerable amount of green finance. Private investment in renewable energy increases from 0.2% of trend GDP in 2022 to almost 4% by 2053. Renewable energy consumption will increase from around 20 million toe in 2022 to 420 million by 2053. This is 460% higher than the baseline projection. The projected impacts on GDP are moderate. During the fiscal consolidation phase, the carbon tax will briefly reduce GDP, peaking in 2027 with a GDP loss of 0.7% compared to the no-policy baseline. Between 2031 and 2037, GDP is projected to exceed the baseline slightly. By 2037, the declining government consumption will increasingly reduce GDP, reaching a projected loss of 1% compared to the baseline in 2053. Regarding welfare, low- income households are broadly indifferent to the transition as long as they receive sufficient compensation for the consumption lost relative to the baseline. In contrast, high-income households considerably benefit from the transition as the renewable capital assets created translate into higher household wealth. Despite the overall moderate GDP effects and positive welfare effects, there are transition risks: While the low-carbon transition eliminates the dependence on fossil fuel imports, core goods exports will decline by up to 11% compared to the baseline. Moreover, there is a manageable but non-negligible risk of fossil asset stranding. In our simulation, stranded assets amount to almost 0.1% of trend GDP in 2053. Delaying the transition until 2030 will considerably increase the risk of stranded assets: 0.65% of trend GDP in 2043. Our simulations stress the positive role of complementary policy. Consistent with Nemet et al. (2017), Battiston et al. (2021), Diluiso et al. (2021), Campiglio et al. (2023), we show that transparent policy communication and a credible commitment to the net-zero agenda benefit the low-carbon transition. It allows the private sector to adjust ahead of time, reducing economic friction. A lack of commitment and the expectation of policy reversals make the transition considerably more costly. Moreover, we find that implementing institutional reforms that improve the rule of law and thereby reduce equity risk premia to the average of China, India, and South Africa drastically improve the macro-economic outlook of the low-carbon transition, with GDP peaking at 2.6% above the no-policy baseline. Finally, reforms that facilitate access to green finance boost renewable energy investment and generation. Overall, complementary structural reforms reduce the required climate-policy interventions and improve the macroeconomic consequences of the low-carbon transition. The present paper seeks to contribute to the literature on the macroeconomic implications of climate-fiscal policy in an emerging-market context. World Bank (2022) estimates the ad- 4 ditional investment and renewable energy capacity needed for achieving the net-zero pathway in T¨urkiye. Hallegatte et al. (2023) impose these investment and capacity needs in a Com- putational General Equilibrium (CGE) and a macro-structural model and study the sectoral and macroeconomic repercussions under various scenarios regarding private/public financ- ing, crowding-out of investment, and labor-market frictions. We seek to complement these contributions by identifying the dynamic climate-fiscal policy mix needed for net zero by 2053 while observing critical feasibility constraints. The integrated framework used projects endogenously investment needs, finance, and, to some extent, productivity gains. Exploiting the value-added of a structural model with inter-temporal decision-making, we seek to offer additional insights on the roles of private sector incentives, policy credibility, and structural reform in supporting the low-carbon transition. The model used in this paper is similar to the E-QUEST and GM models of the European Commission (Varga et al. 2022, Albonico et al. 2019) and the New Area Wide Model of the ECB (Christoffel et al. 2008) but puts the focus on market imperfections relevant for middle-income countries. Our framework is also related to global models such as the Global Macroeconomic Model of the Energy Transition (GMMET) and the Integrated Policy Framework developed by the IMF (Benjamin Carton and Voigts 2022, Vitek et al. 2022) as well as G-Cubed developed by McKibbin and Wilcoxen (1998), but it is considerably smaller in scale to allow for Bayesian estimation. Since estimating the model was a priority, data availability and computational capacity imposed restrictions on its size. Hence, several qualifications are to be addressed in future research: First, while the model comprehensively tracks macroeconomic interactions and includes the critical climate-fiscal policy instruments, it abstracts from sector-specific de- carbonization challenges and features only emissions that arise from fossil fuel combustion. As outlined in World Bank (2022), additional policy interventions such as norms and reg- ulations may be required to improve energy efficiency and address other emissions sources such as industrial processes and land use. Second, we assume that a well-designed energy market regulation is in place such that renewable energy is profitable despite the significant up-front capital costs and volatility in energy prices. This includes complete future markets, access to the power grid, and the absence of excessive administrative obstacles in commis- sioning renewable energy generation facilities. Finally, the relative costs and benefits of the low-carbon transition depend on the choice of the baseline scenario. The present paper uses the BAU scenario as the baseline for T¨ urkiye. This ignores the Carbon Border Adjustment Mechanism the European Union intends to impose on imports. It also neglects potential technology spill-overs from other decarbonizing parts of the world. The remainder of the paper proceeds as follows. Section 2 provides the intuition of the empirical macro framework. It motivates modeling and calibration choices for T¨ urkiye. Ap- pendix A reports the model in detail. Section 3 discusses the Bayesian estimation results. Appendix B details the data sources, parameter calibration, and estimation. Section 4 identi- urkiye on track to net zero and studies its macroeconomic fies a policy mix projected to set T¨ implications. It also investigates how policy credibility and structural reforms can support the low-carbon transition. Section 5 concludes the paper. 5 2 urkiye An empirical macro-economic framework for T¨ The underlying model is designed for data-driven climate policy analysis in T¨ urkiye.4 It is calibrated and estimated using quarterly and annual data covering the period from 2000 to 2022. It comprises a high-skilled household sector, a low-skilled household sector, a core good sector, an energy sector, a public sector, and the Rest of the World (RW). Households use energy for consumption (on average around 48% of total final energy consumption) and firms for non-energy production (around 52%), which can be fuel, heat, or electricity. The energy sector consists of renewable energy (with a sample average of 0.9% of GDP and 17% of total energy consumption) and fossil energy (6.2% of GDP and 83% of total energy consumption). A carbon mining sector (3.4% of GDP) extracts and imports carbon (such as coal, crude oil, and gas) which is an input to fossil energy production. CO2 emissions arise from the combustion of fossil fuels and have fluctuated around 0.46 kilogram (kg) per 1 constant 2015 USD of GDP since the early 2000s.5 Households, firms, the government, and the RW demand and supply finance, respectively. The model features various financial instruments. Households are assumed to be net creditors holding their wealth as bonds, deposits, and private equity. Firms are net debtors financing investments by borrowing from households and issuing equity. The government finances its deficit by issuing bonds. The RW is a net creditor from T¨ urkiye’s perspective. The public, corporate, and RW’s net debt-GDP ratios are moderate on average: around 36%, 57%, and -35% of GDP, respectively. Yet, these debt levels have increased sharply in recent years (OECD 2022). Despite a low net debt vis-a-vis the RW, T¨ urkiye imports around 87% of its total primary energy supply. 2.1 High- and low-skilled labor households Capturing to some extent how the low-carbon transition affects households differently, the model features high-skilled labor H-households (50% of the labor force) and liquidity-constrained low-skilled labor L-households (50% of the labor force). We assume that L-households hold no financial wealth. In T¨ urkiye, the 50% top-income households account for two-thirds of all consumption expenditures. Hence, we assume H-households consume twice as much as L-households while providing the same labor hours. The former receive capital income as well as a higher real wage. The overall wage share in total income (household income and retained earnings) is around 50%. L-households also receive transfer payments from the government (around 7.3% of GDP). H-households own the firms and domestic carbon mines (oil, coal, and natural gas), provide high-skilled labor services, and save. Because a specific household’s labor variety is unique, each H-household has market power. Wage adjustment costs allow the model to capture the persistence in high-skilled wage contracts as observed in empirical data (Taylor 4 Appendix A reports the model in detail. 5 Appendix B reports the model calibration and lists the respective data sources and references. 6 1980, 2016). Taking as given the demand function for their labor variety as well as the budget constraint and facing Rotemberg (1982) wage adjustment costs, H-households choose inter-temporal paths for consumption, governments bonds, domestic deposits, international bonds, firm equity shares, and the wage for their labor variety to maximize inter-temporal utility where per-period utility depends on consumption, labor supply, and wealth. Utility from consumption is subject to external habit formation, which helps the model capture the hump-shaped response of consumption to macro shocks as identified in empirical studies (Christiano et al. 2005). In contrast to that, the L-household’s problem is static. This is because of the assumption that L-households are hand-to-mouth consumers. They do not receive enough income to be able to save. Hence, they hold no wealth and cannot smooth consumption over time. This assumption is empirically accurate and helps the model capture the empirical co-movement of government spending and private consumption (Christiano et al. 2010). Including wealth in the utility function allows the model to capture precautionary saving motives even at a first-order linear approximation (Michaillat and Saez 2021). To see the implications of this assumption, consider the Euler equation (neglecting productivity growth for the sake of clarity), RB,t λH,t = ψA + β Et λH,t+1 , ΠY,t+1 where λH,t equals marginal consumption utility, ψA is the wealth utility scaling parameter, RB,t is the gross bond return, and ΠY,t is the gross inflation rate. Because saving yields utility by itself, the utility from wealth drives a wedge between the value of a unit of income in t to the expected value that this unit would yield if saved and consumed in t + 1.6 A critical implication of this modeling choice for T¨urkiye is that the determinacy of the rational- expectation equilibrium does not necessitate the Taylor principle : an inflation elasticity of the policy rate greater than one (Schoder 2020). While government bonds and deposits are assumed to be risk-free, equity shares or in- ternational bonds are not. These assets are associated with financial intermediation costs which depend on the net-debt position of the issuer and have the interpretation of risk pre- mia (Christoffel et al. 2008). The equity risk premium on top of risk-free government bonds has been around 5.2% in T¨ urkiye7 . The international risk premium is assumed to increase with the net-debt position of the RW vis-a-vis T¨ urkiye. With a persistent Turkish current account deficit, the RW has accumulated net assets of around 40% of Turkish GDP (World Bank 2023). Hence, the international risk premium is negative. 6 As Schoder (2020) explains in detail, this breaks the so-called classical dichotomy at the steady state because it links consumption to the real interest rate. In contrast to conventional Dynamic Stochastic General Equilibrium (DSGE) models, consumption does not drop out from the Euler equation at the steady state. The natural rate of interest is no longer uniquely determined by the Euler equation alone but linked to the real economy through high-skilled consumption. 7 We derive the risk premium as the difference between stock market returns and 10-year government bond interest rates. 7 Regarding the compensation of labor, the high-skilled labor market is standard: H- households have market power and set the wage rate such that the marginal disutility from supplying another hour of labor equals the marginal utility of the additional consumption which that hour of work allows for. The high-skilled labor market clears. L-households choose their supply of low-skilled hours such that the marginal disutility of another hour equals the marginal utility of the corresponding consumption. While H-households set their wage, L-households take their wage as given. One core assumption of the model is that the low-skilled nominal wage is administered or partly exogenous, as in Benigno and Fornaro (2018). This is equivalent to the conventional assumption in New Keynesian Dynamic Stochastic General Equilibrium (DSGE) models that the nominal interest rate is a policy variable. In particular, the low-skilled wage rate is determined by a Phillips curve relationship. Low-skilled wage inflation responds to employ- ment (to capture the effects of labor-market tightening on wage formation) and to the price inflation rate (to capture wage indexing). This assumption generates dis-equilibrium unem- ployment (around 10%) in the low-skilled segment of the labor market. With idle labor, the economy can swiftly respond to demand shocks by adjusting low-skilled employment, given constraints and costs imposed by input substitution technologies. Schoder (2020) provides a detailed review of the macroeconomics of disequilibrium models. It is critical to note that an estimated parameter determines where the model settles between two extremes: per- fectly inelastic low-skilled wage (constant wage or pure Keynesian case) and perfectly elastic low-skilled wage (labor market clearing or pure Walrasian case). 2.2 CES aggregators The production sector is decomposed into multiple layers along the value chain, as illustrated in Figure 1. The economy is populated by four different representative Constant Elasticity of Substitution (CES) aggregators: Retailers (Y-firms), wholesale firms (W-firms), energy firms (E-firms), and labor firms (L-firms). Each combines two different input goods into an output good using a CES aggregator without generating value added. Moreover, the economy features three value-added sectors: core goods firms, renewable energy firms, and fossil energy firms. They use capital and labor as production inputs. Finally, a carbon mining sector extracts carbon without inputs or costs and sells it on a world market. The value added in this sector is matched by rents split between the household and RW sectors. Production functions are of the CES type. The advantage of this functional form is that only two parameters (input share and elasticity of substitution) characterize the technology. Moreover, extreme cases such as perfect complements or perfect substitutes are nested. At every node in Figure 1 an elasticity controls the degree by which inputs can be substituted with each other for given price changes. The more rigid the production structure or the higher the cost at which input substitutes can be produced, the greater will be the extent by which cost increases of a particular input are passed through the value chain and translate into overall inflation. 8 tmp tmp Final good Exported core good tmp good Wholesale Energy service tmp Domestic core good Imported core good Core value-added Energy service tmp Core capital service Labor service Labor service Core private capital Core public capital High-skilled labor Low-skilled labor Energy service Renewable energy Fossil energy Rnewable value-added Fossil value-added Carbon Renewable cap- Fossil capital service Labor service Labor service ital service Renewable pri- Renewable pub- Fossil private capital Fossil public in- vate capital and lic infrastructure and public generation frastructure public generation Figure 1: Schematic representation of the supply side of the model. The model features the following aggregators: At the end of the value chain, Y-firms produce the homogeneous final good used for private and public consumption as well as investment by combining wholesale goods and energy. W-firms produce wholesale goods and sell them to the Y-firms. As inputs, they combine domestic and international core goods. Note that, apart from carbon imports, only core 9 goods are traded which is a strong simplification of the complexity of international trade. Yet, it allows the model to capture the domestic-demand and competitiveness transmission channels of domestic and international shocks. Non-carbon imports in T¨ urkiye amount to around 22.8% of GDP (and to 26% including coal, crude oil, and gas). E-firms produce an energy service (with a value of around 7.1% of GDP) using renewable energy (with a share of 12%) and fossil energy (with a share of 88%). The energy service is input to two sectors: intermediate core Z-good production (around 52% of total energy consumption or 3.7% of GDP) and final Y-good production (around 48% of total energy consumption or 3.4% of GDP).8 E-firms pay a fuel tax on their fossil energy input (1.4 % of GDP) and receive fossil energy subsidy (0.4 % of GDP). They also receive a subsidy for each unit of renewable energy input. Finally, we allow E-firms to accrue zero-mean windfall profits which are distributed to the H-households. Given the transformational nature of the low-carbon transition, a simple CES production structure is too restrictive for the energy sector. Similar to Guerrieri et al. (2008), we, therefore, introduce input adjustment costs which add short-run friction to the long-run elasticity of substitution between renewable and fossil energy. L-firms combine high-skilled and low-skilled labor (both measured in hours) to produce a labor service. High-skilled households receive, on average, 61% of the total wage bill, and low-skilled labor the remaining 39%. The labor service is sold to all sectors that use labor inputs for production: core goods (98.64%), renewable energy (0.23%), and fossil energy (1.13%). 2.3 Core good producers Core goods are exported or sold domestically to the W-firms. They operate under monopo- listic competition and, therefore, have price-setting power. They face standard nominal and real rigidities: Adjusting the output price and investment is costly. Utilizing the installed capital stock beyond the technological optimum accrues costs at an increasing rate. Production is subject to a hierarchical CES structure. At the top of the hierarchy, core goods are produced by combining a capital-labor composite and energy services (purchased from E-firms and amounting to 3.7% of core good revenues). The capital-labor composite is produced by combining a capital service and labor (purchased from L-firms and amounting to 45.6% of the revenues including payroll taxes). The capital service is produced using private and public physical capital measured in units of the final Y-good. While the public capital stock is a free input to production, the private capital expenditures amount to 22.5% of the revenues. About 3.7% of the revenues are used for purchasing energy. The compensation of em- 8 In reality, 46% of the Turkish energy output is used by the energy sector itself as an input. For the sake of simplicity, we neglect this share and only consider the energy used for final consumption and non-energy production. 