Policy Research Working Paper 10387 Simulating the Effect of Business Tax Abolition through a New Regional CGE Model Evidence from Italy Alessio Baldassarre Valerio Ferdinando Calà Danilo Carullo Hasan Dudu Elisa Fusco Pasquale Giacobbe Carlo Orecchia Macroeconomics, Trade and Investment Global Practice March 2023 Policy Research Working Paper 10387 Abstract The main goal of regional computable general equilibrium model of the World Bank. A new regional social accounting models is to analyze how different regions within a spe- matrix for Italy (20 regions at the Nomenclature of territo- cific area react to certain shocks. Therefore, countries with rial units for statistics level) has been constructed. The social high heterogeneity among regions, like Italy, constitute an accounting matrix is used as input data to simulate the abo- interesting case study for regional computable general equi- lition of the regional tax on productive activities (regional librium model analysis. This paper presents the regional part business tax) through three different scenarios, focusing of the new (recursive) dynamic single-country computable on the effects on gross domestic product, regional value general equilibrium model called the Italian Regional and added, and welfare. The results show that under the mod- Environmental Computable General Equilibrium of the eling assumptions, the complete abolition of the regional Department of Finance, based on the Mitigation, Adapta- tax on productive activities would positively impact Italian tion and New Technologies Applied General Equilibrium economic growth and regional welfare. This paper is a product of 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 hdudu@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 Simulating the Effect of Business Tax Abolition through a New Regional CGE Model: Evidence from Italy∗ Alessio Baldassarre1 , Valerio Ferdinando Calà1 , Danilo Carullo1 , Hasan Dudu2 , Elisa Fusco3 , Pasquale Giacobbe3 , and Carlo Orecchia1 1 Department of Finance, Italian Ministry of Economy and Finance, Rome, Italy 2 The World Bank 3 Department of Economic modelling and statistical analysis for policy making, SOGEI SpA, Rome, Italy Keywords: CGE model, SAM, business tax, fiscal policy JEL Codes: C68, E16, R13 ∗ This paper has been developed at the Department of Finance-Italian Ministry of Economy and Finance under the project ‘Assistance for the assessment of environmental tax reforms’, funded by the European Commission-DG Reform, in collaboration with The World Bank. We are particularly grateful to Maria Teresa Monteduro, Marco Manzo, Paolo Di Caro, Silvia Carta, Gavino Mura, Jan Ignacy Witajewski- Baltvilks, Maria Alessandra Tullio, Francesca Laratta, Lorenzo Marchetti, and Alessio Marabucci. We wish to thank Sogei spa for the support. The views expressed in the paper do not reflect those of the institutions of affiliation. The findings, interpretations and conclusions expressed herein are those of the authors and do not necessarily reflect the view of the European Commission, Italian Ministry of Economy and Finance, Sogei spa, World Bank Group, its Board of Directors or the governments they represent. The authors declare that they have no conflict of interest. 1 Introduction The Italian Regional and Environmental Computable General Equilibrium of the Department of Finance (IRENCGE-DF) Model is a (recursive) dynamic single-country computable general equilibrium (CGE) model. It was developed by the Italian Ministry of Economy and Finance (MEF) and the World Bank (WB). The IRENCGE-DF model is based on the MANAGE WB model of the World Bank, which is, in turn, based on the MANAGE model as in Van Der Mensbrugghe (2021). It is developed under the project "Assistance for the assessment of environmental tax reforms (20IT41)" by the World Bank for the Italian Ministry of Economy and Finance (MoEF) and funded by the European Commission. Several modeling novelties are integrated: in addition to the standard features of a single country CGE model, the IRENCGE-DF model is dynamic, using the neo-classical growth specification. Labor growth is exogenous, and capital accumulation derives from savings/investment decisions. Finally, the model has a vintage structure for capital that allows for putty/semi-putty assumptions with sluggish mobility of installed capital. IRENCGE-DF is composed of two modules: i) the environmental part, designed to focus on energy, emissions, and climate change at the national level; and ii) the regional part, built to evaluate shocks at the regional level. In this paper, the regional module of the IRENCGE-DF model is briefly reported. One of the biggest issues in building regional CGE models is the lack of specific regional data for building a proper Social Accounting Matrix (SAM). Our work represents a key contribution to the existing literature on regionalized SAMs: we show a step by step construction procedure for a completely new regional SAM. This matrix includes all economic and fiscal flows so that it represents an important source of information to use as input data for simulating the effect of the abolition of business tax (IRAP) in Italy through the IRENCGE- DF model. The model can be calibrated to a regional SAM that follows a standard set of conventions representing the economic structure.1 The key changes in IRENCGE-DF include a regional specification for Italy (20 regions at NUTS-II level), detailed demographics by age, revised tax equations with additive and multiplicative shifters, a simple debt framework similar to that of transfers to allow an analysis of distributional impacts of borrowing and debt payment, calculation of metric money utility, improved price indices in-demand system, and endogenous capital supply-to-capital stock ratio (i.e., excess capacity in the economy). Our main goal is to simulate how fiscal shocks affect the structure of the Italian regional system, give insights to reform, modernize the tax system, and improve regional economic 1 The model is implemented in the GAMS software, and an aggregation facility is used as a front-end to the model to allow for complete aggregation flexibility. 1 outcomes. In the specific, we simulate the abolition of IRAP (regional production tax) through different scenarios. The IRAP is a local tax on productive activities realized within a regional territory. It was introduced as a financial source for local governments, with a high tax base and homogeneity among the regions. The standard rate is 3.9%, but higher IRAP rates are, for example, applicable to banks and financial institutions (4.65%) and insurance companies (5.90%). Regional authorities have the right to increase or decrease the IRAP rates within the limit of 0.92%. Political debates proposed the abolition of IRAP as one of the ways to reform the Italian tax system. The numerous interventions removed the cost of labor from the tax base and transformed IRAP from a tax on the added value to a tax that affects only the profits and passive interests of the company. In 2019, IRAP guaranteed revenues equal to €24 billion, of which almost 42% derived from public administrations and €600 million from the tax levers operated by the regions. Therefore, the effective standard rate of revenue on the tax base of the private sector to be compensated would be equal to €13.7 billion. In this regard, some hypotheses emerged. Among these, the replacement of the IRAP with an additional IRES (corporate income tax) to be attributed to the regions. In this case, the redistributive effects that would derive from the differences between the two taxes in terms of taxpayers, tax base, and territorial distribution of the tax base should be carefully considered. For this reason, the IRENCGE-DF model seems appropriate to simulate the effects of such tax reform on the Italian economic system. The rest of the paper is organized as follows: Section 3 presents the input data for IRENCGE-DF model by introducing the concept of regional SAM and showing block by block the regional SAM accounts for Italy. Section 4 introduces the main concepts about CGE models and how IRENCGE-DF can contribute to the CGE literature. The new Italian regional SAM is applied to our model in Section 5, where simulations on IRAP abolition are presented. Finally, the main results, strengths, weaknesses, and future developments are discussed in Section 6. 2 Literature Review Computable General Equilibrium (CGE) models reflect, in the general framework, countries’ fiscal and economic situations characterized by interacting sectors and agents. In opposition to country-specific models, it can be interesting to investigate social-economic differences across the administrative units inside countries. This interest is typically justified by the vast differences within the same country. The main goal of regional CGE models is to analyze how different regions respond to a given shock. From this perspective, Italy is a fascinating country for applying a SAM-based regional CGE model, given the high heterogeneity among 2 its regions. A SAM is a comprehensive, economy-wide data framework, typically representing a nation’s economy, linking production, primary factors, and institutions (typically, households, government, and the rest of the world) (Pyatt and Round, 1977; Reinert and Roland-Holst, 1997). Given their explanatory power, SAMs are widely used as data inputs in economics and computable general equilibrium (CGE) models. Consistency of input data ensures that an analysis of shocks and effects is not based on contradictory pieces of information (Cicowiez and Lofgren, 2017). Mainly, a SAM refers to a country’s economy for one year. A SAM is a square matrix in which a row and a column cross represent each account. Each matrix cell contains the payment from its column (expenditure) to its row (revenue) account. The main property of SAMs is to respect the principle of double-entry so that, for each account in the SAM, total revenue (row total) equals total expenditure (column total) (Lofgren et al., 2002). The SAM incorporates three macro balances for government deficit, trade deficit, and savings-investment balance. All SAM-based CGE models must specify how balance is achieved in the three macro accounts and incorporate Walras Law in some form. Usually, the SAM accounts are partitioned, separating endogenous and exogenous accounts so that the endogenous accounts are a function of the values of the exogenous accounts. The partition of the SAM accounts determines the macro “closure” of the model (Robinson, 2006). One of the most critical parts of building such a regional model is the construction of the SAM containing data for each region. This point may be critical due to the lack of data, even if several authors in the literature have proposed techniques to overcome this issue. Research in the regional CGE models presents differences in their structure, but generally, a common aim is to simulate fiscal-economic shocks at a regional level. In the specific case of SAM- based regional CGE models, there is relatively little debate in the CGE community. Naqvi and Peter (1996) developed the MRF (Multi-Regional Forecasting) model. It simulates tax/environmental policy in Australia. Jean et al. (2004) presented Europe’s DREAM- MIRAGE (Deep Regional Economic Analysis Model-Modelling International Relationships in Applied General Equilibrium) model by considering 119 different regions. Canning and Tsigas (2000) built a similar model for eight macro-regions in the United States. Concerning the purposes, there are works with the specific aim of addressing tax reforms (Shoven and Whalley, 1984; Powell and Snape, 1993; Jorgenson and Wilcoxen, 1997; Dixon and Rimmer, 2001), implementation of environmental policies and study of climate change impacts (Darwin and Tol, 2001; Bigano et al., 2008; Eurobarometer, 2008; Skjærseth and Wettestad, 2010), and trade liberalization (Anderson and Martin, 2005; Bouët et al., 2005). In the tradition of Computable General Equilibrium (CGE) models, Mercenier et al. (2016) presented RHOMOLO, which relies on an equilibrium framework à la Arrow-Debreu, 3 where supply and demand depend on the system of prices. Policies are introduced as shocks; the system changes with optimal supply and demand behaviors adjusting, and the allocation and the supporting price system evolve towards a new equilibrium. Therefore, as with all CGE models, RHOMOLO evaluates the interaction effects between all agents through markets, imposing complete consistency. Given the regional focus of RHOMOLO, particular attention is devoted to the explicit modeling of spatial linkages, interactions, and spillovers between regional economies. For this reason, models such as RHOMOLO are referred to as Spatial Computable General Equilibrium (SCGE) models. Each region is inhabited by households aggregated into a representative agent with preferences characterized by a love for variety à la Dixit-Stiglitz (Dixit and Stiglitz, 1977). Transport costs for trade between and within regions are assumed to be of the iceberg type and are sector- and region-pair specific. This framework implies a 5 x 267 x 267 asymmetric trade cost matrix derived from the European Commission’s transport model TRANSTOOLS (see Brandsma et al., 2015). 