10 ployees including payroll taxes amounts to 45.6% of the revenues. The remaining 50.7% of the revenues constitute the gross operating surplus which is used for capital expenditures (22.5% of revenues), corporate income tax (1.7% of revenues), and dividends plus interest payments (26.5% of revenues). Similar to Albonico et al. (2019), we assume capital is firm-specific rather than traded on a spot market. Hence, capital is not rented out to the most productive enterprise but sits with the firm. Capital can only be changed by investing and disinvesting. This adds critical frictions to the low-carbon transition and creates the problem of asset stranding in the fossil energy sector. Borrowing constraints are crucial. The high-skilled household’s problem has shown that investors ask for an equity risk premium over the risk-free rate. Since households own the firms, this has implications for the firm’s preferred source of finance. Because of higher risks, the return on equity RS,t is higher than the interest rate on corporate borrowing RB,t . Hence, the owners of the firms prefer to boost the cash flow in the short run by excessive borrowing. This is because future profits are discounted by a rate RS,t (with discount factor 1/RS,t ) that is higher than the rate of interest for new borrowing RB,t . In other words, the benefit of an additional unit of cash flow today exceeds the value of the debt repayment due tomorrow. To maximize the cash flow, households ask the firms to exploit all available credit lines until borrowing constraints become binding. Following Gerali et al. (2010), we assume that the debt obligation in t + 1 cannot exceed the a given fraction λ of the expected value of capital in t + 1 which serves as the collateral. That is, we impose RB,t BZ,t ≤ λEt PY,t+1 QZ,t+1 (1 − δ )KPZ,t , in the firm’s profit maximization problem where RB,t is the gross interest rate for corporate borrowing, BZ,t is corporate debt, λ is a constant, PY,t is the price level, QZ,t is Tobin’s q or the value of capital, δ is the rate of capital depreciation, and KPZ,t is the private capital stock in the core good sector Z . This introduces to our model a financial-accelerator mechanism along the lines of Kiyotaki and Moore (1997): Borrowing constraints are pro-cyclical as they vary with capital valuation. λ = 33.8 is calibrated to achieve the empirical corporate debt- GDP ratio of 57% at the steady state. The optimality condition for end-of-period corporate borrowing implies 1 1 µZ,t = − RB,t RS,t where µZ,t is the shadow price of one additional borrowing unit. The equation collapses to µZ,t = 0 if the returns on equity and borrowing are equal, RS,t = RB,t . In this special case, the firm’s financial structure is irrelevant from the household’s perspective. With a positive equity risk premium, however, RS,t > RB,t and µZ,t > 0. External finance is preferred over internal finance. The end-of-period borrowing constraints are always binding. 11 2.4 Renewable energy, fossil energy, and carbon mining Renewable energy firms (R-firms) are broadly symmetric to Z-firms, with a few important exceptions: They produce renewable energy using capital and labor only, without energy inputs. In contrast to Z-firms, R-firms operate under perfect competition. Nevertheless, R-firms accrue profits because of the assumption of firm-specific capital. Consistent with a large body of literature on renewable learning rates, total factor productivity is subject to a learning-by-doing externality which increases with renewable energy production (Mercure 2012, Rubin et al. 2015, Way et al. 2022). Fossil energy firms (F-firms) are symmetric to R-firms with one crucial difference: The underlying primary energy source – carbon – is a costly input to energy production. We allow some substitution between the capital-labor composite and carbon to capture the switching between coal, oil, and gas. The corresponding elasticity of substitution is estimated. As with renewable sources, we assume that the supply of carbon is perfectly elastic. We distinguish between public investment in energy production and infrastructure to capture the critical role of adequate green infrastructure as emphasized by Way et al. (2022). The capital stock associated with the former adds to the private capital stock as a perfect substitute. Hence, increasing public investment in energy production tends to crowd out private capital. Public infrastructure, however, complements production capital and tends to crowd in private investment as it increases its marginal return. Yet, there are limits: We impose a satiation level of public infrastructure beyond which additional public investment does not increase private capital returns. The cost structures of renewable and fossil energy production are quite different since R-firms are more capital-intensive than F-firms: The compensation of employees, including payroll taxes as a share of total revenues, amount to 12.7% and 11.9%, respectively. The F-sector spends around 71% of its revenues on raw coal, oil, and gas – the primary energy sources. The gross operating surplus in the R-sector is sizable (88.3% of revenues) compared to F-sector (17.1% of revenues). Yet, the bulk of it is needed for purchasing capital goods (58.8% of revenues in the R-sector vs. 8.2% in the F-sector). The corporate income tax shares are 1.9% and 0.6%, respectively. Differences in the dividend and interest payment shares reflect the different capital intensities: 26.6% and 10.4%, respectively. For the sake of simplicity, we assume one global market for carbon sourced by domestic and international carbon mining. Carbon mines are owned by high-skilled households (13%) and the RW (87%). Any demand is fully accommodated at constant shares and a given exogenous world-market price denominated in USD. This implies that the domestic use of domestically extracted carbon is subjected to exchange rate fluctuations. We assume that domestically produced and imported carbon is used only by the fossil energy sector.9 We further assume that carbon can be extracted without capital or labor input. Hence, carbon sales constitute pure rents. 9 In reality, about 14% of the carbon mining output goes into sectors different from fossil energy. 12 CO2 emissions are linked to the combustion of fossil fuels. There is a direct relationship between carbon input and emissions. The model captures the fact that different coal, oil, and gas have different carbon intensities by allowing the F-firms to substitute carbon with capital and labor, thereby adjusting the carbon intensity of fossil energy (for instance, in response to an upstream carbon tax). 2.5 The public sector and monetary policy The public sector features a rich set of revenue and expenditure instruments. The revenue side comprises final consumption taxes (on average around 11% of GDP), taxes on the H- households’ stock of wealth (1% of GDP), taxes on capital income including capital gains, high- and low-skilled labor income taxes (4% of GDP), payroll taxes (7% of GDP), corporate income taxes (2% of GDP), fuel taxes (1.4% of GDP) and upstream carbon taxes (which T¨urkiye does not currently have). The expenditure side of fiscal policy consists of public consumption of the final good (on average around 14% of GDP) and public investment in the core good sector (3.7% of GDP) as well as in the renewable and fossil energy production (0.1% and 0.1% of GDP, respectively) and infrastructure (0.1% and 0.1% of GDP, respectively). In the short run, public consumption and investment respond endogenously to the final output to capture automatic stabilizers. Finally, the government grants fossil and renewable energy subsidies to E-firms and transfers to L-households. A possible deficit is financed by issuing government bonds. Even though government urkiye short-term bond bonds (and domestic deposits) are assumed to be risk-free, the T¨ return has exceeded the policy rate by around 0.4%. We model this as a flight-to-safety premium as in Smets and Wouters (2003). The monetary authority sets the policy rate according to the Taylor rule. Given an average inflation rate of around 8%, over 2010-2016, which is close to the central bank’s target of 5%, the corresponding policy rate consistent with a 10% long-term unemployment rate is around 8.8%. The simulation will adjust these numbers to reflect the current monetary policy stance. Over the business cycle, the policy rate responds to core inflation and the output gap. In the model, inflation targeting and exchange rate targeting are observationally equivalent as the uncovered interest rate parity condition implies that inflation and exchange rates move together. Note further that the Taylor principle does not necessarily hold in the model due to dis-equilibrium unemployment and steady-state wealth. With a sufficiently low labor- market sensitivity of low-skilled wages, the inflation sensitivity of the policy rate may be smaller than one (Schoder 2020). Note that around 50% of Turkish wage earners receive the statuary minimum wage (DISK 2022). 13 3 Model estimation Given the limited information in macroeconomic data, not all model parameters can be esti- mated. A few weakly identified parameters have been calibrated according to microeconomic evidence; others have direct counterparts in the data and have been calibrated accordingly. Some parameters have been restricted for specific model variables to match their empiri- cal counterpart at the steady state. All structural parameters have been estimated using standard Bayesian techniques as outlined in Herbst and Schorfheide (2015). We exploit annual and quarterly time series for 31 observed variables over 2000Q2-2022Q3. Appendix B discusses all the details of our estimation strategy: data and sources, calibration, prior distributions, and the estimated posterior distributions. It also compares our findings to the empirical literature. Almost all structural parameters are identified and broadly in line with the literature. A few findings are worth noting: First, we estimate the households’ risk aversion parameter lower than what C ¸ ebi (2012), Smets and Wouters (2007), and Albonico et al. (2019) find for T¨urkiye, the US, and the EU, respectively. This is unsurprising given that our model explicitly features wealth in the utility function to capture precautionary saving motives. Second, all elasticities of substitution are identified by data except the capital-labor elasticities for renewable energy firms. We keep unidentified parameters in the estimation to consistently account for parameter uncertainty in the simulation results below. Third, we estimate the value-added and energy elasticity as 0.75, slightly above the industry estimates range in Van der Werf (2008). This is because our core goods sector is a mixture of sectors with different degrees of energy intensity. Therefore, our elasticity is also capturing substitution between sectors. Fourth, our estimate for the elasticity of substitution between carbon and value-added in the fossil energy sector is very low (0.05), which indicates the difficulty of substituting out carbon in fossil energy production and implicitly the low degree of substitution between fossil fuels with different carbon contents in T¨urkiye. Note that this makes the incidence and effects of carbon and fuel taxes almost identical in our model. Fifth, the data strongly identify the long-run elasticity of substitution between renewable and fossil energy with a posterior mean of 18.5 and a 90% High-Density Interval (HDI) of [10.1, 26.6]. As Figure 2 shows, our short-run elasticity starts from low values, covering the low elasticity case of Acemoglu et al. (2012) and estimations of Papageorgiou et al. (2017), and converges to its long-term mean in around 20 quarters.10 Sixth, the employment elasticity of low-skilled wage inflation is around 1.29. Recall that 10 Acemoglu et al. (2012) consider the value of 3 as low and 10 as a high elasticity. Papageorgiou et al. (2017) estimates the short-run elasticity of substitution between dirty and clean energy around 2 for electricity and around 3 for non-energy sectors. Argentiero et al. (2017) finds an elasticity around 1.2 for EU countries. We need to note, however, that the empirical literature is rather thin on the elasticity of substitution between clean and dirty energy. 14 Figure 2: Mean and 90% credibility interval of the time-varying elasticity of substitution between renewable and fossil energy. an elasticity of zero corresponds to the pure Keynesian case of dis-equilibrium unemployment while an elasticity going to infinity captures the pure Walrasian case of a general equilibrium (Schoder 2017, 2020). The estimated elasticity is relatively high and favors the Walrasian case over the Keynesian case.11 This is not surprising, however, given the rapid growth of the Turkish economy during the sample period. 4 The macroeconomic effects of a climate-fiscal policy mix for net-zero Employing the estimated macro model outlined above, this section identifies a climate-fiscal policy mix to achieve net zero in T¨ urkiye by 2053 while observing significant feasibility constraints and reports its projected macroeconomic repercussions. For this endeavor, we use conditional forecasts. This method restricts the future paths of constrained variables such as carbon emissions and computes the policy shocks needed to achieve the predetermined paths. Regarding the information set of the agents, we assume they do not anticipate the policy changes but, once implemented, perceive them as perma- nent. Conditional forecasting requires a linear approximation of the model. To capture the non-linearity of the transition path, we iteratively update the steady state around which the model is linearized.12 The benchmark for the low-carbon transition scenario is a no-policy or business-as-usual baseline scenario. Because of persistence in macroeconomic adjustment, the dynamics at 11 In comparison, we find an elasticity of 0.19 for the Euro Area (Schoder and Tercioglu 2023). 12 In the last period of the sample, we obtain the smoothed value of each variable. We then linearize around the estimation sample’s steady state and obtain a one-period ahead forecast starting from the economy’s current position and imposing the policy changes necessary to achieve the targeted paths. Taking these policy shocks as permanent, we update the steady state and linearize around it. We obtain another one- period ahead forecast from the previously predicted position of the economy. The process is repeated until the end of the forecast horizon. 15 the end of the estimation sample spill over to the projection period. Consistent with OECD (2022), we limit the policy rate hike, which the model predicts, to capture the current monetary policy stance in T¨urkiye. This affects the public debt trajectory, too. Hence, we adjust government consumption to maintain the initially projected path of public debt. This path matches the recent IMF (2022) forecast. In the baseline, variables converge to their steady growth path in the long run. Labor productivity is assumed to grow by 4% per year, which drives the growth of all trending variables, including baseline emissions.13 4.1 Dynamic multiplier effects of climate-fiscal interventions To identify a climate-smart policy mix under feasibility constraints, it is conducive to first study the macro implications of the core climate-fiscal policy interventions. We consider a carbon tax, a renewable energy subsidy, public investment in renewable energy generation, and public investment in renewable energy infrastructure below and above the satiation point. The policy interventions are permanent and phased in with an auto-regressive pa- rameter of 0.8. We normalize each policy shock to 1 billion (bn) 2015 USD before any macroeconomic adjustment occurs. This corresponds to around 0.1% of 2021 GDP. Figure 3 reports the impulse-response functions (IRFs) for critical macroeconomic indi- cators. First, note that while the policy impulse of $1bn is the same for each instrument, the fiscal implications differ over time. For instance, the carbon tax decreases the public deficit only by around $0.7bn. This is because the tax successfully reduces emissions and erodes its base. In contrast, raising the renewable subsidy rate increases the fiscal stimulus over time as renewable energy expands. Hence, the primary deficit slightly overshoots 1bn after ten years. Renewable infrastructure investments of $1bn cause a primary deficit of $1.5bn after ten years, assuming the satiation point is unmet. This is because of the decline in fuel tax revenues. Second, infrastructure investment below the satiation point is highly effective in reducing emissions. This is consistent with Way et al. 2022 who argue that insufficient infrastructure investment may be a bottleneck for the low-carbon transition. Ten years after the policy change, $1bn of additional annual spending reduce emissions by around 40 million (mn) tonnes per year. By the same time, annual renewable energy production increases to around 6mn tonnes of oil equivalent (toe), and annual fossil energy production decreases by 5mn toe. In comparison, the carbon tax and the renewable energy subsidy reduce emissions only by around 6mn tonnes per year after 10 years. The carbon tax achieves this mainly by reducing the energy intensity of production and the subsidy by spurring renewable energy production. Both public investment in renewable generation and infrastructure at the sa- 13 Note that the baseline scenario abstracts from two important developments: First, it neglects the Carbon Border Adjustment Mechanism which the European Union intends to implement on its imports. Hence, the baseline projections of Turkish exports may be too optimistic. Second, the baseline neglects the productivity spill-overs from scaling up renewable energy production in the RW. It is, therefore, likely to underestimate the baseline emission reductions. 16 Public primary deficit Carbon emissions Renewable energy production 2 10 7 6 Millions of tonnes of CO2 emissions 1.5 0 Millions of tonnes of oil equivalent 5 Billions of 2015 USD 1 -10 4 0.5 -20 3 2 0 -30 1 -0.5 -40 0 -1 -50 -1 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 Quarters after policy shock Quarters after policy shock Quarters after policy shock Carbon tax Renewable energy subsidy Public investment in RE generation Pub. inv. in RE infrastructure (below saturation) Pub. inv. in RE infrastructure (above saturation) Fossil energy production Gross domestic product Private investment in renewable energy 1 6 4 0 5 3 Millions of tonnes of oil equivalent -1 4 Billions of 2015 USD Billions of 2015 USD 2 -2 3 1 -3 2 0 -4 1 -5 0 -1 -6 -1 -2 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 Quarters after policy shock Quarters after policy shock Quarters after policy shock Carbon tax Renewable energy subsidy Public investment in RE generation Pub. inv. in RE infrastructure (below saturation) Pub. inv. in RE infrastructure (above saturation) Private investment in fossil energy Exports Non-oil imports 0.2 0.1 1.5 0 0 -0.2 -0.1 1 Billions of 2015 USD Billions of 2015 USD Billions of 2015 USD -0.4 -0.2 -0.6 -0.3 0.5 -0.8 -0.4 -1 -0.5 0 -1.2 -0.6 -1.4 -0.7 -0.5 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 Quarters after policy shock Quarters after policy shock Quarters after policy shock Carbon tax Renewable energy subsidy Public investment in RE generation Pub. inv. in RE infrastructure (below saturation) Pub. inv. in RE infrastructure (above saturation) Oil imports Headline inflation Energy price inflation 0.5 0.2 2 0 0 1 -0.5 -0.2 Billions of 2015 USD Percentage points Percentage points 0 -1 -0.4 -1.5 -0.6 -1 -2 -0.8 -2 -2.5 -1 -3 -1.2 -3 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 Quarters after policy shock Quarters after policy shock Quarters after policy shock Carbon tax Renewable energy subsidy Public investment in RE generation Pub. inv. in RE infrastructure (below saturation) Pub. inv. in RE infrastructure (above saturation) Figure 3: The macroeconomic effects of selected permanent climate-fiscal policy interven- tions. Each policy shock is normalized to move 1 billion of constant 2015 USD before any macroeconomic adjustment. 