2.1 The Italian Context Regarding the regional CGE models presented for Italy, Gesualdo and Rosignoli (2013) presented TUSCANI, a single-region comparative-static CGE model for the Tuscan economy. It is a SAM-based model built by the Institute for Regional Economic Planning of Tuscany (IRPET) and refers to 2008. It focuses on the income generation process and re-distribution for the households. In TUSCANI, the regional SAM consists of 37 industries, 54 commodities, and 17 institutional sectors. The households are divided into ten groups by percentiles of gross income. The interregional trades are estimated using a gravity model. This model is consistent with the neoclassical theory. TUSCANI does not explicitly model margins. In 2018, Paniccià and Rosignoli (2018) proposed the most recent IRPET methodology as a NUTS2 level multiregional Supply and Use Table for 2011. This new model updates the previous one by considering more regional data and differentiating multiregional and foreign trade in final and intermediate flows. Fondazione ENI Enrico Mattei (Standardi et al., 2014) proposed an innovative method to estimate bilateral trade flows across sub-national areas and analyze the implications of different assumptions. They used the GTAP model (Hertel, 1997) for Italy and split the Italian economic system into ten sectors and three macro-regions (North, Centre, and South). The regionalization is obtained through a two-step approach: (1) working on a database creating a sub-national SAM for interactions between sub-national areas and with the rest of the world; and (2) working on the model structure, which allows the introduction of different degrees of factors and goods mobility. The model consists of 57 sectors, and the 4 reference year is 2004. For the estimation of trade flows, the authors used data on transport and economic data on exports and imports (Canning and Tsigas, 2000; Dubé and Lemelin, 2005) to overcome the lack of data. Then, a cross-entropy optimization method is applied to make the information inside the model consistent. Finally, the trade flows across sub- national regions are adjusted by the RAS statistical method to increase the consistency of transportation flows with the production data. Cherubini and Los (2016) follow a macro-oriented approach, introducing the results of a first attempt to integrate sub-national input-output tables for Italy into the World Input- Output Database (WIOD).2 In this way, they assess the effects of changes in the structure of value chains on regional employment patterns. The regionalization is applied by considering 4 Italian macroregions (Northwest, Northeast, Centre, and South and Islands). The approach adopted to integrate the Italian IRSUTs into the WIOD system of international SUTs was to replace the tables for Italy as a whole with 4 Supply and 4 Use tables, a couple for each region (the "Four Italies"). As a result, the final database, where each of the "Four Italies" is a country itself, has 44 countries (comprehensive of the "rest of the world") instead of the original 41. Firstly, for each product, the interregional imports and imports from abroad have been divided into intermediate consumption, and capital use of goods and services, according to the proportions observed for the imports in the international Use for Italy as a whole. The international transport margins (ITMs) of each of the four regions when trading goods with foreign countries, as they result from the international WIOD tables, have been distributed according to the proportion of the total regional imports from abroad of each product. More recent works for Italy are contained in Gesualdo et al. (2019) and Severini et al. (2019). The first (Gesualdo et al., 2019) presents Italy’s multisectoral computable general equilibrium tax model. It shows a methodology for modeling and accommodating the full range of direct and indirect taxes. The authors provide a powerful tool for acquiring new insights into fiscal policy analysis through a commodity tax matrix and a production tax matrix. Simulations are performed to evaluate a value added tax rates reform, finding that a budget-neutral uniform tax rate reform would improve GDP and welfare. The second paper (Severini et al., 2019) focuses on gender disparity in Italy, showing another powerful aspect of CGE models. The authors present a multisectoral analysis to bridge gender disaggregation within income formation. A gender-aware SAM is used to evaluate the impact of different fiscal policies to reduce female labor costs. 2 The WIOD database provides annual time series of world input-output tables (WIOTs) from 1995 to 2011, including 35 industries and 59 products. 5 3 A New Regional SAM for Italy Here, some novelties concerning the cited works for Italy are proposed. In particular, the IRENCGE-DF model is built on a regional SAM with the following characteristics: • 20 sectors at Nace Rev.2 level; • 20 commodities by converting through a bridging matrix the 12 COICOP and 10 COFOG commodities included in the national SAM3 ; • 20 Italian regions at Nuts-II level4 ; • 2 factors: capital and labor at the national and regional levels, respectively; • 4 agents: households, enterprises, central government, and local governments; • 10 tax categories: Social Security Contributions (SSC) from employers; property tax (IMU); Personal Income Tax (PIT) at the national, regional, and municipal level; taxes on production; taxes on products; business tax (IRAP); Corporate Income Tax (IRES); taxes on commodities; value added tax (IVA); excises; taxes on imports. At the current version, the Italian regional SAM refers to the year 2016. However, from a computational perspective, one of the key features of IRENCGE-DF is the ability to update the SAM automatically through a cross-entropy approach (Robinson and El- Said (2000)) with a Higher posterior density (HPD) estimator (Britz (2020)). The SAM estimation module is separate from the model code and is run independently. Users can introduce new information (e.g., split any accounts in SAM, update macroeconomic totals or the whole of the macro SAM), and the code balances the SAM automatically. The code also allows for incorporating GTAP data to split the activities. Regionalization of the SAM is done in a separate process and is combined of two components: regionalization and balancing. The first component prepares the proto-SAM using regional data, while the latter balances it. The second component is the standard SAM balancing algorithm cited above (see Section 3.10 for a further description of cross-entropy with HPD estimator). Regional Input-Output coefficients are estimated using Augmented Flegg Location Quotients (Flegg et al. (1995); Flegg and Webber (1997); Flegg and Webber (2000)). We used the 3 The national SAM was built in a previous project between the Italian Department of Finance, the World Bank, and the European Commission under the project "Improving the evaluation of VAT and excise tax policies in Italy." 4 We consider the two autonomous provinces of Trento and Bolzano, which are counted separately in the NUTS-II regional classification, as a single region labeled Trentino Alto-Adige, to be consistent with the Italian regional framework. 6 standard formulation of those estimations. These coefficients are used to calculate the intermediate use part of the regional SAM by multiplying the total supply in the national SAM. The regional tax accounts and value added are introduced afterward by assuming that the residual is the capital value. Output by region and sector is then calculated by summing each activity’s intermediate use, activity taxes, and value added accounts. Investment, government (central and local), household, and export demand for commodities are introduced next. Margins are calculated by disaggregating national margins based on total consumption in the regions. Lastly, when available, commodity taxes are raised using regional tax data (for VAT and tariffs) and total consumption in regions when not (sales and excises). Next, components of household accounts are regionalized. We start by introducing labor income by compensating employees at the regional level as labor income for the households in that region. We assume all compensation paid to labor in a region goes to the households in that region without transfers to the other areas. We use regional data from household surveys (EU-SILC data5 ) for debt payment receipts and lending to government, enterprise transfers, and government transfers. The national accounts are recorded under the region "Italy". National government, debt, capital, investment, savings, and taxes are kept at the national level. We do not have one of those accounts in every region. However, payments or receipts from those accounts can be regional. For instance, direct tax payments by households are regional as households are regional. Once all blocks are calculated, we estimate the flow of commodities for final and intermediate consumption using a non-linear programming model that minimizes the deviation from supply and demand from the gravity estimations given the supply and demand from each region. The remaining imbalances, which are generally very small, are moved to stock changes if the commodity has stock changes in national SAM. If not, they are moved to household consumption by adjusting household savings. As those are pretty small, they do not cause significant changes. The estimated flows from commodity accounts of the importing region to commodity accounts of the exporting region are then moved to household demand and intermediate demand accounts of importing region. So, the SAM treats commodity flows as if households and activities of the importing region are buying commodities from the commodity accounts of the exporting region. We assume the share of each importing region in the commodity flows from each exporting region is equal to the share of household or intermediate consumption 5 EU-SILC is a cross-sectional and longitudinal sample survey, coordinated by Eurostat, based on data from the European Union member states. 7 of the imported commodity in total demand for this commodity in the regions. That is, the imports of commodity c in the region A from region B are distributed across household demand and intermediate use of activities in region A based on the share of households and intermediate use of activities in the total local supply of commodity A. All these calculations result in negative demand for some commodities in some regions by some Armington agents (Armington, 1969). We set those to zero and leave balancing to the SAM balancing algorithm as the imbalances are generally small. Most of the data are from ISTAT (Italian National Institute of Statistics), such as domestic wages and salaries, consumption of general government, final consumption expenditure of households, employment in persons, domestic compensation of employees, gross value added, gross fixed capital formation, and imports/exports. Tax data are administrative data from the Italian Department of Finance (DF) database. As already mentioned, interregional flows are estimated from DF administrative data (REF), and other data are from the national SAM built under the previous project. For the sake of reproducibility, we present in the following subsections a detailed description of calculations for each account of the new Italian regional SAM. 3.1 Regional Market Output The activities row represents the regional market output. The dimensions are activity- commodity-region, resulting in a 400x400 account. We follow three steps: 1. to calculate regional output tables using the demand components: Drc = HHrc + GOVrc + RegGovrc + IOrc + IN Vrc + EXPrc − IndT ax (1) where HH , GOV , and RegGov are the households, government, and local governments’ consumption, respectively. The subscripts r and c indicate the region and commodity, respectively. IC is the intermediate consumption, IN V the investment, EXP the exports, and IndT ax the indirect taxes; 2. to distribute national output (N atOutc ) across the region using shares: Drc RegOutrc = × N atOutc (2) r Drc 3. to distribute regional output among commodity and sectors (a) through a bridging 8 matrix (BMca ) which links commodities to sectors: RegOutrc,rc = RegOutrc × BMca (3) The total regional output, as sum over regions, activities, and commodities is equal to €3,159.8 billion. 3.2 Total Demand The commodity row represents the total demand (or total supply of commodities for use in the economy). Its sum provides a measure of aggregate welfare. It can be formulated as the sum of the following accounts: • Intermediate Consumption calculated as: N atIOac ICrc,ra = AF LQ (RegOutrc − V Ara ) (4) c N atIOac r c where AFLQ is the Augmented Flegg’s Location Quotient, a measure for considering the concentration of a sector and the region’s dimension when there is a lack of data or adjustment is required. The location quotients are here calculated on employment data. N atIOac represents the national input output account, while V Ara is the regional value added by activity sector. Finally, the total national intermediate consumption is equal to €1,637 billion; • Margins calculated for the commodities trade, transport, and energy based on the national demand for the respective commodity. The total national margins results equal to €309.5 billion; • Households final consumption from the estimation of interregional flows through administrative micro-firm VAT data DF estimation. The estimation results show a total national final consumption by households equal to €1,019.5 billion; • Government Consumption calculated starting from 10 COFOG commodities and obtaining 20 commodities through a bridging matrix. The results are reproportioned on the national demand for government consumption. The total government consumption is equal to €282.3 billion; • Local Government consumption follows the same procedure as the national government consumption. The local government consumption account is diagonal in the SAM and 9 the total national amount is equal to €40.2 billion; • Investment demand calculated as: Invrc IN Vrc = N atIN Vc (5) r Invrc where N atIN V is the national investment from the national SAM. Finally, the total national investment demand, as sum over regions and commodities of IN Vrc , is equal to €297,797 billion; • Export, from regional export (ISTAT-ICE from Bank of Italy report REF) re-proportioned on national demand of commodities from the rest of the world. Total regional exports are equal to €497.3 billion. The total demand results equal to €4,083 billion, which is the sum of the accounts listed above. The imbalances in the total demand account are considered as stocks (this aspect will be better described in Subsection 3.10 dedicated to the balancing procedure). 