17 tiation point only have minor effects on CO2 emissions. For better comparability, consider the 10-year cumulative emission multiplier effects of the carbon tax, the renewable subsidy, and infrastructure investment below the satiation point. Moving $1bn over ten years re- duces emissions by 4.3mn, 3.9mn, and 22.2mn tonnes, respectively. These results are highly consistent with Deleidi et al. (2020), whose panel study covering 17 countries suggests that public investment is considerably more effective in mobilizing private finance than carbon taxes or renewable subsidies. Third, there are critical macroeconomic differences between the policy instruments. The carbon tax contracts GDP by around $0.4bn after ten years. The renewable subsidy raises GDP by around $1bn. At the margin, green infrastructure investment below the satiation point boosts GDP even by $5.6bn. These instruments’ 10-year cumulative GDP multipliers are -0.4, 0.74, and 3.37, respectively. Public investment in generation and infrastructure at the satiation point only have minor GDP effects. The economics behind these results is straightforward: The carbon tax increases energy prices and reduces private consumption and investment. The subsidy encourages private activity by lowering energy prices and stimulating the demand for renewable energy. Public infrastructure investment not only stimulates aggregate demand but also increases the marginal returns of private capital. Forth, none of the climate-fiscal instruments considered here affects net exports consid- erably. Nevertheless, the composition of trade changes fundamentally, causing significant friction. Depending on the policy instrument, exports decrease by up to $0.6bn. Only public investment in renewable energy generation has no impact on exports. Infrastructure invest- ment below the satiation point raises non-oil imports strongly by up to $1.5bn. Given an import share of fossil fuels of 87%, the responses of the oil imports mirror those of CO2 emissions. The carbon tax and the renewable subsidy reduce oil imports by up to $0.4bn. Infrastructure investment reduces oil imports by up to $2.6bn. Fifth, to the extent that climate-fiscal policy instruments push the unemployment rate below (or above) the non-accelerating inflation rate of unemployment (NAIRU), headline inflation may settle above (or below) the inflation target. For instance, the carbon tax increases the labor intensity of production. This tightens the low-skilled labor market and translates into higher wages and, eventually, price inflation of around 0.08%-points above target. Infrastructure investment at the satiation point has the same effect. The renewable subsidy reduces the labor intensity of production, causing inflation to settle 0.1%-points below target. Infrastructure investment below the satiation lowers inflation considerably by up to 1%-point. As discussed in the previous sensitivity analysis, the central bank’s response to transition-induced deviations of inflation from the target critically determines the GDP response. 4.2 A climate-fiscal policy mix for net zero under constraints The net-zero scenario involves the following targets and constraints: Broadly consistent with (IPC 2021, World Bank 2022), we require CO2 emissions to decrease linearly from 400mn 18 Figure 4: A set of climate-fiscal policy interventions to realize the pathway to net zero under urkiye. The shaded area is the 90% HDI. feasibility constraints in T¨ tonnes in 2022 to 68mn tonnes in 2053. Regarding debt stability, we require public debt to decrease until it reaches 34% of trend GDP. After that, it stays at this level. For political economy reasons, we require that low-skilled consumption is not lower than in the baseline scenario. We also require that public investment ensures fossil and renewable infrastructure are at their respective satiation levels. Finally, a public bad bank is assumed to purchase stranded private assets at replacement cost. urkiye can achieve the pathway to net zero by 2053 under the Figure 4 illustrates how T¨ above-mentioned constraints. It displays projections at the posterior mean as well as an approximation of the 90% HDI reflecting posterior parameter uncertainty.14 We broadly distinguish between three phases: (a) the fiscal consolidation, (b) the renewable energy boom, (c) and the final push. urkiye The phase of fiscal consolidation starts in 2022Q3 and lasts for about 5 years. T¨ seeks to achieve the twin goals of reducing emissions and public debt. Fossil fuel subsi- dies, amounting to 0.4% of GDP in 2022, are gradually phased out within ten years. The government introduces a carbon tax that quickly increases to $70 per tonne of CO2 until 2026Q4. With economies of scale kicking in only at higher levels of renewable generation, the initial phase requires a big-push policy (Ploeg and Venables 2022). Note that, in the initial phase, the posterior parameter uncertainty illustrated in Figure B1 of Appendix B 14 The HDI approximation has been obtained from a Monte-Carlo simulation of the conditional forecasts. In particular, we took 500 draws from the posterior distribution and ran the conditional forecasts. We excluded draws that resulted in explosive forecasts or computational errors. 19 does not translate into considerable forecast uncertainty of the projected carbon tax. Public investment in fossil infrastructure decreases from 0.1% of trend GDP to 0.01%. Note that private investment is sensitive to public infrastructure investment. Hence, the phase-out of fossil infrastructure investment – including the sharp decline in the first few quarters – is designed to keep the risk of private asset stranding small.15 Nevertheless, the 90% HDI of stranded asset purchases indicates a considerable risk of asset stranding already in the early stages of the transition. To compensate for the negative effect of the carbon tax on low-skilled consumption, transfer payments increase from 8.1% of trend GDP in 2022 to 8.7% in 2026Q4. While the rapid phase-out of fossil investment reduces the risk of asset stranding, it generates uncertainty regarding the required public support of low-skilled households as indicated by the wide 90% HDI during the fiscal-consolidation phase. Renewable infrastructure investment rises from 0.1% of trend GDP to 0.195% within ten years. We assume that this exactly matches the renewable infrastructure satiation level. It corresponds to the $2bn of annual investment needs in renewable infrastructure identified in World Bank (2022). By 2027, emissions decline below 340mn tonnes – the reference value urkiye to stay on the 2°C path in 2030 (Voyvoda and Yeldan 2015). calculated for T¨ At the posterior mean, the target public debt-GDP ratio of 34% is achieved in 2027Q1. This is when the phase of the renewable energy boom starts.16 Instead of reducing public debt, carbon tax revenues are now used to finance renewable energy subsidies while keeping public debt at a constant share of trend GDP.17 The subsidy peaks in 2029Q1 at $1400 per toe, which amounts to a sizable bill of 0.62% of GDP. As emissions decrease, the revenues from the fuel tax and the carbon tax decline. To continue meeting the public debt target, renewable subsidies are phased out gradually until 2037. During the renewable energy boom, carbon taxes can be temporarily lowered as the renewable subsidies sufficiently reduce emissions to stay on target. Note that there is considerable parameter uncertainty regarding the size of the renewable subsidy and the duration of the boom. At the peak, the 90% HDI ranges from $300 to $2800 per toe. Regarding phase-out time, the 90% HDI ranges from 2032 to 2042. Given strict public debt and low-skilled consumption targets, the uncertainty about the macroeconomic effects of the policy mix considered translates into uncertainty about how much public funds there will be available for renewable subsidies. At the posterior mean, revenues will be insufficient to sustain a constant debt-trend GDP ratio, even without subsidies, by 2037. This is the beginning of the phase of the final push. The carbon tax increases from $90 in 2037 to $175 in 2053. The carbon tax required to stay on the path to net zero is subject to only moderate parameter uncertainty. For instance, the 15 We define stranded capital assets as the private capital stock which F-firms do not wish to hold any longer. Stranded assets are valued at replacement costs. 16 In the Monte Carlo simulation of the 90% HDI, we fix the start date of the renewable subsidy to 2027Q1 and hold the public debt-trend GDP ratio constant at whatever value it has in that period. 17 To ensure a smooth transition, the subsidies are phased in already in 2026Q2. 20 Figure 5: The historical and projected paths of private investment and production in the renewable and fossil energy sectors. 90% HDI in 2053 ranges from $160 to $220. Despite increasing tax rates, carbon tax revenues decline. To ensure debt stability, gov- ernment consumption declines from around 14% of trend GDP in 2037 to 12.1% in 2053. Low-skilled households are compensated with transfers up to 9.1% of trend GDP until 2053 – 1%-point higher than in 2022. During the final push, the risk of fossil asset stranding is high. In our posterior-mean simulation, the public bad bank’s annual purchases of stranded capital assets almost reach 0.02% of trend GDP in 2053. 4.3 Macroeconomic repercussions of the net-zero policy mix The policy package proposed for the Turkish pathway to net zero is projected to critically affect key macroeconomic indicators. We focus on energy investment and consumption, GDP, trade, household welfare, fossil equity share prices, and stranded assets. 21 4.3.1 Energy investment and consumption Figure 5 plots the historical and projected paths of private investment and production in the renewable and fossil energy sectors. The climate-fiscal policy mix previously discussed is expected to mobilize considerable private capital to finance the low-carbon transition – especially during the renewable energy boom triggered by the renewable subsidy. We project private investment in renewable energy to increase from around $4bn in 2022 to $150bn in 2053. This is almost 4% of trend GDP or around 12 times the renewable investment of the baseline scenario. Note that it is also about ten times the baseline fossil investment. It matches the annual investment needs of 4% to 4.5% of GDP identified by Ranger and Volz (2023). The overall energy capital requirements of the transition exceed the baseline needs be- cause renewable energy is entirely produced domestically. Recall that, in the baseline, T¨urkiye imports 87% of its total primary energy supply. Moreover, the renewable energy sec- tor is considerably more capital-intensive than the fossil energy sector (which, in our model, excludes carbon extraction). It is important to emphasize that the projected expansion of renewable investment presumes ideal energy market regulation for energy production to be profitable. Private investment in the fossil energy sector starts from $4bn in 2022 and disappears entirely. Given the policy mix, fossil investment is expected to turn negative by 2042. That is, capital assets are stranding. They are decommissioned and purchased by the public bad bank. Renewable energy consumption increases from around 20mn toe in 2022 to 420mn toe by 2053. This is 460% higher than the baseline projection. By the early 2030s, renewable energy will reach 50% of the energy mix. This is similar to net-zero pathway findings of IPC (2021) for 2030. Fossil energy consumption will decrease from 130mn toe in 2022 to around 20mn toe by 2053. This is 96% below the baseline projection. By 2053, the overall energy consumption will be 20% below the baseline. The carbon tax not only triggers the substitution to renewable energy but also to capital and labor, reducing the energy intensity of GDP. 4.3.2 GDP and components For the climate-fiscal policy mix previously identified, the first three panels in Figure 6 plot the projected responses of GDP, consumption, investment, exports, and imports. Overall, the GDP effects are moderate, staying within 1% of GDP. While consumption does not deviate significantly from the baseline, investment reaches levels of 4% above baseline by 2030 and almost 6% by 2053. The low-carbon transition is investment driven. On the flip side, net exports decline to a similar extent. Overall, considerable parameter uncertainty exists regarding the net effects on GDP and its components. During the phase of fiscal consolidation, GDP declines by up to 0.7% in 2027 compared to 22 Figure 6: The macroeconomic implications of a policy mix to achieve net zero in T¨ urkiye by 2053. Historical and projected paths of carbon taxes, renewable energy subsidies, renewable and fossil energy production, renewable and fossil private investment, GDP, consumption, and investment. baseline – a finding that is small but in line with the literature (Voyvoda and Yeldan 2015). The significant expansionary effect of public spending on renewable infrastructure consider- ably dampens the contractionary effects of the carbon tax and the fossil subsidy phase-out. During the renewable energy boom, GDP recovers slightly exceeding baseline until the mid- 2030s. This is driven by the renewable energy subsidy, which stimulates GDP through two important channels: lower energy prices and an investment boom. Note that endogenous productivity gains from scaling up renewable production reinforce the subsidy effects on en- ergy prices and investment. During the final push, when government consumption declines to maintain public debt stability, GDP decreases to 1% below baseline in 2053. To put the GDP effects in context, we consider a pure fiscal-consolidation scenario that does not seek to decarbonize the economy but adjusts public consumption to achieve the 23 same public-debt trajectory we obtain from the decarbonization scenario. It is worth noting that the initial contraction of GDP is much more substantial: 4.5% below baseline in 2025Q1. Hence, combining carbon pricing and public investment in renewable infrastructure is a cost- effective and climate-smart way of fiscal consolidation. Trade is subject to transition risks. On the positive side, the low-carbon transition will allow T¨urkiye to eliminate its dependence on fossil fuel imports which currently amount to 87% of the total primary energy supply. Total net exports decrease by around 4% of GDP, which may be considered moderate. Yet, exports will steadily decline by up to 11% in 2053 compared to the baseline. The inflationary pressure of higher energy prices will lead to a real appreciation of the Turkish lira. The loss in competitiveness translates into a decline in exports, whose stabilization may require additional policy measures that support exporting firms. As a qualification of this result, note that our baseline scenario (against which the losses of the policy scenarios are measured) does not reflect the EU’s Carbon Border Adjustment Mechanism. Given that half of the Turkish exports are going to the EU, the baseline, no-transition exports are likely to be overestimated. 4.3.3 Household welfare Even though the GDP effects of the low-carbon transition are expected to be moderate in T¨urkiye, this may not hold for household welfare. At the first order, households care about how much they can consume in goods and services and how much leisure they can enjoy. High-skilled households additionally care about how much financial wealth they hold. Therefore, we follow the method of Born and Pfeifer (2020) and assess unconditional welfare effects over time using the concept of consumption equivalence.18 As shown in the fourth panel of Figure 6, the low-skilled household welfare effects along the transition path are moderate, with the consumption equivalent moving between -1.5% and 2% of baseline consumption. This is because of the transfer payments that keep low- skilled consumption at the level of the baseline scenario. Without these payments, low-skilled households would strongly prefer the baseline over the transition scenario. The consumption equivalent would reach around -5% (not shown). That is, they would have to forsake up to 5% of their baseline consumption to be indifferent between the baseline and the transition. Overall, high-skilled households benefit more from the transition than low-skilled house- holds – even though they do not receive transfers. For comparison, we first consider the case in which high-skilled households do not derive utility from wealth.19 The aggregated high- and low-skilled utility functions are identical in that case. We find that the decline in GDP initially translates into a reduction in high-skilled consumption, reducing welfare. 18 The welfare difference between the transition and baseline scenario is measured as the share of consump- tion which a household living in the baseline has to receive to be indifferent between the baseline and the low-carbon transition. Note that we evaluate welfare only at the first order. 19 Assume, for instance, that household-level wealth enters the utility function as the difference to aggregate wealth. The first-order conditions remain the same, but the wealth term in the utility function now has the interpretation of Keeping Up with the Joneses rather than precautionary saving. 24 Starting with the renewable energy boom, the high-skilled household’s welfare improves with a consumption equivalence of up to 2% by 2053. In the case of wealth utility, welfare effects are much more substantial. The consumption equivalent during the fiscal consolidation reaches -14% in 2025. Yet, during the renewable boom, it goes up to 43% and later stabilizes around 20%. In other words, to make the baseline as appealing as the transition scenario, high-skilled households in the baseline would require a significant amount of additional consumption. The welfare effects are substantial because the policy interventions, which are perceived as permanent once implemented, significantly affect the net present value of wealth. Since the renewable energy sector is more capital- intensive than the fossil energy sector, the low-carbon transition implies a net increases in capital assets which are held by the high-skilled households. Since wealth is a source of utility, high-skilled households benefit from the transition. As shown in the fourth panel of Figure 6, the wealth channel is a major driver of household welfare. While these results highlight the importance of studying the welfare impacts of the low- carbon transition on households, they likely draw too gloomy a picture as the model leaves out critical transmission channels: First, as many studies have shown, the benefits of the reduction in CO2-related air pollution are sizable and outweigh the purely consumption and leisure-related welfare loss of the low-carbon transition (Parry et al. 2015, Karlsson et al. 2020). Second, the model’s assumption that the high- and low-skilled consumption bundles are identical further distorts the picture as the former tends to be more energy- and carbon- intensive than the latter (Sager 2019, Dorband et al. 2019). 4.3.4 Stranded assets Financial risks arise from the devaluation of private equity shares and the stranding of productive capital assets in the fossil energy sector. These risks are manageable because the volume of fossil capital assets in T¨ urkiye imports most urkiye is relatively tiny. Recall that T¨ of its fossil fuels. Nevertheless, a disorderly transition and lousy policy communication may increase these transition risks. The last two panels in Figure 6 show the responses of equity share prices and stranded assets to an orderly and delayed transition. The longer the transition is delayed, the more sudden the drop in asset prices and the higher the financial risks. Regarding asset stranding, the public bad bank needs to purchase excess capital assets only starting in 2044, according to our simulation of an orderly transition. Evaluated at replacement costs, the volume of the stranded capital stock will reach almost 0.1% of trend GDP by 2053. In comparison, the total public capital stock is 29% of trend GDP. A delayed transition and an ill-timed phase-out of public fossil infrastructure investment increase the risk of asset stranding. For instance, delaying the start of the transition until 2030 will generate significant stranded assets already by 2033, and the volume will reach 0.63% of trend GDP in 2043. Policy communication and commitment to the net-zero agenda reduce the risk of asset stranding (Bretschger and Soretz 2022). 25 Figure 7: The response of selected variables to the policy mix identified in Section 4.3 in various counterfactual scenarios. 4.4 Supporting policies: Policy commitment and structural re- forms This section investigates how policy commitment and structural reforms can support the low-carbon transition and improve the macroeconomic outlook of the climate policies imple- mented. In particular, we simulate the macroeconomic effects of the policy mix identified in Section 4.3 in various counterfactual scenarios: (a) credible policy commitment, (b) expected policy reversal, (c) improved rule of law, (d) and facilitated access to green finance. 