3.3 Value Added Value added is composed of: • Regional labor value added equal to the compensation of labor net SSC and the total amount is equal to €490.6 billion; • Capital value added equals the total value added net regional labor VA, SSC, taxes on production, IRAP, IRES, and IMU. The total national amount is equal to €787.2 billion. As sum of labor and capital regional value added, the total amount for Italy in 2016 is equal to €1,277.9 billion. 3.4 Income We must redistribute the regional labor VA and national capital VA across the agents. • The Households’ income equals the sum of labor income (€490.6 billion), households’ capital income (€396.8 billion), surplus from enterprises to households (€112.2 billion), transfers from Government to households (€329 billion), and interests paid to households (€25.4 billion). The labor income is obtained as the sum over commodities of the Labor 10 VA net SSC. All the other components are calculated by applying Households’ income shares (σhh from DF estimation, see Table 1) to the capital income of households from the national macro SAM (KYhh macroSAM ). The macro SAM is a SAM that includes only macro aggregates. The shares σhh represent the income source for households across the regions (e.g., 7% of income from capital goes to families in Piemonte ). As an example, the household’s capital income KY is the following: capital KYhh = σhh macroSAM KYhh (6) • Enterprises income is the sum of enterprises’ capital income (€342.2 billion) and interest paid to enterprises (€239.2 billion). The first is the enterprises’ capital income from macro SAM redistributed on Gross Operating Surplus (GOS). The GOS is equal to the sum of value added over the regions minus compensation of labor, social security contributions (SSC), taxes on production, business tax (IRAP), and corporate income tax (CIT or IRES). The second, interest to enterprises, is obtained as in Equation 8; • Government income is the sum of depreciation, SSC by employer, personal income tax (PIT) , taxes on production, CIT, taxes on commodities, value added tax (VAT), excises, import tax, and interest payments to the Government. The government capital income is reconstructed from the national macro SAM. • Local Governments’ income is the sum of property tax (IMU), regional PIT, municipal PIT, and IRAP. 3.5 National Taxes In our regional SAM, the Italian national taxes are disaggregated as follows: • PIT at national level net tax for the tax year 2016. The aggregation is conducted on regions and the total national amount is equal to €238 billion; • Production taxes redistributed on value added plus IRAP paid by public administrations. The production taxes are the sum of other taxes on production and negative taxes on production. The total collected production taxes are equal to €34 billion; • IRES (CIT) is regionalized starting from national SAM 2016 data. The total national amount is equal to €30 billion; 11 Labor Capital Surplus Unempl. Child Pensions In-work Interest Piemonte 8% 7% 7% 8% 6% 8% 8% 13% Valle D’Aosta 0% 0% 0% 0% 0% 0% 0% 0% Lombardia 21% 19% 18% 17% 13% 17% 17% 17% Trentino 3% 2% 2% 2% 2% 2% 2% 5% Veneto 10% 8% 10% 8% 7% 8% 11% 9% Friuli 2% 2% 2% 2% 1% 2% 2% 4% Liguria 3% 4% 3% 2% 2% 3% 3% 4% Emilia-Romagna 10% 9% 8% 9% 7% 9% 9% 9% Toscana 7% 8% 8% 6% 6% 7% 6% 7% Umbria 1% 1% 1% 1% 2% 2% 2% 2% Marche 2% 2% 2% 3% 2% 3% 3% 4% Lazio 10% 12% 11% 9% 8% 10% 9% 6% Abruzzo 2% 2% 2% 2% 2% 2% 2% 3% Molise 0% 0% 0% 0% 1% 0% 1% 2% Campania 6% 6% 7% 7% 13% 7% 7% 4% Puglia 4% 5% 6% 7% 9% 6% 6% 4% Basilicata 1% 1% 1% 1% 1% 1% 1% 2% Calabria 2% 3% 3% 2% 4% 3% 3% 3% Sicilia 5% 6% 7% 9% 10% 7% 7% 2% Sardegna 2% 2% 2% 3% 3% 3% 2% 1% Table 1: The Table shows for each Italian region (rows) data of Households’ income shares (DF estimation). These shares represent the income source for households across the regions. The Italian regions are listed from North regions to South and islands (Sicilia and Sardegna) • SSC employer from ISTAT data 2016. The total social security contributions by employers are equal to €176.5 billion; • Taxes on commodities, composed of other net taxes on products from national SAM, are redistributed across regions using the demand for the respective commodity as a share. We took data from the national SAM 2016, which report a total national amount of taxes on commodities equal to €9 billion; • VAT is regionalized using data from the DF microsimulation model on VAT (Cirillo et al., 2021). Only the VAT relating to households is regionalized. The total amount (€102 billion) was re-proportioned using the totals from the ISTAT national accounts (i.e., those reported in the 2016 national SAM); • Excises from national SAM (€59,2 billion) reproportioned on the national demand for the respective commodity; • Taxes on imports are regionalized as regional tariffs and the total amount is equal to €2.2 billion. 12 3.6 Regional Taxes The regional taxes taken into consideration in our model are the following: • PIT at regional and municipal level aggregated by regions for the tax year 2016. The former is equal to €11.8 billion, the former €4.5 billion. Note that, given the regional structure of our SAM, we attribute municipal PIT to regions, although this is actually a tax collected by municipal governments and not by the regional ones. • IRAP (BT) minus IRAP paid by public agents. It is regionalized and includes IRAP calculated on the regional-plant breakdown table for the tax year 2016. The sum over regions brings to a total IRAP collected by regions equal to €13 billion; • IMU (property tax) from F24 revenue data, for the tax year 2016. The result is obtained from the sum of the government share and the municipalities share. In addition, the taxes were partitioned between natural persons and legal persons. The percentage of natural persons has been incorporated into the real estate sector. The ATECO sector already broke down the share relating to legal persons in the original data. As for municipal PIT, even IMU should be attributed to municipalities instead of regions. However, here we attribute the total amount (€20 billion ) to regional governments. 3.7 Savings The savings account is the sum of households, the Government, and the rest of the world’s savings. In the specific, these components are calculated as follows: • The households’ savings are equal to €70.3 billion. As for capital income in Equation 6, they are obtained by applying the share of households’ income from public debt interest (σhh interest ) to the households’ savings from the national macro SAM. The calculation is the following interest macroSAM Sr,hh = σhh Shh (7) • Enterprises’ savings are calculated as enterprises’ capital income net transfers from enterprises to households. The imbalances in the account enterprises’ landing/borrowing are considered savings. The total amount of enterprises’ savings is equal to €234 billion; • Government savings is the sum of government savings from national SAM and imbalances so that the imbalances in the account central government account are considered as savings. The total government savings are equal to €46 billion; 13 • Rest of the world (RoW) savings are taken from national SAM, and imbalances in the account RoW are treated as savings. The total amount is equal to €52.5 billion and are reported with the negative sign in the regional SAM as representing a negative investment. 3.8 Debt The debit account is the sum of households’ debt, enterprises’ debt, central Government debt, and RoW debt. In the specific, each element is calculated as follows: • Households’ debt (€24.8 billion) is redistributed on interest shares (the same process seen in Equation 7). We assume debt payment and borrowing from households are distributed with the same shares; • Enterprises’ debt (€234 billion) is calculated as Sr,ent EntDebtr = EntInterestsmacroSam (8) r Sr,ent where EntInterestsmacroSam are the interest on debt payed by enterprise from national macro SAM. This value is regionalized by using regional enterprises’ savings as share; • Government debts (€414.3 billion) from national SAM; • Rest of the world debts (€146.4 billion) from national SAM. 3.9 Rest of the World We consider the Rest of the World (RoW) to take into account the relationships between Italy and all the other countries. In this context, a detailed description of how imports and exports were treated seems to be essential for replications. • Import of goods: data refer to 2016 and are taken from the ISTAT-ICE yearbook "Commercio estero e attività internazionali delle imprese ", edition 2017. Data refers to imports by economic activity and region. Data on imports of goods describe different sectors linked to the sectors of our interest. For the electricity/water category, we had to split them so that we used additional 2 digit ICE-ISTAT data for the water and electricity sectors. Specifically, the data are not divided between regions for the electricity sector, but we have only a national figure (about €1.6 billion). We 14 distributed this value for the 20 regions using a re-proportioning method through domestic demand. For instance, Piemonte ’s share is equal to the domestic demand for the Piedmont electricity sector, divided by the domestic demand for the Italian electricity sector. The shares of each region were then multiplied by the amount of €1.6 billion to obtain the regional figure for importing the electricity sector; • Import of services: data refer to 2016 and are taken by the ISTAT-ICE report from the Bank of Italy. Data refers to “ debiti per importazioni di servizi per tipologia e ripartizione territoriale ” (i.e., debts on imports of services). These data are reported for macro-regions (northwest, northeast, middle, south, and islands), so we need to regionalize these macro-regions into our 20 Nuts-II levels. For the regionalization of the import of services, the domestic demand is used in the same way it is presented for the import of goods. The initial data are not broken down by macro-areas for the transport category but fall under the item "non-distributable data" (approximately €20 million). For this reason, data on the number of companies in the transport sector whose reference markets are foreign countries were used to distribute the €20 million (data on permanent ISTAT censuses). Once the regional shares were calculated, they were applied to the €20 million to obtain the final regional figure; • Export of goods: as for goods imports, the first step is to split the electricity/water category through 2 digit data. However, the 2 digit data shows €350 million not distributed among the regions. While we solved the import problem using shares calculated on domestic demand, shares on the number of employees by region and sector have now been used for services exports. It is plausible that the value of exports of goods can be approximated using the shares of the number of employees in the same sector within a specific region. In addition to the shares on employees, shares relating to the number of companies with more than ten employees operating within the electricity sector were also used. In this way, a double re-proportioning was carried out to obtain the final data; • Export of services: data from 2016 from the Bank of Italy are used for the export of services. They refer to “ Crediti per esportazioni di servizi per tipologia e ripartizione territoriale ” (i.e., credits on exports). The services reported by the Bank of Italy are grouped as for import of services. The unique difference is that data are regionalized using shares calculated on the number of employees. We used the same procedure for the transport sector as used for the import of services. 15 3.10 Balancing the SAM In almost every step of the regional SAM building procedure, the cross-entropy algorithm is applied as a balancing method with a Highest Posterior Density (HPD) estimation function (Britz, 2020). Therefore, the SAM is balanced through a constrained optimization problem. A set of identities must be fulfilled by the final balanced data set, such as: • closed market balances • macroeconomic accounting identities • various exhaustion conditions • non-negativity (e.g., to exclude negative consumption) The constraints are in the form of (in)equalities and represent additional information so that a Bayesian approach is followed. The penalty function to correct the raw data based on the constraints minimizes the difference between the original and final data set. Finally, the entropy criterion is applied to provide a distance measure between the prior and posterior distributions. Once the balancing procedure is run, we assume that imbalances are due to the estimation of interregional trade for intermediate use plus stock changes. The imbalances in interregional trade are solved through an optimization problem (nonlinear programming). We want to minimize the difference between new and old supply/demand under the constraint of their equality. We obtain the optimal level of trade, and the excess is attributed to margins. The rest of the imbalances in commodities becomes stocks if the investment demand exceeds zero; households’ savings in the opposite case. At the end of this procedure, we apply the cross-entropy algorithm again to balance the final SAM. 4 IRENCGE-DF Model The IRENCGE-DF model is a recursive dynamic computable general equilibrium (CGE) model. For each year, a scenario is solved as a static equilibrium, with dynamic equations linking exogenous factors (such as population growth and capital accumulation) across years and updated equations for productivity factors. Each static equilibrium relies on a relatively standard set of equation specifications. Production is modeled using a series of nested constant-elasticity-of-substitution (CES) functions designed to capture the substitutions and complements across the different inputs. 