4.4.1 Expectations regarding policy commitment In the previous analysis, we assumed that agents do not anticipate the policy interventions required to achieve net zero or do not deem the government’s net-zero commitment credible. Yet, once implemented, agents expect the policy changes to be permanent. No policy reversal is expected. In this section, we consider two additional scenarios regarding the credibility of the net zero pledge. In each expectation scenario, we take the same dynamic policy mix as identified above and simulate its macroeconomic effects. This allows us to isolate the effects of expectations regarding policy commitment. In the credible-commitment scenario, the entire path of future policy interventions is communicated to the public. The government is fully committed to the announcement and the public deems it credible. In the policy-reversal scenario, agents observe a policy change once implemented but, instead of perceiving it as 26 permanent, they expect it to slowly phase out again – for instance, due to anticipated social unrest or vested interests. We assume an auto-regressive parameter of 0.95. That is, agents expect that any given policy change will be reversed by 50% after about 4 years. As Campiglio et al. (2023) points out, there are multiple real-world examples of policy reversals that justify this type of expectation. As shown in Figure 7, transparent policy communication and fully credible commitment can crowd in private sector behavior, slightly raising renewable investment and generation. Net zero is already reached in the late-2040s. Since the different stages of the transition phase are anticipated, agents can smooth behavior and reduce adjustment costs and friction. This reduces the GDP contraction of the fiscal-consolidation phase and raises and prolongs the GDP expansion during the scaling-up phase. The private-sector expectation that a future government will take back already imple- mented measures makes the transition considerably more costly. Even with an expected slow policy reversal of 5% each quarter, GDP is projected around 0.5% below the bench- mark policy scenario for most of the transition period. Even worse, the policy mix that is sufficient for net zero in the benchmark scenario, will fail to achieve the transition in the weak-commitment scenario. Emissions are projected to decline by only 55% compared to the no-policy baseline. Hence, to achieve net zero the carbon tax would have to be considerably higher than in the benchmark policy scenario further reducing GDP. Credible commitment reduces the costs of decarbonization. Yet, as is well-understood since Kydland and Prescott (1977), policy commitments are credible only with the corre- sponding institutional setting. Barrett (2008) suggests to amend climate treaties with an enforcement mechanism, such as a trade restriction. Nordhaus (2015) proposes a climate club intended to eliminate the problem of free-riding. On a national level, Chile pioneers sustainability-linked sovereign bonds, which link the bond return to green performance in- dicators and partly align the government’s interests with the low-carbon agenda (Gir´aldez and Fontana 2021). Our results suggest that these measures would considerably lower the costs of the low-carbon transition by anchoring private-sector expectations. 4.4.2 Structural reforms The OECD (2021) has identified a weak rule of law as a main obstacle for private-sector investment dynamics in T¨ urkiye. To assess the role of the business uncertainty for the low- carbon transition, Figure 7 also illustrates the scenario in which the government implements reforms that reduce the equity risk premium to the average of China, India, and South Africa within five years. These reforms drastically improve the macroeconomic outlook of the low- carbon transition, with GDP peaking at 3% above the no-policy baseline. Renewable energy investment and generation are considerably higher than in the scenario with the benchmark risk premium. The economics of these results is straightforward: With a lower equity-risk premium, firms reduce the discounting of future profits. Capital investment becomes more appealing. Since renewable energy is very capital-intensive, this sector mainly benefits from 27 a lower risk premium. Regarding bank lending, the OECD’s recommendation for T¨ urkiye is to reduce private sector debt OECD (2021). Nevertheless, we consider the counterfactual scenario in which the government implements reforms that gradually reduce constraints on green finance within five years. In particular, we assume that the share of capital which banks accept as collat- eral for lending to renewable energy firms increases by 20%-points. Figure 7 illustrates the results. Similar to the scenario of an improved rule of law, renewable energy investment and generation increase considerably. The GDP response demonstrates the importance of the financial accelerator mechanism forcefully. With a higher debt-capital ratio and borrowing constraints proportional to the value of capital, business activity becomes more pro-cyclical. As the value of the renewable capital stock decreases in the fiscal-consolidation phase, borrow- ing constraints tighten up, and GDP contracts considerably. On the flip side, the subsequent boom initiated by the renewable energy subsidy alleviates borrowing constraints, reinforcing the expansion of GDP. 5 Concluding remarks In conclusion, this paper examines the feasibility of T¨ urkiye’s net-zero emissions target by 2053, given the country’s structural deficits and recent climate-related risks. The country’s dependence on fossil fuel imports and the volatility of energy prices due to exchange rate fluctuations pose significant risks to energy security, while its primary export market, which is more carbon-intensive than its European counterpart, is also under threat from the Euro- pean Union’s low-carbon agenda. Therefore, T¨ urkiye must implement effective climate-fiscal policy interventions that enable the country to grow along the 2053 net-zero pathway, taking into account significant feasibility constraints. The analysis employs an empirical dynamic stochastic open-economy macro framework, allowing for the simulation of low-carbon transitions. The policy instruments evaluated include a carbon tax, a renewable energy subsidy, transfer payments, public infrastructure investment, a bad bank for stranded assets, and the phase-out of fossil fuel subsidies and public investment. The study also explores how the macroeconomic effects of the policy interventions differ in counterfactual scenarios of a fully credible commitment to net zero, an improved rule of law, and facilitated access to green bank finance. The study finds that T¨ urkiye’s transition to net zero is feasible with a mix of policy instruments that address the country’s structural deficits. The model simulations show that a credible commitment to net zero, improved rule of law, and access to green bank finance would improve the effectiveness of climate-fiscal policy interventions in mobilizing private finance. Moreover, renewable energy subsidies and public infrastructure investments have a positive effect on economic growth, while the carbon tax and phase-out of fossil fuel subsidies contribute to the decarbonization of the economy. urkiye’s climate policy should focus on the imple- Therefore, the study suggests that T¨ 28 mentation of an integrated set of policy interventions that take into account the country’s structural deficits, particularly the weak rule of law, and low business confidence in public institutions. A credible commitment to net zero, improvements in access to finance and the rule of law, coupled with a mix of policy instruments, can provide a path to sustainable and equitable economic growth while meeting T¨ urkiye’s climate goals. References Acemoglu, D., Aghion, P., Bursztyn, L., Hemous, D. (2012): The environment and directed technical change, American Economic Review, 102(1), pp. 131–166. e, F., Mihoubi, F., Mutschler, W., Pfeifer, J., Adjemian, S., Bastani, H., Juillard, M., Karam´ Ratto, M., Rion, N., Villemot, S. (2022): Dynare: Reference Manual, Version 5, Dynare Working Papers 72, CEPREMAP. es, L., Cardani, R., Croitorov, O., Dio, F. D., Ferroni, F., Giovannini, M., Albonico, A., Cal´ Hohberger, S., Pataracchia, B., Pericoli, F., Pfeiffer, P. (2019): The Global Multi-Country Model (GM): An Estimated DSGE Model for Euro Area Countries, European Economy - Discussion Papers 2015 - 102, Directorate General Economic and Financial Affairs (DG ECFIN), European Commission. An, S., Kang, H., et al. (2011): Oil shocks in a DSGE model for the Korean economy, Commodity Prices and Markets, 20, pp. 295–321. Annicchiarico, B., Di Dio, F. (2015): Environmental policy and macroeconomic dynamics in a new Keynesian model, Journal of Environmental Economics and Management, 69, pp. 1–21. Argentiero, A., Atalla, T., Bigerna, S., Micheli, S., Polinori, P. (2017): Comparing renewable energy policies in EU-15, US and China: A Bayesian DSGE model, The Energy Journal, 38(KAPSARC Special Issue). Barrett, S. (2008): Climate treaties and the imperative of enforcement, Oxford Review of Economic Policy, 24(2), pp. 239–258. Barro, R. J., Grossman, H. I. (1971): A General Disequilibrium Model of Income and Em- ployment, American Economic Review, 61(1), pp. 82–93. Battiston, S., Monasterolo, I., Riahi, K., van Ruijven, B. J. (2021): Accounting for finance is key for climate mitigation pathways, Science, 372(6545), pp. 918–920. Benigno, G., Fornaro, L. (2018): Stagnation Traps, Review of Economic Studies, 85(3), pp. 1425–1470. 29 Benjamin Carton, D. M., Christopher Evans, Voigts, S. (2022): Getting to Know GMMET: The Theoretical Structure and Simulation Properties of the Global Macroeconomic Model for the Energy Transition., international Monetary Fund. Born, B., Pfeifer, J. (2020): The New Keynesian Wage Phillips Curve: Calvo Vs. Rotemberg, Macroeconomic Dynamics, 24(5), pp. 1017–1041. Bretschger, L., Soretz, S. (2022): Stranded Assets: How Policy Uncertainty affects Capital, Growth, and the Environment, Environmental & Resource Economics, 83(2), pp. 261–288. Calvo, G. A. (1983): Staggered prices in a utility-maximizing framework, Journal of Mone- tary Economics, 12(3), pp. 383–398. Campiglio, E., Lamperti, F., Terranova, R. (2023): Believe me when I say green! Heteroge- neous expectations and climate policy uncertainty, LEM Papers Series 2023/12, Labora- tory of Economics and Management (LEM), Sant’Anna School of Advanced Studies, Pisa, Italy. Cantore, C., Ferroni, F., Leon-Ledesma, M. A. (2017): The dynamics of hours worked and technology, Journal of Economic Dynamics and Control, 82, pp. 67–82. ¸ ebi, C. (2012): The interaction between monetary and fiscal policies in Turkey: An esti- C mated New Keynesian DSGE model, Economic Modelling, 29(4), pp. 1258–1267. u¸ Cengiz, D., Tekg¨ c, H. (2022): Is it merely a labor supply shock? Impacts of Syrian migrants on local economies in Turkey, ILR Review, 75(3), pp. 741–768. Chiarella, C., Flaschel, P., Franke, R. (2005): Foundations for a Disequilibrium Theory of the Business Cycle, Cambridge University Press. Christiano, L., Eichenbaum, M., Evans, C. L. (2005): Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy, The Journal of Political Economy, 113(1), pp. 1–45. Christiano, L. J., Trabandt, M., Walentin, K. (2010): DSGE Models for Monetary Policy Analysis, in: Friedman, B. M., Woodford, M., eds., Handbook of Monetary Economics, Elsevier, vol. 3 of Handbook of Monetary Economics, chap. 7, pp. 285–367. Christoffel, K., Coenen, G., Warne, A. (2008): The New Area-Wide Model of the euro area: a micro-founded open-economy model for forecasting and policy analysis, Working Paper Series 0944, European Central Bank. etre, A., Fabre, A., Kruse, T., Planterose, B., Chico, A. S., Stantcheva, S. (2022): Dechezleprˆ Fighting Climate Change: International Attitudes Toward Climate Policies, NBER Work- ing Papers 30265, National Bureau of Economic Research, Inc. Deleidi, M., Mazzucato, M., Semieniuk, G. (2020): Neither crowding in nor out: Public direct investment mobilising private investment into renewable electricity projects, Energy Policy, 140, p. 111195. 30 Diluiso, F., Annicchiarico, B., Kalkuhl, M., Minx, J. C. (2021): Climate actions and macro- financial stability: The role of central banks, Journal of Environmental Economics and Management, 110, p. 102548. DISK (2022): Asgari Ucret Gercegi 2022 Raporu, Tech. rep., Devrimci Isci Sendikalari Kon- federasyonu Arastirma Merkezi. Dissou, Y., Karnizova, L. (2016): Emissions cap or emissions tax? A multi-sector business cycle analysis, Journal of Environmental Economics and Management, 79, pp. 169–188. Dorband, I. I., Jakob, M., Kalkuhl, M., Steckel, J. C. (2019): Poverty and distributional effects of carbon pricing in low- and middle-income countries - A global comparative analysis, World Development, 115, pp. 246–257. Fischer, C., Springborn, M. (2011): Emissions targets and the real business cycle: Intensity targets versus caps or taxes, Journal of Environmental Economics and Management, 62(3), pp. 352–366. Gerali, A., Neri, S., Sessa, L., Signoretti, F. M. (2010): Credit and Banking in a DSGE Model of the Euro Area, Journal of Money, Credit and Banking, 42(s1), pp. 107–141. Gir´aldez, J., Fontana, S. (2021): Sustainability-linked bonds: the next frontier in sovereign financing, Capital Markets Law Journal, 17(1), pp. 8–19. Guerrieri, L., Erceg, C., Bodenstein, M. (2008): Oil Shocks and External Adjustment, Tech. rep., Society for Economic Dynamics. urkaynak, R. S., Kisacikoglu, B., Lee, S. S. (2022): Exchange Rate and Inflation under G¨ Weak Monetary Policy: Turkey Verifies Theory, CEPR Discussion Papers 17248, C.E.P.R. Discussion Papers. Hallegatte, S., Mcisaac, F. J., Dudu, H., Jooste, C., Knudsen, C., Beck, H. A. (2023): The Macroeconomic Implications of a Transition to Zero Net Emissions: A Modeling Framework, Tech. rep., The World Bank. Havranek, T., Irsova, Z., Laslopova, L., Zeynalova, O. (2020): The elasticity of substitution between skilled and unskilled labor: A meta-analysis, unpublished. Herbst, E., Schorfheide, F. (2015): Bayesian Estimation of DSGE Models, The Econometric and Tinbergen Institutes Lectures, Princeton University Press. Hodrick, R., Prescott, E. C. (1980): Post-war U.S. business cycles: an empirical investigation, Discussion Paper 451, Carnegie-Mellon University. IEA (2021): Turkey 2021, Tech. rep., International Energy Agency. IMF (2022): World Economic Outlook. Countering the Cost-of-Living Crisis., International Monetary Fund, Washington, DC. 31 IPC (2021): Turkey’s decarbonization pathway net zero in 2050, Tech. rep., Istanbul Policy Center. Kalman, R. E. (1960): A New Approach to Linear Filtering and Prediction Problems, Trans- actions of the ASME–Journal of Basic Engineering, 82(Series D), pp. 35–45. Karlsson, M., Alfredsson, E., Westling, N. (2020): Climate policy co-benefits: a review, Climate Policy, 20(3), pp. 292–316. Kiyotaki, N., Moore, J. (1997): Credit Cycles, Journal of Political Economy, 105(2), pp. 211–48. Kydland, F. E., Prescott, E. C. (1977): Rules Rather than Discretion: The Inconsistency of Optimal Plans, The Journal of Political Economy, 85(3), pp. pp. 473–492. Lenzen, M., Geschke, A., Abd Rahman, M. D., Xiao, Y., Fry, J., Reyes, R., Dietzenbacher, E., Inomata, S., Kanemoto, K., Los, B., et al. (2017): The Global MRIO Lab–charting the world economy, Economic Systems Research, 29(2), pp. 158–186. a i Canals, L., Pi˜ Lenzen, M., Geschke, A., West, J., Fry, J., Malik, A., Giljum, S., Mil` nero, P., Lutter, S., Wiedmann, T., et al. (2022): Implementing the material footprint to measure progress towards Sustainable Development Goals 8 and 12, Nature Sustainability, 5(2), pp. 157–166. McKibbin, W. J., Wilcoxen, P. J. (1998): The theoretical and empirical structure of the G-Cubed model, Economic Modelling, 16(1), pp. 123–148. Mercure, J.-F. (2012): FTT:Power : A global model of the power sector with induced tech- nological change and natural resource depletion, Energy Policy, 48, pp. 799–811, special Section: Frontiers of Sustainability. Michaillat, P., Saez, E. (2021): Resolving New Keynesian Anomalies with Wealth in the Utility Function, The Review of Economics and Statistics, 103(2), pp. 197–215. Millard, S. (2011): An estimated DSGE model of energy, costs and inflation in the United Kingdom, Bank of England working papers 432, Bank of England. Nemet, G. F., Jakob, M., Steckel, J. C., Edenhofer, O. (2017): Addressing policy credibility problems for low-carbon investment, Global Environmental Change, 42, pp. 47–57. Nordhaus, W. (2015): Climate Clubs: Overcoming Free-Riding in International Climate Policy, American Economic Review, 105(4), pp. 1339–70. OECD (2021): OECD Economic Surveys: Turkey 2021, Organisation for Economic Co- operation and Development. OECD (2022): OECD Economic Outlook, Volume 2022 Issue 2, Organisation for Economic Co-operation and Development. 32 Papageorgiou, C., Saam, M., Schulte, P. (2017): Substitution between clean and dirty energy inputs: A macroeconomic perspective, Review of Economics and Statistics, 99(2), pp. 281– 290. Parry, I., Veung, C., Heine, D. (2015): How Much Carbon Pricing is in Countries’ Own In- terests? The Critical Role of Co-Benefits, Climate Change Economics, 06(04), p. 1550019. Ploeg, R. v. d., Venables, A. (2022): Radical Climate Policies, CEPR Press Discussion Paper No. 17677, CEPR. Ranger, N., Volz, U. (2023): Climate Change, Fiscal risks, and Debt Sustainability., mTI World Bank. Rotemberg, J. J. (1982): Sticky Prices in the United States, Journal of Political Economy, 90(6), pp. 1187–1211. Rubin, E. S., Azevedo, I. M., Jaramillo, P., Yeh, S. (2015): A review of learning rates for electricity supply technologies, Energy Policy, 86, pp. 198–218. Sager, L. (2019): Income inequality and carbon consumption: Evidence from Environmental Engel curves, Energy Economics, 84(S1), p. S0140988319302968. Schoder, C. (2017): Are Dynamic Stochastic Disequilibrium Models Keynesian or Neoclas- sical?, Structural Change and Economic Dynamics, 40(C), pp. 46–63. Schoder, C. (2020): A Keynesian Dynamic Stochastic Disequilibrium model for business cycle analysis, Economic Modelling, 86, pp. 117 – 132. Schoder, C., Tercioglu, R. B. (2023): Revisiting Macroeconomic Fluctuations in the Euro Area: An Estimated Dynamic Stochastic Disequilibrium Model., university of Bamberg. Smets, F., Wouters, R. (2003): An Estimated Dynamic Stochastic General Equilibrium Model of the Euro Area, Journal of the European Economic Association, 1(5), pp. 1123– 1175. Smets, F., Wouters, R. (2007): Shocks and Frictions in US Business Cycles: A Bayesian DSGE Approach, The American Economic Review, 97(3), pp. pp. 586–606. Taylor, J. B. (1980): Aggregate Dynamics and Staggered Contracts, The Journal of Political Economy, 88(1), pp. 1–23. Taylor, J. B. (2016): The Staying Power of Staggered Wage and Price Setting Models in Macroeconomics, NBER Working Paper Series, 22356. TurkStat (2022): Data Portal for Statistics, data retrieved from, https://data.tuik.gov. tr/Kategori/GetKategori?p=enflasyon-ve-fiyat-106&dil=2. 33 van der Ploeg, F., Rezai, A. (2021): Optimal carbon pricing in general equilibrium: Tem- perature caps and stranded assets in an extended annual DSGE model, Journal of Envi- ronmental Economics and Management, 110, p. 102522. Van der Werf, E. (2008): Production functions for climate policy modeling: An empirical analysis, Energy economics, 30(6), pp. 2964–2979. Varga, J., Roeger, W., et al. (2022): E-QUEST: A multisector dynamic general equilib- rium model with energy and a model-based assessment to reach the EU climate targets, Economic Modelling, p. 105911. Vitek, F., Gaspar, V., Adrian, T. (2022): A Medium-Scale DSGE Model for the Integrated Policy Framework, IMF Working Papers 2022/015, International Monetary Fund. Voyvoda, E., Yeldan, A. E. (2015): Low Carbon Development Pathways and Priorities for Turkey, Tech. rep., Istanbul Policy Center. Way, R., Ives, M. C., Mealy, P., Farmer, J. D. (2022): Empirically grounded technology forecasts and the energy transition, Joule, 6(9), pp. 2057–2082. urkiye Country Climate and Development Report, CCDR Series, World Bank (2022): T¨ World Bank Group. World Bank (2023): World Development Indicators, data retrieved from World Development Indicators, https://databank.worldbank.org/source/ world-development-indicators. 