16 Figure 1: Production nest This outstanding feature is embodied in a vintage capital structure that captures the semi- putty/putty relations across inputs with more elastic long-run behavior than the short- run. The model allows for both multi-input and multi-output production. The former, for example, would allow for electricity supply to be produced by multiple activities-thermal, hydro, solar and other renewable forms of electricity production. The latter allows for a single activity to produce more than one product. For example, oil seed crushing makes vegetable oils and oil cakes (for feed). Labor and capital income are primarily allocated to households with pass-through accounts to enterprises. Government revenue is derived from both direct and indirect taxes. The production nest of the model is synthesized in Figure 1. Household demand is modeled using the constant-differences-in-elasticity (CDE) demand function, the standard utility function used in the GTAP model. The model allows for a different specification of demanded commodities from supplied commodities. A transition matrix approach is used to convert demanding goods to supplied goods that rely on a nested CES approach. The current version’s transition matrix is mainly diagonal, with consumed commodities directly mapped to supplied commodities. Final demand is handled similarly, though the aggregate expenditure function is a CES function rather than the CDE. Goods are evaluated at basic prices with tax wedges. The model incorporates trade and transport margins that add an additional wedge between basic and end-user prices. The trade and 17 transport margins are differentiated across transport nodes-farm/factory gate to domestic markets and the border (for exports), and from port to end-user (for imports). Import demand is modeled using the ubiquitous Armington assumption, i.e., goods with the same nomenclature are differentiated by region of origin. It allows for imperfect substitution between domestically produced goods and imported goods. The level of the CES elasticity determines the degree of substitutability across regions of origin. Domestic production is analogously differentiated by destination region using the constant-elasticity- of-transformation (CET) function. The CET elasticity level determines producers’ ability to switch between domestic and foreign markets. The model allows for perfect transformation so that the law-of-one price must hold. Market equilibrium for domestically produced goods sold domestically is assumed through market clearing prices. The small country assumption is assumed by default for export and import prices. Thus they are exogenous, i.e., export levels do not influence the price received by exporters, and import demand does not affect (CIF) import prices. The model does allow for the implementation of an export demand schedule and an import supply schedule, in which case it would endogenously determine the terms of trade. The current version of the model assumes market-clearing wages in the labor markets with the possibility of an upward- sloping labor supply schedule and sluggish labor mobility across sectors. Future works could readily implement more labor market segmentation (for example, rural versus urban) and some form of wage rigidity. In dynamic simulations, new capital, i.e., new investments, is allocated across sectors to equalize the rate of return. Old capital remains installed in its original industry unless the sector declines. A declining sector is one in which potential supply, as measured by the capital/output ratio, exceeds ex-post demand. This can occur from various shocks that lower the need for a specific commodity. If a sector declines, it releases its installed capital using an upward-sloping supply schedule, and its ex-post return on capital is less than the economy-wide average. Old capital in expanding sectors earns the same rate of return as new capital. The dynamics of IRENCGE-DF are composed of three elements: (i) population and labor stock growth is exogenous-the latter is often equated to the development of the working-age population; (ii) the aggregate capital stock grows according to the overall level of saving (enterprises, households, public, and foreign). Still, it will also be influenced by the investment price index and the depreciation rate; (iii) service labor productivity is assumed or dynamically calibrated to achieve a per capita growth target. Labor productivity in other sectors is calculated relative to labor productivity in services using a linear schedule that allows for both multiplicative and additive components. This aims to calibrate inter-sectoral labor productivity to historical trends (domestic or international). 18 The coding of the model is relatively independent of the dimensionality of the SAM and other functional dimensions of the data. The input SAM has a dimensionality of is × is, and most of the remaining sets and subsets derive from is. Sectors have three classifications: a, c, and k , respectively (production) activities, marketed commodities and consumed commodities. In a traditional model, the three sets are identical. In the IRENCGE- DF model, with its multi-input multi-output production structure, output from activities a is combined with imports to supply (or ’produce’) commodities c. This allows multiple activities to produce a single commodity (for example, electricity) and having single activities produce multiple commodities. In addition, IRENCGE-DF allows for commodities in final demand (indexed by k ) to differ from marketed commodities c. A consumer-based ’make’ or transition matrix maps consumed commodities to supplied commodities. It allows for more realistic demand behavior in the context of an energy model. For example, household demand for transportation services can combine the demand for fuel and automobiles. If the fuel price increases, the combined demand for fuel/autos will decline. It also allows for specific treatment of the demand for fuel and intra-fuel substitutability. 5 Simulations In this section, we present the simulations regarding the abolition of IRAP. This exercise is justified by the political debate around potential reforms of the Italian fiscal system. The necessity of macroeconomic stability after the financial and sovereign debt crises, and, more in general, the harmonization of the tax system, leads to the will to reform local taxes. Italian fiscal federalism guarantees local governments fiscal independency. However, this concept is conditioned by the progressive emptying of the IRAP tax base. Consequently, the reduction of revenues imposed rebalancing of funding sources to the benefit of VAT sharing, which cannot be managed by the regions and constitutes a transfer. The corresponding payments at standard rates are supplemented by state transfers to guarantee the total financing of regional health needs within the National Health Fund. The Chamber approved the draft law for the revision of the tax system, to be implemented through one or more legislative decrees (A.C. 3343-A, sent to the Italian Parliament on 29 