34 A An open-economy energy macro model This appendix provides the details of the open-economy macro framework used for the policy analysis in the present paper. For a description of the variables and parameters, see Tables B2 to B5 in Appendix B. Since the model includes a trend arising from a growing labor- embodied productivity zt , we stationarize any trending real variable Xt by X ˜ t = Xt /zt . Financial assets are nominal variables and, therefore, additionally exhibit a price trend. We therefore divide them by the aggregate price of the final good: X ˆ t = Xt /(zt PY,t ). Note that international bonds denominated in foreign currency are detrended by the foreign price level. For the nominal price of a good or asset X we use the following notation: pX,t = PX,t /PY,t . A.1 The household sector The model features two types of households: high-skilled labor H-households and income- constrained low-skilled labor L-households. A.1.1 High-skilled labor households (H-households) High-skilled labor aggregator. A perfectly competitive high-skilled labor aggregator purchases differentiated labor varieties from the high-skilled labor households. The labor i aggregator’s demand for the high-skilled labor variety NH,t can be obtained from the following cost minimization problem: 1 i i min i PLH,t NH,t di NH,t 0 subject to 1 1 1+εWH,t µH i 1+εWH,t µH LH,t = NH,t di , 0 where LH,t is total homogeneous high-skilled labor measured in hours and 1 + 1/εWH,t µH > 1 is the elasticity of substitution between different high-skilled labor varieties. Noting that the Lagrangian multiplier of the constraint is equal to the aggregate high-skilled wage index, PLH,t , one can show the first-order condition (FOC) to read 1 i − 1+ ε i PLH,t WH,t µH NH,t = LH,t . PLH,t Differentiated high-skilled labor households. A continuum of high-skilled labor house- holds (H-households) exclusively owns the firms and supplies differentiated high-skilled labor 35 services to the labor aggregator. The problem of an H-household i is to optimize inter- temporal utility. That is,  1−ϱ  ϱ (CH,t −κCH,t−1 ) i ∞ εCH,t ψC zt max E0 βt  i 1+ηH NH,t 1−ϱ  i i i i i i i i ∞ {CH,t ,BG,t ,BP,t ,BW,t ,SZ,t ,SR,t ,SF,t ,PLH,t }t=0 Ai t=0 −εNH,t ψH zt 1+ηH + ψA PY,t t subject to the demand schedule for its labor variety 1 i − 1+ ε i PLH,t WH,t µH NH,t = NH,t PLH,t and subject to the budget constraint i Ait (1 + tC )CH,t + (1 + tA ) + PY,t i i ∗ i PLH,t i PLH,t St PO,t +TR,t + tH NH,t + Ψi W,t = N i + (1 − αO ) Ot PY,t PY,t H,t PY,t RB,t−1 i i RW,t−1 i + BG,t −1 + BP,t−1 + (1 − ΓW,t−1 ) St BW,t−1 PY,t PY,t PSZ,t i + (1 − ΓS εΓS,t−1 ) + DZ,t SZ,t −1 PY,t PSR,t i + (1 − ΓS εΓS,t−1 ) + DR,t SR,t −1 PY,t PSF,t i + (1 − ΓS εΓS,t−1 ) + DF,t SF,t −1 + DE,t + Ξt PY,t where 2 i τW PLH,t Ψi W,t = i − ΠH zt 2 PLH,t−1 Ai i i i i i i t = BG,t + BP,t + St BW,t + PSZ,t SZ,t + PSR,t SR,t + PSF,t SF,t  RB,t−1 −1 i i RW,t−1 −1 i  PY,t BG,t −1 + BP,t−1 + (1 − ΓW,t−1 ) PY,t St BW,t−1 +    PSZ,t P −1 i PSR,t PSR,t−1 i i TR,t = tR  PY,t + DZ,t − SZ,t PY,t SZ,t−1 + PY,t + DR,t − PY,t SR,t −1 +  +(1 − ΓS εΓS,t−1 )   P SF,t PSF,t−1 i + PY,t + DF,t − PY,t SF,t−1 + DE,t γ 1 St BˆW,t ΓW,t = γW εΓW,t W exp −1 ˜ t PY,t GDP 36 RW,t−1 − tR (RW,t−1 − 1) Ξt = ΨW,t + ΓW,t−1 St BW,t−1 + PY,t PSZ,t PSR,t PSF,t + ΓS εΓS,t−1 + DZ,t + + DR,t + + DF,t PY,t PY,t PY,t and where zt = gt zt−1 is labor-embodied productivity growing at rate gt as specified below. Note that all quantities are in real terms except assets which are nominal. The third term in the utility function states that wealth yields utility (cf. Schoder 2020). Following Christoffel et al. (2008), we introduce financial intermediation premia which depend on the riskiness of the respective asset and are rebated to the households in a lump-sum manner. This introduces wedges between risky asset returns and the risk-free domestic bond return. As in Christoffel et al. (2008), the external financial intermediation premium increases with the ratio of foreign bonds valued in domestic currency to GDP. After aggregation over all H-households, the FOCs imply that −ϱ ˜H,t−1 /gt ˜H,t − κC (1 + tC )λH,t = εCH,t ψC C (1) RB,t − tR (RB,t − 1) (1 + tA )λH,t − ψA = β Et λH,t+1 (2) ΠY,t+1 RB,t − tR (RB,t − 1) = (1 − ΓW,t ) (RW,t − tR (RW,t − 1)) ∆S,t+1 (3) RB,t − tR (RB,t − 1) = (1 − ΓS,t ) (RS,t − tR (RS,t − 1)) (4) εNH,t ψH L1+ ηH H,t (1 − tH )˜ pLH,t LH,t = (1 + εWH,t µH ) (5) λH,t λH,t+1 − τW εWH,t µH (ΠH,t − ΠH ) ΠH,t − β Et (ΠH,t+1 − ΠH ) ΠH,t+1 λH,t ˜H,t + A C ˜CH,t + ˆt + T ˜H,t + T ˜A,t + T +T ˜R,t = p ˜LH,t LH,t + (1 − αO ) Et p∗ ˜ O,t Ot (6) RB,t−1 ˆ ˆP,t−1 /gt + RW,t−1 Et B ˆW,t−1 /gt + BG,t−1 + B (7) ΠY,t Π∗ Y,t RS,t−1 ˜ E,t + pSZ,t−1 + p (˜ ˜SR,t−1 + p ˜SF,t−1 ) /gt + D ΠY,t where Et ΠY,t ∆S,t = (8) Et−1 Π∗ Y,t 37 is the rate of depreciation of the nominal exchange with the real exchange rate defined as ∗ Et = St PY,t /PY,t , γ 1 ˆW,t Et B ΓW,t = γW εΓW,t W exp −1 (9) GDP˜ t is the external risk premium, ΓS,t = ΓS εΓS,t (10) is the equity risk premium, p ˜SZ,t = Et Λt,t+1 gt+1 p ˜ Z,t+1 ˜SZ,t+1 + D (11) p ˜SR,t = Et Λt,t+1 gt+1 p ˜ R,t+1 ˜SR,t+1 + D (12) p ˜SF,t = Et Λt,t+1 gt+1 p ˜ F,t+1 ˜SF,t+1 + D (13) are the respective sectors’ expected discounted streams of future dividends, −1 RS,t−1 Λt−1,t = (14) ΠY,t ˜A,t , T ˜CH,t , T is the stochastic discount factor, and T ˜R,t are tax revenues as specified ˜H,t , and T below. Eq. (1) states that the utility from relaxing the budget constraint by one unit is equal to the marginal utility of the consumption which this unit provides. Eq. (2) is a variant of the Euler equation and determines how the households smooth consumption optimally. It relates the marginal consumption utility of a unit of income in t to the utility this unit would yield if saved and consumed in t + 1. Eq. (3) is the risk-adjusted uncovered interest parity condition and it pins down the change of the exchange rate which adjusts to eliminate any arbitrage opportunity making the household indifferent between holding international debt or any other asset. Eq. (4) states that equity returns exceed the risk free bond rate by an equity risk premium, ΓS εΓS,t RS,t (abstracting from capital taxes). Note that RSZ,t = RSR,t = RSF,t ≡ RS,t and note that the equity return rate is defined as PSx,t + Dx,t PY,t RSx,t−1 ≡ for x ∈ {Z, R, F } PSx,t−1 Eq. (5) states that the rate of high-skilled wage inflation should be such that the marginal disutility from increasing labor supply by one unit should be equal to the marginal utility which the additional consumption provided from that unit of labor generates. The mark- up, εWH,t µH , arising from monopoly power drives a wedge between the marginal rate of substitution between differentiated labor varieties and the high-skilled real wage. Eq. (6) is simply the budget constraint. Note that we normalize the total number of equity shares to one. Define pS,t = PS,t /PY,t as the price of an equity share expressed in the price of the final good. 38 A.1.2 Low-skilled labor households (L-households) There is a continuum of low-skilled labor L-households each of which supplies labor hours and receives wage income. L-households have no access to the financial market and are therefore hand-to-mouth consumers. The low-skilled labor household’s utility maximization problem has a corner solution w.r.t. consumption. The L-household’s problem is to i 1−ϱ i L1+η ϱ CL,t − κCL,t−1 NL,t max εCL,t ψC zt − εNL,t ψL zt i ,N i CL,t L,t 1−ϱ 1 + ηL i i i s.t. (1 + tC )CL,t + tL pLL,t NL,t = pLL,t NL,t + TLL,t The FOCs imply −ϱ ˜L,t−1 /gt ˜L,t − κC (1 + tC )λL,t = εCL,t ψC C (15) ηL εNL,t ψL NL,t = (1 − tL )λL,t p ˜LL,t (16) ˜CL,t + T ˜L,t + T C ˜L,t = p ˜LL,t ˜LL,t LL,t + T (17) Eq. (15) relates marginal utility to the shadow price of relaxing the budget constraint. Eq. (16) is the labor supply function. In the optimum, the disutility of supplying one additional unit of labor has to equal the utility it generates through consuming the additional labor income. Eq. (17) is the aggregated budget constraint. Note that only a fraction of low-skilled labor households finds employment. For the sake of simplicity, we assume unemployment only on the extensive margin and not on the intensive margin. A.2 CES bundlers A.2.1 Retail firms (Y-firms) Retailers or Y-firms combine wholesale goods Wt and energy goods EY,t and, using a Constant- Elasticity-of-Substitution (CES) bundler, form the final retail good Yt which is used for consumption and capital investment. Their cost minimization problem reads min PW,t Wt + PE,t EY,t Wt ,EY,t subject to a CES production structure, σY 1 σY −1 σY −1 σY −1 1 σY σY Yt = εY,t αY Wt + (1 − αY ) σY EY,t σY . 39 The FOCs of this problem imply ˜ t = εσY −1 αY pW,t −σY Y W ˜t (18) Y,t ˜Y,t = εσY −1 (1 − αY ) pE,t −σY Y E ˜t (19) Y,t ˜ t + pE,t E ˜t = pW,t W Y ˜Y,t (20) where pW,t ≡ PW,t /PY,t , pE,t ≡ PE,t /PY,t and εY,t is the productivity shock. Eqs. (18) and (19) are the Y-firms’ demand functions for wholesale goods W ˜Y,t . Eq. ˜ t and energy goods E (20) is the Y-firm’s budget constraint. A.2.2 Wholesale firms (W-firms) Wholesale firms produce a wholesale good Wt which they sell to the retailers. As inputs, they ∗ combine domestic and foreign core goods, Zd,t and Ze,t and they are subject to input specific productivity shocks εWD,t and εWI,t . Note that the asterisk indicates that the variable is interpreted from the view of the Rest of the World. Competition is assumed to be perfect. ∗ A unit of the domestic core good costs PZd,t , while a unit of the foreign core good is St PZe,t . St is the nominal exchange rate and converts foreign prices into domestic prices. Profit maximization implies the following input demand functions: −σW ˜d,t = εσW −1 αW pZ,t ˜t Z WD,t W (21) pW,t −σW W −1 Et p∗ Z,t ˜e,t ∗ = εσ ˜t Z WI,t (1 − αW ) W (22) pW,t ˜d,t + Et p∗ Z ˜ t = pZ,t Z pW,t W ˜∗ (23) Z,t e,t where pZ,t ≡ PZd,t /PY,t and p∗ ∗ ∗ ∗ ∗ Z,t ≡ PZd,t /PY,t = PZe,t /PY,t where the last equality follows from the law of one price in the absence of trade transaction costs. The real exchange ∗ rate is Et ≡ St PY,t /PY,t . Eq. (21) is the domestic demand function for domestic core ˜ goods Zd,t which together with the foreign demand for domestic core goods Z ˜e,t (exports) to be specified below constitute the total domestically produced core goods. Eq. (22) is the domestic demand function for foreign core goods Z ˜e,t ∗ which constitute the only imports except the carbon imports. Eq. (23) is the W-firm’s budget constraint. 40 A.2.3 Labor firms (L-firms) Labor firms combine high-skilled and low-skilled labor hours into a homogeneous labor service which they sell to the core good firms, renewable energy firms, and fossil energy firms. They use a CES technology with input specific productivity shocks εLH,t and εLL,t . Profit maximization implies the following aggregated FOCs: −σL σL −1 ˜LH,t p LH,t = εLH,t αL (LZ,t + LR,t + LF,t ) (24) p˜L,t −σL L −1 ˜LL,t p LL,t = εσ LL,t (1 − αL ) (LZ,t + LR,t + LF,t ) (25) p˜L,t ˜L,t (LZ,t + LR,t + LF,t ) = p p ˜LL,t LL,t ˜LH,t LH,t + p (26) Eqs. (24) and (25) are standard input demand functions for high-skilled and low-skilled labor hours for a given unit of labor services. Eq. (26) is the labor firm’s budget constraint. The growth rate of the real wage is linked to wage and price inflation according to p˜LH,t gt = ΠH,t /ΠY,t (27) ˜LH,t−1 p p˜LL,t gt = ΠL,t /ΠY,t (28) ˜LL,t−1 p A.2.4 Energy firms (E-firms) Using a CES technology, E-firms combine renewable energy ER,t and fossil energy EF,t to produce a homogeneous energy composite EY,t + EZ,t which they sell to the core good firms and the retail firms. Energy firms are subject to a productivity shock εE,t . Similar to Guerrieri et al. (2008), we introduce renewable input adjustment costs ΨE,t to get a time varying elasticity of substitution, which is low in the short-run and converges to σE in the long-run. σE 1 σE −1 1 σE − 1 σE − 1 ˜Z,t = εE,t α (ΨE,t ER,t ) ˜Y,t + E E σE σE + (1 − αE ) σE EF,t σE . E 2 τE ER,t ΨE,t =1− −1 . 2 ER,t−1 41 Competition is assumed to be perfect. To account for the stickiness of both price and quantity contracts in the energy market, we allow for windfall profits DE,t which are dis- tributed to high-skilled households. These windfall profits are random and zero at the steady state. They help the model reconcile the consumer price index for energy, the price indices for fossil and renewable energy, as well as the respective quantities, all of which are observed. Energy firms receive a subsidy sER per unit of renewable energy input and pay a fuel tax tEF per unit of fossil energy. Profit maximization implies the following aggregated FOCs: σE pER,t − sER E −1 ˜Z,t αE (ΨE,t ER,t )−1 ˜Y,t + E ∂ ΨE,t = εσ E,t E ER,t + ΨE,t (29) pE,t ∂ER,t σE ˜Z,t ˜Y,t + E E pEF,t + tEF E −1 = εσ E,t (1 − αE ) (30) pE,t EF,t ∂ ΨE,t ER,t 1 = −τE −1 (31) ∂ER,t ER,t−1 ER,t−1 ˜Z,t = (pER,t − sER ) E ˜Y,t + E pE,t E ˜F,t + D ˜R,t + (pEF,t + tEF ) E ˜ E,t . (32) Eqs. (29) and (30) are input demand functions for renewable and fossil energy. Eq. (32) is the energy firm’s budget constraint. A.3 The value-added firm sector A.3.1 Core goods firms (Z-firms) Core good aggregator. A perfectly competitive unity continuum of core goods firms aggregates differentiated goods varieties from the intermediate goods firms. Taking as given j the prices PZ,t and PZ,t , the representative firm chooses the optimal selection of intermediate goods varieties in order to minimize costs. The core good aggregator’s demand for the intermediate good Ztj supplied by an intermediate good firm j of a continuum of mass one can be obtained from the following cost minimization problem: 1 j min j PZ,t Ztj dj Zt 0 subject to 1 1 1+εPZ,t µZ Zd,t + Ze,t = Ztj 1+εPZ,t µZ dj , 0 42 where Zd,t + Ze,t is total core output sold domestically and exported to the Rest of the World. Further, 1 + 1/εPZ,t µZ > 1 is the elasticity of substitution. Noting that the Lagrangian multiplier of the constraint is equal to the aggregate price index, PZ,t , one can show the FOC to read 1 j − 1+ ε PZ,t µZ PZ,t Ztj = (Zd,t + Ze,t ) . PZ,t Note that the law of one price requires that the domestically produced core good has the same price domestically and abroad. Differentiated core goods firms. Differentiated core goods firms purchase labor hours from L-firms and energy from E-firms. Using these inputs as well as private and public capital, a Z-firm j produces a differentiated variety of the intermediate core good Ztj and sells it to the core goods aggregators. Differentiated Z-firms operate in monopolistic markets and, hence, have price setting power. We assume that the capital stock is firm-specific and cannot be traded on a spot market. They face various frictions such as price adjustment costs, capital utilization costs, and investment adjustment costs. They pay corporate income taxes as well as payroll taxes. Labor productivity growth is stochastic and follows a deterministic long-run trend, g ρ 1−ρg gt = gt− 1g εg,t . (33) Taking as given total demand for core goods Zd,t + Ze,t , the final goods price level PY,t , the core price level PZ,t , the real price for energy pE,t , and the real wage rate of a homogeneous labor hour pL,t as well as the law of motion of capital, adjustment costs, the production functions, the demand function for intermediate goods, and the requirement to maintain a j j debt-capital ratio λ, the firm j chooses {BZ,t , PZ,t , Lj j j j j ∞ Z,t , IPZ,t , uZ,t , KPZ,t , EZ,t }t=0 to maximize j j PSZ,t its real beginning-of-period value, DZ,t + Sj PY,t Z,t−1 . Evaluating at period t = 0, the optimization problem reads ∞ t s=0 Λs−1,s j max E0 DZ,t j {BZ,t j ,PZ,t ,Lj j j j j ∞ Z,t ,IPZ,t ,uZ,t ,KPZ,t ,EZ,t }t=0 t=0 Λ−1,0 where j j PZ,t DZ,t = (1 − tP ) Z j − (1 + tS )pL,t Lj j j j Z,t − pE,t EZ,t − ΨP,t + ΨP,t − ΨZ,t + ΨZ,t + PY,t t   j 2 τI I PZ,t RB,t−1 j + tP Qj j δKPZ,t −1 + εK,t − 1 IPZ,t j j  − IPZ,t j − BZ,t + B Z,t j 2 gt IPZ,t−1 ΠY,t Z,t−1 43 subject to j 2 τP PZ,t Ψj P,t = j − ΠY (Zd,t + Ze,t ) 2 PZ,t−1 τZ2 j 2 Ψj j Z,t = τZ1 uZ,t − 1 + uZ,t − 1 KZ,t−1 2   j 2 j τI IPZ,t j j KPZ,t = εK,t 1 − j −1  IPZ,t + (1 − δ )KPZ,t −1 2 gt IPZ,t −1 σZ 1 σZ −1 1 σZ −1 σZ − 1 Ztj j j σZ = αZ VZ,t σZ + (1 − αZ ) σZ εZE,t EZ,t σZ αVZ j VZ,t = εVZ,t uj j Z,t KZ,t−1 (zt LZ,t )(1−αVZ ) j j αKZ KZ,t = εKZ,t KPZ,t KGZ,t (1−αKZ ) 1 j − 1+ ε PZ,t µZ PZ,t Ztj = (Zd,t + Ze,t ) PZ,t j RB,t BZ,t ≤ λEt ΠY,t+1 Qj j Z,t+1 (1 − δ )KPZ,t , where zt = gt zt−1 and −1 RS,t+k−1 Et Λt,t+k = ΠY,t+k is the stochastic discount factor. Note that price and capital adjustment costs are refunded to the core firm sector in a lump-sum manner. Hence, adjustment costs affect economic behavior but not the aggregate flow of funds. 44 The FOCs imply 1 1 µZ,t = − (34) RB,t RS,t 1 φZ,t (1 + εPZ,t µZ ) − pZ,t  pZ,t ΠY,t  Π pZ,t−1 Y,t − ΠY pZ,t−1 −(1 − tP )τP εP,t µZ  pZ,t+1 ˜d,t+1 +Z (Z ˜e,t+1 )gt+1  = 1 − tP 1 pZ,t+1 −Et Λt,t+1 ΠY,t+1 − ΠY Π Y,t+1 ˜d,t +Z˜e,t ) pZ,t pZ,t pZ,t (Z (35) 1 1 σZ ˜e,t ˜d,t + Z Z σZ ˜Z,t V (1 − tP )(1 + tS )˜ pL,t = φZ,t αZ (1 − αVZ ) (36) V˜j LZ,t Z,t   2 τI ˜PZ,t I ˜PZ,t I ˜ IPZ,t  1 = QZ,t εK,t − (1 − tP )QZ,t εK,t  −1 + τI −1 (37) 2 ˜PZ,t−1 I ˜PZ,t−1 I ˜PZ,t−1 I 2 ˜PZ,t+1 I ˜PZ,t+1 I + Et Λt,t+1 (1 − tP )QZ,t+1 εK,t+1 τI −1 gt+1 ˜PZ,t I ˜PZ,t I 1 1 ˜d,t + Z Z ˜e,t σZ ˜Z,t V ˜ Z,t−1 /gt = φZ,t α αVZ (1 − tP ) (τZ1 + τZ2 (uZ,t − 1)) K σZ Z ˜Z,t V uZ,t (38)  1  1 ˜e,t+1 ˜d,t+1 +Z Z σZ V˜Z,t+1 K˜ Z,t αVZ uj σZ QZ,t = Et Λt,t+1 φZ,t+1 αZ ˜Z,t+1 V Z,t+1 ˜ Z,t /gt K αKZ ˜ PZ,t K  (39) +(1 − (1 − tP )δ )QZ,t+1 1 1 + λEt ΠY,t+1 (1 − δ )QZ,t+1 − RB,t RS,t 1 1 σZ − 1 σZ ˜d,t + Z Z ˜e,t σZ (1 − tP ) pE,t = φZ,t (1 − αZ ) σZ εEZ,t (40) E˜Z,t where   2 ˜PZ,t ˜ PZ,t = εK,t 1 − τI K I −1 ˜ PZ,t−1 /gt ˜PZ,t + (1 − δ )K I (41) 2 ˜PZ,t−1 I 45 σZ 1 σZ −1 σZ −1 σZ −1 1 σZ ˜e,t = ˜d,t + Z Z ˜ σZ αZ V σZ + (1 − αZ ) σZ ˜j εZE,t E (42) Z,t Z,t αVZ ˜Z,t = εVZ,t uZ,t K ˜ Z,t−1 /gt (1−αVZ ) V LZ,t (43) ˜ Z,t = εKZ,t K K ˜ (1−αKZ ) ˜ αKZ K (44) PZ,t GZ,t ˜ Z,t = pZ,t Z D ˜e,t − p ˜d,t + Z ˜PZ,t − T ˜Z,t − T ˜L,t LZ,t − pE,t E ˜SZ,t − (45) ˆZ,t − RB,t−1 B ˜PZ,t + B −I ˆZ,t−1 /gt ΠY,t Eq. (34) relates the shadow price of one additional unit of borrowing µZ,t to the equity risk premium. The equation collapses to µj Z,t = 0 if the returns on equity and borrowing are equal, RS,t = RB,t . In this special case, the financial structure of the firm is irrelevant from the household’s perspective. With a positive equity risk premium, however, RS,t > RB,t and µj Z,t > 0. External finance is preferred over internal finance. This is because households discount risky future dividend payouts by more than the firm’s financial obligations from borrowing increase over time. They rather have an additional unit of dividend today at the expense of a lower cash flow tomorrow, then forego the unit of dividend today and enjoy a higher cash flow tomorrow.20 Hence, households ask firms to distribute as much as possible today exploiting all available credit. The end-of-period borrowing constraints are always binding. The demand for end-of-period loans satisfies ˜ PZ,t ˆZ,t = λEt ΠY,t+1 QZ,t+1 (1 − δ )K RB,t B (46) Eq. (35) states that the mark-up εPZ,t µZ arising from monopoly power and the price ad- justment costs τP drive a wedge between the real marginal revenue (1 − tP )pZ,t and the real marginal cost φZ,t . Eq. (36) is the labor demand function and equates the marginal cost of labor and its marginal revenue product. Eq. (37) translates a change in the shadow price of capital QZ,t into a change in investment. Note that without capital adjustment costs QZ,t = 1, i.e. investment will adjust instantly to realize the desired capital stock. Eq. (38) determines the optimal rate of capital utilization. Eq. (39) links QZ,t to the desired capital stock. Without adjustment costs, the optimal capital stock ensures the marginal return of a dollar spent on capital is equal to its cost which is depreciation. Note that the marginal return includes the savings from labor costs. Also note that firm’s debt-capital ratio λ en- ters into desired capital stock decisions. Eq. (40) is the energy demand function and states 20 To increase the expected cash flow by one unit in t + 1, the firm has has to retain RB,t /Et ΠY,t+1 units in t. Yet, by the definition of the stochastic discount factor, households are only willing to give up Et Λt,t+1 units for a one unit cash flow in t + 1. 46 that energy input will equate its marginal revenue which includes savings in labor costs and taxes to its marginal costs which is the price. Eq. (41) is the aggregated law of motion of private capital. Eqs. (42) to (44) are the aggregated CES technologies. Eq. (45) is simply the rearranged firm’s budget constraint. Note that T ˜S,t are the Z-firms’ corporate ˜PZ,t and T income and payroll taxes, respectively. A.3.2 Renewable energy firms (R-firms) R-firms are modeled equivalently to Z-firms except that they don’t have market power, don’t have variable capital utilization, and don’t use energy input to production. Moreover, public capital can either take the form of generation capital and infrastructure. In the former case, public capital simply adds to private capital as a perfect substitute. As infrastructure, public capital constitutes an externality and enhances the marginal product of private capital. Yet, there is an exogenous satiation level of how much infrastructure can be used in production. Also, the productivity shock process in renewable energy production VER,t features learning by doing. In the literature, learning-by-doing elasticity is usually estimated using costs and capacity of production. Here, we relate productivity to energy output. We assume a one-to- one relationship between increase in productivity and decline in levelized cost of renewable energy, which we confirm by stochastic simulations. Taking as given price of its output PER,t , the final goods price level PY,t , and the real wage rate of a homogeneous labor service pL,t as well as the law of motion of capital, capital adjustment costs, the production functions, and the requirement to maintain a debt-capital j j ratio λ, the firm j chooses {BR,t , ER,t , Lj j j ∞ R,t , IPR,t , KPR,t }t=0 to maximize its real beginning- j j PSR,t of-period value, DR,t + Sj PY,t R,t−1 . Evaluating at period t = 0, the optimization problem reads ∞ t s=0 Λs−1,s j max E0 DR,t j {BR,t j ,ER,t j j j ,LR,t ,IPR,t ,KPR,t }∞ t=0 t=0 Λ−1,0 where j DR,t = (1 − tP ) pj j j ER,t ER,t − (1 + tS )pL,t LR,t +   j 2 j  j τI IPR,t j j j RB,t−1 j + tP QR,t δKPR,t−1 + εK,t −1 IPR,t  − IPR,t − BR,t + B j 2 gt IPR,t −1 ΠY,t R,t−1 subject to   j 2 j τI IPR,t j j KPR,t = εK,t 1 − j −1  IPR,t + (1 − δ )KPR,t−1 2 gt IPR,t−1 47 σVR 1 σVR −1 1 σVR −1 σVR −1 j j zt Lj σVR ER,t = VER,t αVR KR,t −1 σVR + (1 − αVR ) σVR R,t σVR ϕER (1−ρER ) ρER ER,t VER,t = VER,t−1 ϵER,t ER j j αKR used (1−αKR ) KR,t = KPR,t + KGRgen,t KGRinf,t used max KGRinf,t = min KGRinf , KGRinf,t j RB,t BR,t ≤ λEt ΠY,t+1 Qj j R,t+1 (1 − δ )KPR,t The FOCs imply 1 1 µR,t = − (47) RB,t RS,t 1 1 σVR −1 σVR ˜R,t E σVR pL,t = pER,t (1 − αVR ) (1 + tS )˜ σVR VER,t (48) LR,t   2 τI ˜PR,t I ˜PR,t I ˜ IPR,t  1 = QR,t εK,t − (1 − tP )QR,t εK,t  −1 + τI −1 2 ˜PR,t−1 I ˜PR,t−1 I ˜PR,t−1 I (49) 2 ˜PR,t+1 I ˜PR,t+1 I + Et Λt,t+1 (1 − tP )QR,t+1 εK,t+1 τI −1 gt+1 ˜PR,t I ˜PR,t I  σVR −1 1  1 σVR σVR E˜R,t+1 σVR K˜ R,t QR,t = Et Λt,t+1 (1 − tP ) pER,t+1 αVR εER,t+1 ˜ R,t /gt+1 K αKR ˜ PR,t K  (50) +(1 − (1 − tP )δ )QR,t+1 1 1 + λEt ΠY,t+1 (1 − δ )QR,t+1 − (51) RB,t RS,t where   2 ˜PR,t ˜ PR,t = εK,t 1 − τI K I −1 ˜ PR,t−1 /gt ˜PR,t + (1 − δ )K I (52) 2 ˜PR,t−1 I 48 σVR 1 σVR −1 σVR −1 σVR −1 σVR 1 ˜R,t = VER,t E αVR σVR ˜ R,t−1 /gt K + (1 − αVR ) σVR σVR LR,t (53) αKR (1−αKR ) ˜ R,t = K K ˜ GRgen,t ˜ PR,t + K ˜ GRinf,t K used (54) ˜ GRinf,t K used ˜ GRinf = min K max ˜ GRinf,t ,K (55) ˜ R,t = pR,t E D ˜R,t − p ˜SR,t − I ˜PR,t − T ˜L,t LR,t − T ˆR,t − RB,t−1 B ˜PR,t + B ˆR,t−1 /gt (56) ΠY,t ˜ PR,t ˆR,t = λEt ΠY,t+1 QR,t+1 (1 − δ )K RB,t B (57) A.3.3 Fossil energy firms (F-firms) For the sake of simplicity, the model assumes that domestic carbon extractors (crude oil, coal and natural gas) produce for the international market only. They set the carbon fuel ∗ price according to the global price denominated in foreign currency PO,t even though the 21 resource is used domestically. As before, capital is firm-specific and subject to quadratic adjustment costs. We assume that fossil energy firms operate under perfect competition. The fossil energy production technology allows for substitution between capital and carbon when producing the carbon-capital composite. We assume quadratic adjustment costs for carbon input going into fossil fuel production proportional to past period’s aggregate carbon input. Taking as given the fossil energy price PEF,t , the overall price level PY,t as well as the law of motion of the F-firm’s capital stock, the fossil energy production function, adjust- ment costs, and the requirement to maintain a debt-capital ratio λ, the firm j chooses j j {BF,t , EF,t , Lj j j j ∞ j F,t , IPF,t , KPF,t , Ot }t=0 to maximize its real beginning-of-period value, DF,t + j PSF,t Sj PY,t F,t−1 . Evaluating at period t = 0, the optimization problem reads ∞ t s=0 Λs−1,s j max E0 DF,t j {BF,t j ,EF,t ,Lj j j j ∞ F,t ,IPF,t ,KPF,t ,Ot }t=0 t=0 Λ−1,0 where j DF,t = (1 − tP ) pj j j j ∗ EF,t EF,t − (1 + tS )pL,t LF,t − ΨO,t + ΨO,t − tO Ot − Et pO,t Ot   j 2 τI IPF,t RB,t−1 j + tP Qj j δKPF,t −1 + εK,t j − 1 IPF,t j  − IPF,t j − BF,t + B F,t j 2 gt IPF,t−1 ΠY,t F,t−1 21 Note that currently we only assume carbon imports but not carbon exports. 49 subject to 2 j τO Ot Ψj O,t = j −1 Ot−1 2 gt Ot−1   j 2 j τI IPF,t j j KPF,t = εK,t 1 − j −1  IPF,t + (1 − δ )KPF,t −1 2 gt IPF,t −1 σEF 1 σEF −1 σEF −1 σEF −1 j σEFj 1 j, σEF EF,t = αEF VF,t σEF + (1 − αEF ) σEF Ot σVF 1 σVF −1 1 σVF −1 σVF −1 j j zt Lj σVF VF,t = εEF,t αVF KF,t−1 σVF + (1 − αVF ) σVF F,t σVF j j αKF used (1−αKF ) KF,t = KPF,t + KGF gen,t KGF inf,t used max KGF inf,t = min KGF inf , KGF inf,t j RB,t BF,t ≤ λEt ΠY,t+1 Qj j F,t+1 (1 − δ )KPF,t , The FOCs imply 1 1 µj F,t = − (58) RB,t RS,t 1 1 1 σEF ˜F,t E σEF 1 σVF −1 σVF ˜F,t V σVF (1 + tS )˜ pL,t = pEF,t αEF (1 − αVF ) σVF εEF,t (59) ˜F,t V LF,t   2 τI ˜PF,t I ˜PF,t I ˜PF,t I 1 = QF,t εK,t − (1 − tP )QF,t εK,t  −1 + τI −1  (60) 2 ˜PF,t−1 I ˜PF,t−1 I ˜PF,t−1 I 2 ˜PF,t+1 I ˜PF,t+1 I + Et Λt,t+1 (1 − tP )QF,t+1 εK,t+1 τI −1 gt+1 ˜PF,t I ˜PF,t I 50  1 σVF −1 1  1 1 σEF ˜F,t+1 E σEF σVF σVF V˜F,t+1 σVF K˜ F,t QF,t = Et Λt,t+1 (1 − tP ) pEF,t+1 αEF ˜F,t+1 V αVF εF,t+1 ˜ KF,t /gt αKF ˜ KPF,t  +(1 − (1 − tP )δ )QF,t+1 (61) 1 1 + λEt ΠY,t+1 (1 − δ )QF,t+1 − RB,t RS,t ˜t O 1 tO + Et p∗ O,t + τO −1 − ˜ t−1 O g 1 ˜ t+1 O ˜ t+1 O 1 ˜F,t E σEF −Et Λt,t+1 τO −1 = pEF,t (1 − αEF ) σEF (62) ˜t O ˜t O O˜t where   2 ˜PF,t ˜ PF,t = εK,t 1 − τI K I −1 ˜ PF,t−1 /gt ˜PF,t + (1 − δ )K I (63) 2 ˜PF,t−1 I σEF 1 σEF −1 σEF −1 σEF −1 1 ˜F,t = E ˜ σEF αEF V σEF + (1 − αEF ) σEF ˜t O σEF (64) F,t σVF 1 σVF −1 σVF −1 1 σVF −1 σVF ˜F,t = εEF,t α V σVF ˜ F,t−1 /gt K + (1 − αVF ) σVF LF,t σVF (65) VF αKF (1−αKF ) ˜ PF,t + K ˜ F,t = K K ˜ GF gen,t ˜ GF K used inf,t (66) ˜ max , K ˜ used = min K K ˜ GF inf,t (67) GF inf,t GF inf ˜F,t − p ˜ F,t = pF,t E D ˜L,t LF,t − Et p∗ ˜ ˜ ˜ ˜ O,t Ot − TPF,t − TSF,t − TO,t − (68) −I˜PF,t + BˆF,t − RB,t−1 BˆF,t−1 /gt ΠY,t ˜ PF,t ˆF,t = λEt ΠY,t+1 QF,t+1 (1 − δ )K RB,t B (69) Eq. (62) determines the carbon input to fossil energy production. At the steady state, the marginal cost of an additional unit of carbon consisting of extraction costs plus carbon tax and adjustment costs (left hand side) is equal to the marginal revenue (right hand side). Carbon adjustment costs ensure that carbon inputs adjusts sluggishly to changes in carbon pricing. 51 A.4 Low-skilled wage, fiscal, and monetary policy The wages for unskilled labor are administered similar to the monetary policy rate. We assume that the productivity corrected nominal wage inflation depends on its lagged value, employment (Phillips curve) and lagged price inflation (wage indexation): ρW L ϕWE (1−ρW L ) ϕWP (1−ρW ) ΠL,t ΠL,t−1 EL,t ΠY,t−1 = εWL,t . (70) ΠL ΠL EL ΠY Since hours are more volatile than employment, we follow Smets and Wouters (2003) and Christoffel et al. (2008) and assume only a fraction ξ of firms can modify employment rela- tionships. The FOC implies β 1 (1−βξ)(1−ξ) EL,t EL,t+1 1+β EL,t−1 1+β LL,t /NL,t (1+β )ξ = (71) EL EL EL EL,t The government’s budget constraint is ˜GZ,t + I ˜G,t + I C ˜GRinf,t + I ˜GRgen,t + I ˜GF inf,t + I ˜GFgen,t + I ˜Gbb,t + (72) +T ˜ER,t + RB,t−1 B ˜LL,t + S ˜t + B ˆG,t−1 /gt = T ˆG,t ΠY,t where the sum of all taxes is ˜CH,t + T ˜t = T T ˜A,t + T ˜CL,t + T ˜H,t + T ˜R,t + T ˜L,t (73) +T ˜SR,t + T ˜SZ,t + T ˜SF,t + T ˜PR,t + T ˜PZ,t + T ˜EF,t + T ˜PF,t + T ˜O,t with ˜H,t ˜CH,t = tC C T (74) ˜L,t ˜CL,t = tC C T (75) ˆt ˜A,t = tA A T (76)   RB,t−1 −1 ˆG,t−1 + BˆP,t−1 /gt + RW,t∗ −1 −1 ˆW,t−1 /gt + B Et B ˜R,t = T tR,t  ΠY,t ΠY,t  (77) RS,t−1 −1 ˜ + ΠY,t (˜ pSZ,t−1 + p ˜SF,t−1 ) /gt + DE,t ˜SR,t−1 + p ˜H,t = tH p T ˜LH,t LH,t (78) 52 ˜L,t = tL p T ˜LL,t LL,t (79) ˜SZ,t = tS p T ˜L,t LZ,t (80) ˜SR,t = tS p T ˜L,t LR,t (81) ˜SF,t = tS p T ˜L,t LF,t (82)   ˜e,t − p ˜d,t + Z pZ,t Z ˜SZ,t − ˜L,t LZ,t − T ˜PZ,t = tP  T 2  (83) I˜PZ,t ˜ PZ,t−1 /gt − εK,t τI −δ K −1 ˜PZ,t I 2 ˜ IPZ,t−1 ˜R,t − p pER,t E ˜SR,t − ˜L,t LR,t − T ˜PR,t = tP T ˜PR,t I 2 (84) ˜ PR,t−1 /gt − εK,t τI −δ K −1 ˜PR,t I 2 ˜PR,t−1 I ˜F,t − p pEF,t E ˜SF,t − T ˜L,t LF,t − T ˜O,t − Et p∗ O ˜ O,t t ˜PF,t = tP T ˜PF,t 2 (85) −δ K ˜ PF,t−1 /gt − εK,t τI I −1 I ˜PF,t 2 ˜PF,t−1 I ˜F,t ˜EF,t = tEF E T (86) ˜t. ˜O,t = tO O T (87) Government consumption and investment depend on the output gap and are subject to an auto-regressive policy shock, −ϕCY ˜G,t C ˜ t GDP = εCG,t (88) C˜G ˜ GDP −ϕIY ˜GX,t I ˜ t GDP = εIX,t (89) I˜GX ˜ GDP for X ∈ {Z, Rgen, Rinf, F gen, F inf }. The public bad bank, purchasing stranded fossil investment goods, follows 53 ˜Gbb,t−1 + εIGbb,t ˜Gbb,t = I I (90) The laws of motion of the various public capital stocks normalized by trend growth are   2 ˜GX,t K˜ GX,t = εK,t 1 − τI I − 1 I ˜ GX,t−1 /gt ˜GX,t + (1 − δ )K (91) 2 I ˜GX,t−1 for X ∈ {Z, Rgen, Rinf, Fgen, Finf }. The law of motion of detrended bad bank capital stock is: ˜Gbb,t + (1 − δ )K ˜ Gbb,t = I K ˜ Gbb,t−1 /gt (92) The renewable energy subsidies are ˜R,t . ˜ER,t = sER E S (93) Monetary policy follows a Taylor rule which responds to core inflation which is the price inflation of the wholesale goods. It excludes energy input to the retail goods (but includes energy input to core good production) as well as imports (input to wholesale goods). ρR ϕRΠ (1−ρR ) ϕRY (1−ρR ) Rt Rt−1 pZ,t ΠY,t ˜ t GDP = εR,t (94) R R pZ,t−1 ΠY ˜ GDP As in Smets and Wouters (2007), a risk premium capturing flight to safety creates a wedge between the policy rate and the return on risk-free assets. Since a positive shock increases the required return on domestic assets and the cost of capital, it reduces current consumption and investment simultaneously and helps explaining the co-movement of consumption and investment. 1/Γ Rt = 1 − ΓB εΓB,tB RB,t (95) where the exponent of the risk premium shock ensures that the variance of the shock does not depend on the level of the risk premium ΓB . A.5 Rest of the World The budget constraint of the rest of the world is ˜e,t − B ˆW,t = p∗ ˜∗ ∗ ˜ RW,t−1 ˆ pZ,t /Et Z Z,t Ze,t + αO pO,t Ot − BW,t−1 /gt (96) Π∗ Y,t 54 where the second line uses the law of one price for domestically produced goods. Recall that we assume that there is only one international market for carbon with price p∗O,t . The final demand abroad for retail goods is assumed to be a first-order auto-regressive process ∗ Y ˜t∗ ˜∗ = Y ρY ˜ ∗1−ρ∗ −1 Y Y ε ∗ Y ,t (97) t Domestic exports are the domestic core goods bought abroad. We assume export adjustment costs to capture sluggish response of exports to relative price changes. Assuming a symmetric economic structure in the Rest of the World, foreign retailers will buy domestic core goods according to  Z˜e,t −σW ∗ 1 pZ,t τZ,e ˜e,t−1 Z −1 g∗ /Et ˜e,t = (1 − αW ∗ )  Z  ˜ ∗. Y (98) t p∗ Z,t Finally, the foreign price of core goods deflated by the foreign price level is assumed to evolve according to ∗ρp∗ ∗1−ρp∗ p∗ Z,t = pZ,t−1 pZ Z Z εp ∗ Z ,t (99) We assume the interest rate of the rest of the world follows a first-order auto-regressive process ρRW 1−ρRW RW,t = RW,t−1 RW εRW,t (100) We also assume the inflation rate of the rest of the world follows a first-order auto-regressive process ρΠ∗ 1−ρΠ∗ Π∗ ∗ Y,t = ΠY,t−1 Y Π∗ Y Y εΠ ∗ Y ,t (101) The international price of carbon is exogenous and evolves according to ∗ρp ∗1−ρpO p∗ O,t = pO,t−1 pO O εp ∗ O ,t (102) A.6 Market clearing Capital market clearing implies ˜R,t + B ˜Z,t + B B ˜P,t ˜F,t = B (103) To derive the macroeconomic balance condition, we need to sum over the aggregate budget p ˜ Z,t ˜SZ,t +D p ˜i ˜SR,t +D ˜SF,t +D p ˜i R,t F,t constraints of each sector. Note that RS,t−1 /ΠY,t = p˜SZ,t−1 /gt = p˜SR,t−1 /gt = p˜SF,t−1 /gt . The 55 summing over all budget constraints leads to the macroeconomic balance conditions which equates aggregate supply of and aggregate demand for final goods: ˜t = C Y ˜H,t + C ˜G,t + ˜L,t + C (104) +I ˜PR,t + I ˜PZ,t + I ˜PF,t + I ˜GRgen,t + I ˜GZ,t + I ˜GRinf,t + I ˜GF inf,t + I ˜GFgen,t + I ˜Gbb,t . Note that net exports are not part of this equation as there is no trade in final retail goods but only in core goods. Note further that the the macro balance condition does not have the interpretation of a resource constraint as in conventional DSGE models. This because resources are neither utilized fully nor up to a supply-side-constrained structural level. We define GDP as Gross Domestic Purchases (Yt ) plus net exports: ˜t + pZ,t Z ˜ t=Y GDP ˜e,t − Et (p∗ ˜∗ ∗ ˜ Z,t Ze,t + αO pO,t Ot ). (105) A.7 Structural shocks The model includes the following i.i.d. disturbances: εCH,t (High-skilled consumption preference shock) εCL,t (Low-skilled consumption preference shock) εNH,t (High-skilled labor preference shock) εNL,t (Low-skilled labor preference shock) εΓB,t (Flight-to-safety shock) εΓW,t (Foreign bond risk premium shock) εΓS,t (Equity risk premium shock) εg,t (Long-run productivity growth shock) εWH,t (High-skilled wage mark-up shock) εWL,t (Low-skilled wage inflation shock) εY,t (Y-firms (retail good) productivity shock) εWD,t (W-firms (wholesale good) domestic input productivity shock) εWI,t (W-firms (wholesale good) imported input productivity shock) εEZ,t (Z-firms (core good) energy efficiency shock) εVZ,t (Z-firms (core good) productivity shock) εLH,t (L-firms (labor) high-skilled input specific productivity shock) 56 εLL,t (L-firms (labor) low-skilled input specific productivity shock) εE,t (E-firms (energy good) productivity shock) εER,t (R-firms (renewable energy good) productivity shock) εEF,t (F-firms (fossil energy good) productivity shock) εPZ,t (Core good price mark-up shock) εK,t (Capital efficiency shock) εKZ,t (Core good capital bundler productivity shock) εCG,t (Government consumption shock) εIZ,t (Government investment shock in core good sector) εIGRinf,t (Government investment shock in renewable infrastructure) εIGRgen,t (Government investment shock in renewable energy generation) εIGF inf,t (Government investment shock in fossil energy infrastructure) εIGF gen,t (Government investment shock in fossil energy generation) εIGbb,t (Government investment shock in bad bank) εR,t (Monetary policy shock in the home country) εRW,t (Monetary policy shock in the RW) εW ∗ ,t (RW retail firms productivity shock) εp ∗ Z ,t (Import price shock) εY ∗ ,t (Demand shock in the RW) εΠ ∗ Y ,t (Inflation shock in the RW) εp ∗ O ,t (Global fossil fuel price shock) 57 B Model calibration and estimation This section reports the estimation and calibration of the model. In order for certain model variables to match their empirical counterparts at the steady state, some parameters have been restricted. All structural parameters have been estimated using Bayesian techniques. Here, we report first calibrated parameters and then estimated parameters with their 90% credibility intervals. B.1 Data We use time series data for 31 observed variables over 2000Q2-2022Q3, which are reported in Table B1.22 Twenty-six observables are available at the quarterly frequency.23 However, energy balances from the International Energy Agency (IEA) and CO2 emissions from the Climate Watch (CAIT) database maintained by the World Resources Institute are annual. We use mixed frequency data in the estimations. The missing quarterly observations of annual (and quarterly) data are obtained by the Kalman (1960) filter. This feature of our estimation technique is highly relevant for emerging market economies where data are available but generally with gaps or at lower frequencies than what the models are set at. We discuss the details of how we deal with missing observations below. We detrend policy rate, 3-month interest rate, inflation rate, low-skilled wage inflation rate, high-skilled wage inflation rate, real exchange rate, employment rate, RoW interest rate and RoW inflation rate by the double-sided Hodrick and Prescott (1980) filter. For the capacity utilization rate, we use demeaned data. For all other observables, we use demeaned growth rates. We subtract the growth rate of the labor force from volume-related variables such as macro aggregates, emissions, sectoral inputs/outputs and hours worked. We take prices relative to the GDP deflator to match the definition of prices in the model. Some approximations are needed due to data limitations. The rate of capacity utilization is for manufacturing which, empirically, is the biggest chunk in our core goods sector. We use hours worked in manufacturing only, which covers all of our estimation period.24 For low-skilled wage inflation, we use money wages of construction workers, 80% of whom have less than high school education (Cengiz and Tekg¨ ¸ 2022). For high-skilled wage inflation, uc we take wages of information and communication sector, which attracts mainly university degree holders. For global fossil fuel prices, we use OECD producer prices from the IEA. 22 Data sources are as follows: Components of GDP, employment, interest rates, capacity utilization rate and hours worked (OECD Main Economic Indicators), public-private investment, compensation of employees, unemployment, high-skilled and low-skilled wages (TurkStat), price indexes, EU output, inflation and prices (Eurostat), energy production and prices (IEA), emissions (CAIT) and real effective exchange rate (BIS). 23 Public and private investment expenditure series are not available after 2016Q2. Low-skilled wage inflation is available over 2005Q2-2017Q4 and high-skilled wage inflation is available after 2009Q2. 24 Total hours worked data from TurkStat is available after 2010, and there is a close match between quarterly growth rates of manufacturing hours worked and total hours worked over 2010-2022. 58 Table B1: Time series used in model estimation Real Total Gross Domestic Product, quarterly (demeaned growth rate) Real Private Final Consumption Expenditure, quarterly (demeaned growth rate) Real Gross Fixed Capital Formation, quarterly (demeaned growth rate) Real Government Final Consumption Expenditure, quarterly (demeaned growth rate) Real Exports of Goods and Services, quarterly (demeaned growth rate) Real Imports of Goods and Services, quarterly (demeaned growth rate) Real public investment expenditure, quarterly (demeaned growth rate) Real private investment expenditure, quarterly (demeaned growth rate) Rate of capacity utilization, quarterly (demeaned) Gross Domestic Product deflator inflation, quarterly (double-sided HP filtered) Harmonized Index of Consumer Prices: Overall Index Excluding Energy, quarterly (demeaned growth rate) Harmonized Index of Consumer Prices: Energy, quarterly (demeaned growth rate) Employment rate: number of people employed/total labor force, quarterly (double-sided HP filtered) Real compensation of employees, quarterly (demeaned growth rate) Total Hours worked, quarterly (demeaned growth rate) Low-skilled hourly wage inflation, quarterly (double-sided HP filtered) High-skilled hourly wage inflation, quarterly (double-sided HP filtered) Discount rate, quarterly (double-sided HP filtered) 3-month short-term interest rate, quarterly (double-sided HP filtered) Real Broad Effective Exchange Rate, quarterly (double-sided HP filtered) Total final household (residential+public+half of transportation) energy consumption, annual (demeaned growth rate) Total final firm (industry+half of transportation) energy consumption, annual (demeaned growth rate) Renewable energy supply, annual (demeaned growth rate) Fossil energy supply, annual (demeaned growth rate) CO2 emissions, annual (demeaned growth rate) Domestic fossil fuel producer prices, quarterly (demeaned growth rate) OECD fossil fuel producer prices, quarterly (demeaned growth rate) European Union Real Gross Domestic Product per capita, quarterly (demeaned growth rate) European Union Gross Domestic Product deflator inflation, quarterly (double-sided HP filtered) European Union export price deflator, quarterly (demeaned growth rate) Euro Area 10-year government bond interest rate, quarterly (double-sided HP filtered) B.2 Empirical model parametrization Tables B2 and B3 provide all relevant information regarding the calibration. They list every model parameter including a brief description. They report the calibrated values and how we came up with them: calibrated, or restricted. A set of parameters have been restricted in order for a set of related variables to match their empirical counterparts.25 Typically, parameters have been restricted which have strong effects on the steady state but not so much on the transitional dynamics. To give one example, the consumption demand scaling parameter is restricted to ensure that GDP is equal to one at the steady state. Then, the steady states of many other variables have straightforward interpretations. To give another example, we know from the data that the ratio between renewable and fossil energy is around 0.2 in T¨ urkiye. Hence, we restrict the 25 For the estimation, a parameter restriction implies that the affected parameter is adjusted in every iteration of the sequential estimation procedure in order to uphold the restriction. As the Bayesian algorithm travels the parameter space to assess how likely certain parameter values are the restricted parameters are updated every step of the way. 59 corresponding share parameter in the CES production function of the E-firm such that the model reproduces this ratio at the steady state (but not out of the steady state!). Let us briefly discuss the relevance of each of the model parameters: • The inter-temporal discount factor β controls to what extend households discount future utility to assess its present value. Since households own the firms the discount factor is also relevant for the discounting of future profits. Note that in DSGE models, β is usually calibrated as the inverse of the average of the historical real interest rate which is implied by solution of the households optimization problem. With wealth in the utility function this relationship collapses and β can be estimated. • The parameter scaling the utility derived from wealth ψA captures the precautionary saving preferences of the households. The remaining scaling parameters in the utility functions scale the utility from consumption and the disutility from labor. Since scaling parameters mainly affect the steady state values but not the transitional dynamics, they are all restricted to have certain variables or ratios of variables match their empirical counterparts, as reported in table B2. • The consumption habit persistence parameter κ controls how sluggish the response of consumption to changes in income or the interest rate. This parameter is estimated and helps the model reproduce the empirical observation that aggregate consumption typically gradually adjusts to a shock with only modest jumps. The higher κ, the more gradual the adjustment. • The parameter of relative risk aversion φ is the inverse of the elasticity of inter-temporal substitution and controls how strongly consumption changes in response to a change in the real interest rate. The lower φ and, hence, the higher the elasticity of inter- temporal substitution, the more consumption decreases in response to an increase of the real interest rate. Note that with uncertainty about future events risk aversion and inter-temporal elasticities are related concepts. The more risk averse the household, the more reluctant she will be to give up consumption today and save it for tomorrow. This is relevant for the discounting of future profits which needs to take this preference regarding uncertainty into account in addition to the pure rate of time preference β . • The inverse of the Frisch elasticity of labor supply η controls how strongly the labor hours supplied by the households respond to changes in the real wage. The higher η , the lower the elasticity, and the smaller the increase of labor supply in response to an increase in the real wage. • The risk premia on bond and equity returns have straightforward interpretations. They drive a wedge between the returns and the policy rate. The premia can be made endogenous to the financial structure of the bond and equity issuers. 60 Table B2: Parameter description, calibration, and restrictions (Part A) Parameter Description Value Source Households β Intertemporal discount factor 0.99 Calibrated to literature ψA Utility from wealth scaling 1 Calibrated (only affects restrictions) ψC Utility from consumption scaling 19.631 ˜ Restricted such that GDP = GDP ψH Disutility from labor scaling for H-hhs 554147 Restricted s.t. LH /LL = LH/LL ΓB Flight-to-safety gov. bonds risk premium 0.001 Calibrated to empirical equivalent ΓS Equity risk premium 0.013 Calibrated to empirical equivalent γW RoW net-debt eff. on RoW bond risk prem. 0.003 Restricted s.t. E (BW/GDP ) = E (BW/GDP ) µH High-skilled wage mark-up 0.2 Calibrated to literature Firms µZ Core good price mark-up 0.2 Calibrated to literature δZ Capital depreciation rate for core firms 0.025 Calibrated to literature δR Capital depreciation rate for renew. energy 0.03 Calibrated to literature δF Capital depreciation rate for fossil energy firms 0.02 Calibrated to literature ϕER Learning by doing rate 0.1 Calibrated to literature λ All firms debt-capital ratio 0.339 ˜ ˜P /GDP Restricted s.t. B = BP/GDP τZ1 Linear capital utilization costs 0.046 Restricted s.t. uZ = uZ Input share parameters in Constant-Elasticity-of-Substitution functions αO Imports share of fossil fuel consumption 0.870 Calibrated to empirical equivalent αKZ Private cap. in core good cap. services 0.736 ˜ ˜PZ /GDP Restricted s.t. I = IP Z/GDP αKR Private cap. in renew. energy cap. services 0.502 ˜ ˜ Restricted s.t. IPR /GDP = IP R/GDP αKF Private cap. in fossil energy cap. services 0.537 ˜ ˜PF /GDP Restricted s.t. I = IP F/GDP αE Renew. energy to fossil energy 0.122 Restricted s.t. pER E˜R /pEF E˜F = pER ER /pEF EF αVZ Cap. in value added of core firms 0.414 Restricted s.t. p ˜ ˜L LZ /pZ Z = CEZ /Z αVR Cap. in value-added of renew. energy 0.998 Restricted s.t. LR /LF = LR /LF αVF Cap. in value-added of fossil energy 0.992 Restricted s.t. p˜ ˜ ˜ L (LR + LF )/(pER ER + pEF EF ) = CEE /E αEF Value added in output of fossil firms 0.243 ˜ Restricted s.t. O/GDP = O/GDP αZ Value added in output of core firms 0.921 Restricted s.t. EY /EZ = EY /EZ αY Wholesale goods in retail goods 0.733 Restricted s.t. p ˜E (EY + EZ )/GDP ˜ = pEE/GDP αW Domestic core goods in wholesale goods 0.815 p∗ Restricted s.t. E (˜ Z˜ ∗ + αO p ˜ ˜ ∗ O )/GDP = M/GDP Z e O αW ∗ RoW core goods in RoW wholesale goods 0.742 Restricted s.t. E = E αL High-skilled labor in labor services 0.762 Restricted s.t. C˜H /C˜L = CH/CL Taxes T˜LL Transfers to L-hhs 0.081 ˜ ˜G /GDP Restricted s.t. B = BG/GDP tC Final sales tax rate 0.16 Calibrated s.t. cons. taxes over GDP are 0.11 tA Wealth tax rate 0.0003 Calibrated s.t. property taxes over GDP are 0.01 tR Tax rate on hhs. cap. income and cap. gains 0.008 Calibrated s.t. pers. inc. taxes over GDP are 0.04 tH High-skilled labor income tax rate 0.08 Calibrated s.t. pers. inc. taxes over GDP are 0.04 tL Low-skilled labor income tax rate 0.06 Calibrated s.t. pers. inc. taxes over GDP are 0.04 tS Firm contr. labor tax rate (payroll tax) 0.17 Calibrated s.t. payroll taxes over GDP are 0.07 tP Corporate income tax rate 0.045 Calibrated s.t. corp. inc. taxes over GDP are 0.02 tEF Fuel tax rate 0.025 Calibrated s.t. energy taxes over GDP are 0.02 tO Carbon tax rate 1E-05 Calibrated to empirical equivalent Notes : H-hhs and L-hhs stands for high-skilled labor households and low-skilled labor households, respectively. RoW stands for the rest of the world. Data sources reported in the main text. 61 Table B3: Parameter description, calibration, and restrictions (Part B) Parameter Description Value Restriction such that Steady-state parameters g Deterministic growth rate 1.0083 Calibrated to empirical equivalent ΠY Price inflation rate 1.0199 Calibrated to empirical equivalent ΠH High-skilled wage inflation rate 1.0282 Calibrated to ΠY + g − 1 ΠL Low-skilled wage inflation rate 1.0282 Calibrated to ΠY + g − 1 R Policy rate 1.0219 Calibrated to empirical equivalent RW Interest rate in the RoW 1.0036 Calibrated to empirical equivalent CG Gov. cons.-GDP ratio 0.14 Calibrated to empirical equivalent IGZ Gov. inv. in core goods-GDP ratio 0.036 Calibrated to empirical equivalent inf IG R Gov. infra. inv. in renew. energy-GDP ratio 0.001 Calibrated to empirical equivalent gen IG R Gov. gen. inv. in renew. energy-GDP ratio 0.001 Calibrated to empirical equivalent inf IG F Gov. infra. inv. in fossil. energy-GDP ratio 0.001 Calibrated to empirical equivalent gen IG F Gov. gen. inv. in fossil energy-GDP ratio 0.001 Calibrated to empirical equivalent EB ˜ ˆW /GDP RoW net debt-GDP ratio -1.6 Calibrated to empirical equivalent Y∗ Final demand in the RoW 1 Normalized p∗Z Core goods price in the RoW 1 Normalized p∗O Carbon supply cost 0.07 ˜∗ Restricted s.t. E p O O/GDP = PFES/GDP Steady-state targets GDP GDP 1 Normalized E Real exchange rate 1 Normalized LH/LL High-low-skilled labor ratio 1 Calibrated to empirical equivalent LR/LF Renew.-fossil employment ratio 0.15 Calibrated to empirical equivalent EL Employment rate 0.708 Restricted s.t. ψL ≡ ψH IP Z/GDP Private core investment-GDP ratio 0.233 Calibrated to empirical equivalent IP R/GDP Renew. investment-GDP ratio 0.003 Calibrated to empirical equivalent IP F/GDP Fossil investment-GDP ratio 0.004 Calibrated to empirical equivalent CEZ/Z Wage share of output in core firms 0.39 Calibrated to empirical equivalent CEE/E Wage share of output in energy firms 0.08 Calibrated to empirical equivalent BP/GDP Corp. bonds-GDP ratio 2.3 Calibrated to empirical equivalent pER ER/pEF EF Renew.-fossil energy ratio 0.2 Calibrated to empirical equivalent O/GDP Carbon intensity of GDP as kg of CO2 per 0.46 Calibrated to empirical equivalent USD EY /EZ Energy cons. of households-firms ratio 0.94 Calibrated to empirical equivalent pEE/GDP Total final energy cons. exp.-GDP ratio 0.067 Calibrated to empirical equivalent M/GDP Total imports-GDP ratio 0.26 Calibrated to empirical equivalent CH/CL High-low-skilled consumption ratio 2 Calibrated to empirical equivalent uZ Cap. utilization 1 Calibrated to empirical equivalent BG/GDP Gov. bonds-GDP ratio 1.32 Calibrated to empirical equivalent PFES/GDP Fossil primary energy supply exp.-GDP ratio 0.04 Calibrated to empirical equivalent Notes : RoW stands for the rest of the world. Data sources reported in the main text. 62 • The wage and price mark-ups µW and µP are defined implicitly through the substitu- tion elasticities between labor and core good varieties, respectively. The higher these elasticities, the smaller the market power and, hence, the smaller the mark-ups. • We calibrate sector specific depreciation rates δZ , δR , and δF with respect to results of Argentiero et al. (2017) for EU-15 countries. • The debt-capital ratio λ is a parameter typically not found in DSGE models. This is because there is no steady-state wealth in conventional DSGE models and saving is a pure transitory phenomenon. The Modigliani-Miller theorem holds and the financial structure of the firm sector is indeterminate. With wealth in the utility function capturing precautionary saving, the financial structure does become relevant. It is governed by λ which controls how much of investment is financed by corporate bonds and how much by equity. • All production functions in the model are of the CES type. The share parameters do not have strong effects on the transition dynamics. All of them are restricted as indicated in Table B2. • The corresponding elasticities of substitution are critical for transitional dynamics. They control how flexible the production structure can respond to price and quantity shocks. For instance, an increase in the carbon price will induce fossil energy firms to reduce the carbon intensity of fossil fuels. To the extent they succeed in this endeavor, they can limit fossil fuel price increases. Similarly energy firms will face higher fossil energy costs and try to use renewable energy instead. The elasticity of substitution between renewables and fossil energy will be crucial (as well as the availability and costs of renewables). With a rigid supply chain, price increases will be passed through, cause inflation, and induce the central bank to raise interest rates and the trade balance to deteriorate. Output effects may be considerable. Because substitution elasticities are critical, all of them are estimated. • The learning by doing rate ϕER captures productivity increases in renewable energy production coming with the scale of energy output. We calibrate this parameter to 0.1, which corresponds to a learning by doing elasticity of 0.152, which is close to the lower bound of findings in the literature (Rubin et al. 2015). • The model features various quadratic adjustment costs. These costs do not affect the steady state but the short-run behavior of the model. Hence, the corresponding scaling parameters are all estimated. • Since, empirically, employment measured in persons varies to a lesser extent than labor input measured in hours, the model assumes that only a share of labor firms can adjust low-skilled employment. A share ξEL is stuck with the previous level. This Calvo-type of real rigidity allows the model to capture the different volatilities of employment and hours worked. ξEL is estimated. 63 • Endogenous policy responses are characterized by elasticities. They only affect transi- tional dynamics (not the steady state) and are, therefore, estimated. The low-skilled wage inflation responds to changes in employment and price inflation. The stronger wages respond to the labor market, the closer the model is to a DSGE closure with labor market clearing. • To capture automatic stabilizers, public consumption and investment respond to out- put and output gap elasticities are estimated.26 • The policy interest rate responds to core inflation and the output gap. Note that, contrary to DSGE models, the model does not require (but allow) the inflation elasticity of the interest rate to be larger than 1. Note that the model also requires this so-called Taylor principle to hold when the wage rate responds strongly to the labor market and the model converges to the DSGE model. • The various tax rates are calibrated to match the corresponding tax revenues-to-GDP ratios. Note, however, that we did not strictly restrict them as they crucially affect the transitional dynamics. • Stochastic shocks have been added to the model to improve its empirical fit. Note that without stochastic elements there would be no uncertainty and the model would be able to generate only smooth transitions. Historical data, however, include considerable noise. These stochastic elements are typically specified in an auto-regressive form, and the corresponding auto-regressive parameters are estimated. They capture how persistent shocks are. • Finally, the model includes various parameters which have a direct, empirical equiva- lent. For instance, the deterministic growth rate is the average growth rate of labor productivity. The long-run or steady-state rates of inflation and the interest rate should correspond to their historical averages. The same holds for the steady-state govern- ment consumption and public investments as well as external debt. Since the RoW is specified in reduced form, the steady state values of output and core prices can be calibrated arbitrarily. The steady-state international price of carbon is restricted such that the carbon intensity of GDP matches the empirical counterpart at the steady state. B.3 Empirical steady state values and ratios What are the steady-state values and ratios the calibration seeks to match? As mentioned above, selected model parameters which mainly affect the steady state rather than transi- tional dynamics are calibrated to ensure that the model reproduces certain values and ratios 26 We restrict the output gap elasticities of investment to be equal to output gap elasticity of consumption at each iteration in estimation. 64 of variables at the steady state. The respective restrictions are reported in the last column of Tables B2 and B3. • The consumption utility scaling parameter is set to ensure that GDP is equal to one at the steady state. This is done for convenience only as most other steady-state values then have a clear interpretation as ratios to GDP. • The share parameter in RoW wholesale production function is restricted to give real exchange rate equal to 1 at the steady state. • The ratio of high-skilled and low-skilled labor hours is roughly 1, taken from main labor force indicators by education level data of TurkStat. The corresponding consumption ratio is around 2, calculated using distribution of consumption over income quantiles data of TurkStat under the assumption that low-skilled households constitute the low- est 50% of the income distribution. Hence, H-households consume twice as much as L-households while they provide the same number of labor hours. This is possible because the formers’ real wage is higher, and they receive capital income. • The share of private non-energy, renewable energy and fossil energy investment to GDP ratios are around 23.3%, 0.3% and 0.4%, respectively. Total investment share of GDP comes from GLORIA database. We use World Input Output Database (WIOD) to split investment to core and energy. • The annualized public net debt-GDP, corporate net debt-GDP, and RoW net debt- GDP ratios are around 33%, 57%, and -40%, respectively, from the Financial Accounts urkiye (CBRT). Reports of the Central Bank of the Republic of T¨ • The ratio of renewable and fossil energy is around 0.2, averaged over 2019-2021 from the energy balance of IEA27 . We assume the ratio of final use (residential, public, and half of transportation) and intermediate use (industry, and half of transportation) of the TFC is 0.94, average value over 2000-2022. Finally, about 87% of the Turkish TPES is imported, coming from GLORIA database. In the model, this corresponds to the share of imported carbon in total carbon input. • Wage share of output in core goods and energy sectors are around 39% and 0.08%, calculated from GLORIA database. • Renewable to fossil energy employment ratio is 0.15, calculated from Global Trade Analysis Project (GTAP)-Power and Gender Disaggregated Labor Database (GDLD) of the World Bank. 27 To obtain the renewable-fossil energy ratio, total final consumption (TFC) would be the appropriate measure. Yet, it is unclear to what extent electricity is generated by renewables and fossil energy. Hence, we use the ratio of total primary energy supply (TPES) as a proxy. 65 • We scale the carbon input to energy production to match the carbon intensity of GDP: 0.46 kg CO2 per 2015 $ of GDP, averaged over 2000-2022, from World Development Indicators of the World Bank. • The value of total energy consumption as a share of GDP is 7.1%, calculated from GLORIA database. • The constant USD price of carbon is restricted to attain primary fossil energy supply expenditure-GDP ratio of 3.9%, from GLORIA database. • The ratio of total imports to GDP is around 26%, calculated from GLORIA database. Note that the steady state values of a few important variables have a direct equivalent in the data and have been calibrated accordingly. These values can be set directly and do not require another parameter to be restricted. They are reported in Tables B2 and B3. B.4 Estimation strategy In this subsection, we discuss our Bayesian estimation strategy together with the results. The basic idea of Bayesian statistics comes from the Bayes’ rule: p(y|θ)p(θ) p(θ|y) = ∝ p(y|θ)p(θ) p(y) where θ is the vector of parameters, p(θ|y) is the probability distribution of parame- ters given data (posterior distribution), p(y|θ) is the probability distribution of data given parameters (likelihood), P (θ) is the prior distribution of θ and p(y) is the marginal distri- bution of data. We are interested in marginal posterior distribution of model parameters; however it is not always possible to analytically compute marginal posterior distributions due to complex integral calculations. Bayesians, therefore, have developed numerical methods that allow us to approximate marginal posterior distributions of parameters. The basic idea is to draw from some other distribution (proposal or importance distributions) and find moments of simulated data, which are going to converge to moments of posterior distribution under specific conditions (Herbst and Schorfheide 2015). We use Metropolis Hastings (MH) algorithm, which is part of MCMC family. MH al- gorithm simply evaluates each parameter vector drawn (θi at iteration i) from a proposal distribution at the kernel of the posterior distribution over the proposal distribution, and compares this to the ratio of the vector stored in iteration (i-1). If the ratio of the former to the latter is higher than a random number u ∈ (0, 1), then candidate draw is accepted and stored. If not, θi−1 is stored again among accepted parameter vectors. There are two important components of the MH algorithm. One is the proposal dis- tribution, which is assumed to be a multivariate normal distribution with mean θi−1 (last 66 accepted draw) and a covariance matrix Σ, which is calculated as the inverse Hessian at the mode. The second component is the likelihood function, which must be evaluated before the MH algorithm starts running. We apply the Kalman filter to the linear state-space form, which includes a state equation describing the evolution of states based on their past values and a measurement equation showing how states are related to data, to get the likelihood. st = ϕt st−1 + ϵt−1,s ; ϵt−1,s ∼ N (0, Σs t−1 ) yt = Bt st + ϵt,m ; ϵt,m ∼ N (0, Σm t ) for t=1,...,n where Σs m t and Σt are covariance matrices of state and measurement shocks ϵt,s and ϵt,m . Kalman filter simply operates on two steps: It first makes a prediction about a new state st based on the state transition equation and then corrects for it when new data yt become available. If yt are missing, Kalman filter still guesses st using the prediction step. Then, it predicts missing yt using the measurement equation. This feature of the Kalman filter makes adding mixed frequency or missing observations possible. For example, we have energy balances available at the annual frequency. We simply feed what we have to the Kalman filter and leave missing quarterly observation as NaN. Similarly, public/private investment data are not published after 2016 Q2. We feed only available data to the filter and leave the remaining values as NaN. Kalman filter fills these missing observations by using the best available information. We use the software Dynare 5.3 on Matlab to simulate posterior distribution of 85 model parameters. We run two chains of MH algorithms with 50,000 iterations each, and burn in the first half of the accepted draws. We choose the scaling parameter of the proposal distribution such that the acceptance ratio is around 0.25. B.5 Prior distributions Tables B4 and B5 report the prior and posterior distributions of all estimated model pa- rameters. We generally choose loose priors to let the data speak as freely as possible. We take the means of prior distributions of household preference parameters from Smets and Wouters (2007) and Christoffel et al. (2008). The elasticity of substitution between high- and low-skilled labor is centered at 1.5 with a standard deviation of 0.5 (Cantore et al. 2017). The prior distributions of the domestic Armington elasticity (between domestic and for- eign goods) and the corresponding elasticity of the RW are taken from Albonico et al. (2019). The mean of the prior distribution of the elasticity of substitution between core and energy consumption goods comes from the posterior mean of oil and core consumption elasticity in An et al. (2011). The mean elasticity of substitution between value-added and energy is 0.5 – a common value in the literature reported by Van der Werf (2008). The standard deviation is 0.1. 67 Table B4: Parameter description and estimation Parameter Description Prior Prior Prior Posterior 90% HPD dist. mean std. mean interval Households κ Consumption habit persistence B 0.7 0.05 0.62 [0.54, 0.71] ϱ Inverse elast. of intertemp. subst. G 1.5 0.375 1.08 [0.70, 1.49] η Inverse Frisch elast. of labor supply G 2 0.25 1.73 [1.33, 2.09] Elasticities of substitution in CES functions between σZ Value added & energy for core goods firms G 0.5 0.1 0.75 [0.65, 0.85] σL High & low skilled labor for labor services G 1.5 0.5 2.94 [2.66, 3.20] σY Core goods & energy for retail firms G 0.3 0.1 0.14 [0.07, 0.20] σF Value added & carbon for fossil energy firms G 0.3 0.1 0.05 [0.03, 0.07] σVR Capital & labor for renew. energy firms G 0.3 0.05 0.30 [0.22, 0.39] σVF Capital & labor for fossil energy firms G 0.3 0.05 0.28 [0.20, 0.35] σW Dom. & RoW core goods for wholesale firms G 2 0.4 1.57 [1.45, 1.68] σW ∗ RoW & dom. core goods for RoW wholesale firms G 2 0.4 1.31 [0.82, 1.82] σE Renew. & fossil energy for energy firms G 6 4 18.51 [10.12, 26.56] Scaling parameters ψu Scaling parameter for capacity utilization rate N 2 1 2.68 [1.69, 3.74] Adjustment cost scaling parameters τW Quadratic high-skilled wage adjustment costs G 50 20 92.86 [56.78, 128.61] τP Quadratic core good price adjustment costs G 50 20 59.90 [36.26, 83.11] τI Quadratic investment adjustment costs G 10 4 4.49 [2.84, 6.39] τZ2 Quadratic capital utilization costs G 0.5 0.1 0.34 [0.20, 0.47] τO Quadratic carbon input adjustment costs G 5 1 3.29 [2.21, 4.33] τZe Quadratic export adjustment costs G 5 1 4.33 [2.95, 5.70] τE Quadratic renew. adjustment costs G 2 1.25 0.80 [0.13, 1.48] Adjustment elasticities ξEL Calvo-style employment parameter B 0.5 0.15 0.96 [0.95, 0.97] ϕWE Empl. elast. of low-skilled wage infl. N 1 0.5 1.29 [0.74, 1.82] ϕWP Price infl. elast. of low-skilled wage infl. N 1 0.5 0.67 [0.24, 1.10] ϕCY Output gap elast. of gov. cons. N 0 0.1 -0.12 [−0.24, 0.02] ϕR Π Core infl. elast. of policy rate N 1.1 0.2 1.22 [1.01, 1.42] ϕRY Output gap elast. of policy rate N 0.5 0.2 0.02 [−0.05, 0.09] Persistence parameters ρWD Domestic input productivity shock for wholesale firms B 0.5 0.1 0.75 [0.67, 0.82] ρWI Imported input productivity shock for wholesale firms B 0.5 0.1 0.76 [0.67, 0.85] ρW ∗ Export demand shock for RoW wholesale firms B 0.5 0.1 0.29 [0.18, 0.41] ρR Interest rate smoothing of the home country B 0.8 0.1 0.90 [0.88, 0.92] ρC Consumption preference shock B 0.5 0.1 0.32 [0.20, 0.45] ρNH High-skilled labor supply preference shock B 0.5 0.1 0.40 [0.27, 0.53] ρNL Low-skilled labor supply preference shock B 0.5 0.1 0.45 [0.35, 0.56] ρRB Flight to safety shock B 0.5 0.1 0.53 [0.45, 0.62] ργW Foreign bond risk premium shock B 0.5 0.1 0.65 [0.55, 0.76] ρRS Equity risk premium shock B 0.5 0.1 0.46 [0.33, 0.57] ρPZ Core good price mark up shock B 0.5 0.1 0.66 [0.55, 0.77] ρVZ Productivity shock for core goods firms B 0.5 0.1 0.69 [0.56, 0.81] ρZE Energy efficiency shock to core goods firms B 0.5 0.1 0.75 [0.64, 0.86] ρVF Productivity shock for fossil energy firms B 0.5 0.1 0.67 [0.56, 0.78] Notes : RoW stands for the rest of the world. Data sources reported in the main text. 68 Table B5: Parameter description and estimation Parameter Description Prior Prior Prior Posterior 90% HPD dist. mean std. mean interval Persistence parameters (continued) ρVR Productivity shock for renew. energy firms B 0.5 0.1 0.71 [0.56, 0.86] ρE Productivity shock for energy firms B 0.5 0.1 0.75 [0.68, 0.82] ρKZ Productivity shock for core goods capital bundlers B 0.5 0.1 0.64 [0.55, 0.74] ρCG Government consumption shock B 0.5 0.1 0.66 [0.57, 0.75] ρIG Government investment shock B 0.5 0.1 0.70 [0.61, 0.79] ρY ∗ Demand shock in the RoW B 0.75 0.05 0.78 [0.72, 0.84] ρpZ ∗ Core goods price shock in the RoW B 0.75 0.05 0.89 [0.86, 0.92] ρWL Low-skilled wage inflationsmoothing B 0.5 0.1 0.24 [0.14, 0.34] ρDE Windfall profit shock for energy firms B 0.5 0.1 0.77 [0.71, 0.83] ρLH High-skilled input productivity shock for labor firms B 0.5 0.1 0.59 [0.43, 0.76] ρLL Low-skilled input productivity shock for labor firms B 0.5 0.1 0.76 [0.69, 0.83] ρY Productivity shock for retail firms B 0.5 0.1 0.75 [0.66, 0.84] ρpO∗ Fossil fuel price shock B 0.75 0.05 0.85 [0.81, 0.89] ρΠ∗ RoW inflation shock B 0.5 0.1 0.32 [0.21, 0.42] Shock parameters ϵWL Low-skilled wage inflation shock IG 0.1 1 0.047 [0.0309, 0.0619] ϵR Home country interest rate shock IG 0.01 0.01 0.0045 [0.0039, 0.0051] ϵRW RoW long-term interest rate shock IG 0.001 0.001 0.0011 [0.0009, 0.0012] ϵC Consumption preference shock IG 0.1 1 0.216 [0.118, 0.299] ϵNH High-skilled labor supply preference shock IG 0.1 1 4.153 [2.490, 5.656] ϵNL Low-skilled labor supply preference shock IG 0.1 1 4.200 [2.503, 5.943] ϵRB Flight to safety shock IG 0.01 0.01 0.018 [0.015, 0.020] ϵRS Equity risk premium shock IG 0.1 1 0.140 [0.066, 0.235] ϵPZ Core good price mark-up shock IG 0.1 1 0.806 [0.513, 1.069] ϵZE Energy efficiency shock to core goods firms IG 0.1 1 0.248 [0.115, 0.384] ϵVZ Productivity shock for core goods value-added firms IG 0.1 1 0.033 [0.025, 0.042] ϵVR Productivity shock for renew. energy firms IG 0.1 1 0.088 [0.052, 0.121] ϵE Productivity shock for energy firms IG 0.1 1 0.082 [0.072, 0.092] ϵVF Productivity shock for fossil energy firms IG 0.1 1 0.266 [0.193, 0.333] ϵDE Windfall profit shock for energy firms IG 0.1 1 0.005 [0.004, 0.006] ϵLH High-skilled input productivity shock for labor firms IG 0.1 1 0.043 [0.026, 0.061] ϵLL Low-skilled input productivity shock for labor firms IG 0.1 1 0.258 [0.199, 0.314] ϵKZ Productivity shock for core goods capital bundlers IG 0.1 1 0.076 [0.063, 0.091] ϵCG Government consumption shock IG 0.1 1 0.040 [0.035, 0.045] ϵIG Government investment shock IG 0.1 1 0.100 [0.085, 0.112] ϵY ∗ Demand shock in the RoW IG 0.01 0.01 0.017 [0.015, 0.020] ϵpZ ∗ Core goods price shock in the RoW IG 0.01 0.01 0.010 [0.009, 0.011] ϵγW Foreign bond risk premium shock IG 0.1 1 0.148 [0.100, 0.196] ϵY Productivity shock for retail firms IG 0.1 1 0.037 [0.030, 0.044] ϵWD Domestic input productivity shock for wholesale firms IG 0.1 1 0.049 [0.040, 0.057] ϵWI Imported input productivity shock for wholesale firms IG 0.1 1 0.130 [0.105, 0.155] ϵW ∗ Export demand shock for RoW wholesale firms IG 0.1 1 0.378 [0.209, 0.554] ϵpO∗ Global fossil fuel price shock IG 0.1 1 0.092 [0.081, 0.103] ϵΠ∗ RoW inflation shock IG 0.01 0.01 0.0031 [0.0028, 0.0035] ϵGDPm Output measurement shock IG 0.01 0.01 0.035 [0.031, 0.039] ϵInvm Investment measurement shock IG 0.01 0.01 0.087 [0.076, 0.100] Notes : RoW stands for the rest of the world. Data sources reported in the main text. 69 The elasticity of substitution between value-added and carbon is an original feature of the model which implicitly captures the substitution between different fossil fuels (coal&gas, coal&oil, oil&gas). We prefer a low prior mean at 0.3, with a standard deviation of 0.1. The long-run elasticity of substitution between renewable and fossil energy, is a critical parameter of the model. We choose the mean of 6 based on Varga et al. (2022) and set a loose prior with a standard deviation of 4. Similar to Guerrieri et al. (2008), we introduce a renewable energy-input adjustment cost to energy firms to get time-varying elasticity of substitution which is low in the short-run and converges to the long-run value. For the adjustment cost scaling parameter we choose a prior with mean 2 and standard deviation 1.25. For the adjustment cost scaling parameters, we use common priors from the literature (see Smets and Wouters 2007, Christoffel et al. 2008, Albonico et al. 2019). For the em- ployment and inflation elasticities of nominal wage setting, we use normal distributions with mean 1 and standard deviation 0.5 to cover to a wide range of possible values. For public consumption rule, we choose an output gap elasticity centered at 0 to let data decide whether these policies are counter- or pro-cyclical.28 For monetary policy parameters, we deviate from the standard literature which assumes large inflation response parameters (1.5 and above) in the Taylor rule. Instead, given the weak monetary policy regime in T¨ urkiye post-2010, we choose a lower prior 1.1 for the inflation response parameter with standard deviation of 0.2 (G¨urkaynak et al. 2022). For the output-gap elasticity, we follow the literature and set a normal prior with mean 0.5 and standard deviation 0.2. For most of the shock persistence parameters, we use beta priors centered at 0.5 with a standard deviation of 0.1. For foreign demand, import and fossil fuel price shocks, which are highly autoregressive, we prefer priors with mean 0.75 and standard deviation 0.05. We center the prior of the interest rate smoothing parameter to 0.8, following Christoffel et al. (2008). For almost all shock processes we choose inverse gamma distributions with mean 0.1 and standard deviation 1. For domestic interest rates and all foreign shocks, we prefer lower prior means and standard deviations. B.6 Posterior distributions Posterior distributions of structural parameters are plotted in Figure B1. All behavioural parameters of the households are identified and in line with the literature. Our estimate of consumption habit persistence is similar to the result of Christoffel et al. (2008) for the EU but less than estimates of Smets and Wouters (2007) for the US. Our mean of the inverse Frisch elasticity of labor supply is similar to the finding of C ¸ ebi (2012) for T¨urkiye.29 On the other hand, we estimate the risk aversion parameter lower than C ¸ ebi (2012) and what Smets and Wouters (2007) and Albonico et al. (2019) estimate for the US and EU, respectively. 28 Since they are not identified separately, we assume that the output-gap elasticities of public investment are equal to the elasticity of public consumption at each iteration of the Metropolis Hastings algorithm. 29 We only estimate the inverse Frisch elasticity of labor supply for high skilled households and assume that low-skilled parameters is equal to that in each iteration of the Metropolis-Hastings algorithm. 70 This is not surprising given that our model features wealth in the utility function to capture precautionary saving motives explicitly. All elasticities of substitution are identified by data except the capital-labor elasticities for renewable energy firms. We estimate value-added and energy elasticity as 0.75, which is slightly above the range of industry estimates in Van der Werf (2008). Our estimate for the elasticity of substitution between carbon and value-added in fossil energy sector is very low (0.05), which indicates the difficulty of substituting out carbon in fossil energy production and implicitly the low degree of substitution between fossil fuels with different carbon contents in T¨ urkiye. Note that this makes the incidence and effects of carbon and fuel taxes almost identical in our model. We estimate the elasticity of substitution between high and low skilled labor around 3, which is close to the upper limit in the literature (Havranek et al. 2020). The reason is the high correlation between the high- and low-skilled wage time series. The mean of the domestic Armington elasticity is lower than the prior and close to the Albonico et al. (2019) results for EU. The export demand elasticity has a mean 1.31, close to the calibration of Millard (2011) for the UK. The long-run elasticity of substitution between renewable and fossil energy is strongly identified by data with a posterior mean of 19 and a 90% High Density Interval (HDI) of [10, 27]. We estimate the elasticity of substitution between energy and non-energy final goods around 0.14 which is lower than the posterior mean of An et al. (2011) with DSGE (0.3) and DSGE-VAR (0.2) specifications for the Republic of Korea. Posterior means of wage and price adjustment costs are higher than prior means indicating substantial wage and price adjustment costs in T¨ urkiye. We estimate investment adjustment cost of around 5, in line with the results of Christoffel et al. (2008) and Smets and Wouters (2007), and capital utilization adjustment cost around 0.34. Our mean estimate for export adjustment cost is 4.3. We estimate mean carbon input adjustment cost as 3.3 and renewable input adjustment cost as 0.8. We estimate the Calvo (1983) style employment parameter to be higher than the value found by Christoffel et al. (2008), which indicates that a large share of firms in T¨ urkiye do not adjust employment with variations in low-skilled labor hours. In the Phillips curve equation, the employment elasticity of wage inflation is around 1.29, and the wage indexation parameter is 0.67. They point to a significant demand impact on low-skilled wage inflation. We estimate the output-gap elasticity of public expenditures around −0.12 which suggests urkiye. that these expenditures are pro-cyclical in T¨ We estimate the interest rate smoothing parameter as 0.9 which is higher than C ¸ ebi urkiye over 2002Q1-2009Q3. This is not surprising as the Central (2012)’s finding (0.62) for T¨ Bank of T¨ urkiye has not been changing rates in response to inflation in the second half of our sample. Similarly, our inflation coefficient is around 1.22 capturing loose monetary policy in T¨urkiye post-2010 unlike C ¸ ebi (2012)’s estimate of 1.75 which points to a strictly inflation targeting monetary policy. Our estimate for output gap elasticity is around 0.02. 71 Consumption habit persistence Inverse elast. of intertemp. subst. Inverse Frisch elast. of labor supply 1.5 1.5 6 1 1 4 2 0.5 0.5 0 0 0 0.3 0.4 0.5 0.6 0.7 0.8 0.9 -0.5 0 0.5 1 1.5 2 2.5 3 0.5 1 1.5 2 2.5 3 3.5 E.o.s between capital and labor for renew. E.o.s between capital and labor for fossil E.o.s between value-added and energy 8 8 6 6 6 4 4 4 2 2 2 0 0 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.2 0.4 0.6 0.8 1 1.2 E.o.s between value added and carbon E.o.s between high and low-skilled labor Import price elasticity 2 30 1.5 4 20 1 2 10 0.5 0 0 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 0.5 1 1.5 2 2.5 3 3.5 4 0.5 1 1.5 2 2.5 3 3.5 Export price elasticity E.o.s between renewable and fossil energy E.o.s between core goods and energy 10 1 0.1 5 0.5 0.05 0 0 0 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 -10 0 10 20 30 40 50 -0.2 0 0.2 0.4 0.6 0.8 High-skilled wage adjustment cost Core good price adjustment cost Investment adjustment cost 0.02 0.02 0.3 0.015 0.01 0.2 0.01 0.005 0.1 0 0 0 -50 0 50 100 150 200 250 300 -50 0 50 100 150 -5 0 5 10 15 20 25 30 Renewable input adjustment cost Carbon input adjustment cost Export adjustment cost 0.6 0.8 0.4 0.6 0.4 0.4 0.2 0.2 0.2 0 0 0 -2 0 2 4 6 8 10 -2 0 2 4 6 8 10 -2 0 2 4 6 8 10 Calvo empl. param. Empl. elas. of low-skilled wage infl. Price infl. elas. of low-skilled wage infl. 1.5 60 1 40 1 0.5 20 0.5 0 0 0 0 0.2 0.4 0.6 0.8 1 -1 0 1 2 3 4 -1 -0.5 0 0.5 1 1.5 2 2.5 3 Output gap elas. of gov. cons. Core infl. elas. of policy rate Output gap elas. of policy rate 3 4 8 2 3 6 2 4 1 1 2 0 0 0 -0.6 -0.4 -0.2 0 0.2 0.4 0.5 1 1.5 2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 Figure B1: Prior (gray) and posterior (black) distribution of structural model parameters. Dashed green lines represent modes of the posterior. Two chains of Random Walk Metropolis Hastings simulations, each with 50,000 iterations and 25,000 burn-in. Acceptance rate is around 0.25. 72 B.7 Sensitivity analysis The sensitivity analysis determines which parameters are crucial for the stability of the model. We use Dynare’s global sensitivity analysis toolkit. The tool first samples from a distribution (prior range, prior/posterior distributions) and groups draws into two subsets: Violating behavior and not-violating behavior. Behavior can be defined flexibly such as stability or indeterminacy of the model. Then, it plots cumulative density functions of behavior (blue) and non-behavior (red). It conducts Kolmogorov-Smirnov test under the null hypothesis that two distributions are identical against they are different. In the diagonals of Figure B2, as the vertical distance between blue and red density functions increases, the probability of rejecting the null goes up. We simulate from the prior range to investigate stability in a wide range and plot parameters with p-values less than 0.001 in their order of significance. Model stability is sensitive to the long-run elasticity of substitution between renewable and fossil energy σE , elasticity of substitution between value-added and energy σSZ , and monetary policy - wage inflation rule parameters. High σE reduces the probability of finding a solution. Blue line above red line indicates that stable solutions accumulate at low values of σE . Monetary policy parameters are important for the indeterminacy of the model. The second panel of Figure B2 zooms in indeterminacy and shows two distinct monetary policy regimes in covariance plots of ϕRP and ϕW E . Unlike a standard DSGE model, a less than one-to-one response to inflation in Taylor Rule (ϕRP ≤ 1) does not necessarily cause indeter- minacy. If Phillips curve coefficient is negative (ϕW E ≤ 0), no restriction is needed for ϕRP for the rational expectations solution. On the other hand, if ϕW E > 0, the model converges to the DSGE case: ϕRP must be greater than 1 to rule out indeterminacy.30 30 Given high estimated ϕW E for T¨ urkiye, the model cannot explore the ϕRP < 1 region. Investigating regime switches further is in our research agenda. 73 P E P Y Z iW iW iR iR gE gS ph ph ph ph si si 50 gE si 0 2 1 P iR 0 ph 1 0.5 Z gS 0 si 4 2 P 0 iW -2 ph 4 2 E 0 iW -2 ph Y iR ph 0 50 0 1 2 0 0.5 1 -2 0 2 4 -2 0 2 4 0 1 2 unique Stable Saddle-Path NO unique Stable Saddle-Path E P Y SZ iW iR iR E g g ph ph ph si si 2 1 P iR ph 0 4 2 E 0 Wi ph -2 2 1 Y 0 iR ph 1 0.5 Z gS si 0 gE si 0 1 2 -2 0 2 4 0 1 20 0.5 1 0 50 NO indeterminacy indeterminacy Figure B2: Sensitivity analysis for the stability and indeterminacy of the model. 74