October 2021). The gradual abolition of IRAP was discussed as one of the critical changes of the fiscal reform. However, a long debate has been opened on how to replace IRAP and guarantee adequate resources for financing the health system. The literature suggests that taxes levied on immovable property or consumption are less distortionary and, thus, less harmful to economic growth than those imposed on corporate or labor income (Mankiw et al., 2009; Slemrod, 1990). In that regard, there appears to be significant scope for shifting 19 taxes to more growth-friendly revenue sources in Italy (IMF, 2020). A potential hypothesis could be the substitution of IRAP with the additional tax on IRES (CIT) to regions. In this case, it would be appropriate to carefully consider the redistributive effects deriving from the differences between the two taxes in terms of taxpayers (broader audience in the case of IRAP), the tax base (IRAP taxes profits and interest expenses, IRES only profits), and the territorial distribution of the latter. For these reasons, it seems appropriate to verify the effects of this tax reform on the Italian economic system through the IRENCGE-DF model. Here, three different scenarios are simulated: 1. IRAP-baseline: Complete abolition of IRAP tax-financed by public deficit; 2. IRAP-IRES: Tax shift between IRAP and IRES with a uniform tax rate; 3. IRAP-VAT: Tax shift between IRAP and VAT. The simulation of each scenario with the IRENCGE-DF model is straightforward. We already have seen that the model has clear national and regional tax streams. Moreover, depending on the closure rule, the tax rate may be endogenous to achieve some target. To simulate the first scenario, we set to zero the IRAP tax and allow for a higher public deficit. In the second simulation, the Government revenue is fixed to the baseline scenario level, and we set the IRAP tax to zero. Next, we endogenize the IRES tax schedule to achieve the fiscal target using a uniform shift in the marginal tax rate. The model solves for a unique tax multiplicative adjuster applied to all activities. Finally, we simulate the third scenario using the same logic for the VAT instead of the IRES tax. In the following figures, we present some results regarding these simulations. First, we look at the macroeconomic impacts on GDP in 2030 (end of time horizon for all simulations) regarding percentage change with respect to the baseline scenario. Figure 2 also illustrates the GDP dis-aggregated into its components: consumption, Government expenditure, investment, and net export. All three considered scenarios have a positive impact on GDP. The IRAP- baseline scenario (complete abolition of IRAP tax) has the highest impact on GDP. Figure 3 shows the same effects on GDP but for all the time horizons. The effects observed in the IRAP-baseline scenario are the highest: in 2018, there is more than a +0.35 percent increase in GDP compared to the Business as Usual (BaU) scenario. That increase is slowly decayed but persistently higher than in the other two simulations. Overall, we could consider the IRAP-IRES simulation has the one having no effect on GDP growth compared to the baseline scenario. The IRAP-VAT scenario, on the other hand, starts with a slightly negative effect 20 0,80% 0,60% 0,40% 0,20% 0,00% Consumption Government Investment Export Import GDP Expenditure -0,20% -0,40% -0,60% -0,80% irap_bas irap_ires irap_vat Figure 2: Macroeconomic impacts on GDP and its components in 2030, measured as percentage change with respect to the baseline scenario, in all considered simulations. 21 0,40% 0,35% 0,30% 0,25% 0,20% 0,15% 0,10% 0,05% 0,00% 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 -0,05% irap_bas irap_ires irap_vat Figure 3: GDP growth from 2018 to 2030, measured as percentage change with respect to the baseline scenario, in all considered simulations. 22 Figure 4: Regional Value Added, measured as percentage change with respect to the baseline scenario in 2030, for all considered simulations. on GDP in the first period (the year 2018). After a couple of years, it starts having a positive and, over time, increasing impact on GDP. As already pointed out, the IRENCGE-DF model is the appropriate tool to simulate these scenarios since they all involve territorial distribution considerations. In particular, we focused on the effects of Regional Value Added and welfare. Figure 4 shows the percentage change variation in regional value added at the end of the time horizon (the year 2030) with respect to the Baseline scenario. We adopt a red-green diverging scale, with midpoint zero set to white color. On the left, we have the IRAP-baseline scenario: by the abolition of the IRAP tax, only four regions (Sardegna, Valle d’Aosta, Basilicata, and Trentino Alto Adige ) would experience a negative impact on their value Added. The highest positive results would be observed in Lazio, Lombardia, and Marche regions. The uniform tax shift between IRAP and IRES would exacerbate the adverse effects observed in the IRAP-baseline scenario and, in many other regions (for a total of 13 out of 20), would harm production. Finally, under a tax shift between IRAP and VAT, the positive and negative effects on production observed in the IRAP-baseline scenario are mitigated. In a few cases (Lombardia and Puglia regions), the impact on production has a reversed sign: overall, we have five regions experiencing a negative effect on production. 23 Figure 5: Welfare effects, measured as percentage change with respect to the baseline scenario in 2030, for all considered simulations. Figure 5 tells us another part of the story: welfare effects in terms of household monetary utility functions (percentage change with respect to the baseline scenario, as before). Let us look at the map on the left. We see a positive welfare effect on all regions, which is no surprise since the abolition of IRAP would give households across regions a higher disposable income. Hence, they all would be better off. However, it is worth noticing that under the IRAP-IRES scenario, almost all the regions that would have experienced a negative impact on production in the same scenario (13 out of 20) also would have households worse off (10 out of 13). Regarding the IRAP-VAT scenario, we see that among the regions with a negative impact on production (5 out of 20), two of them (Lombardia and Trentino Alto Adige ) would experience a positive effect in terms of welfare. Few regional households (Piemonte and Umbria ) would be slightly worse off among the regions with boosted production. The general picture tells us that - if it were compatible with fiscal objectives - the complete abolition of IRAP would have the highest positive effects on economic growth and welfare. 24 6 Discussion The development of the IRENCGE-DF model represents an example of a new regional and environmental CGE model and constitutes a novelty in the comparative literature. IRENCGE-DF is dynamic and, using the neo-classical growth specification, treats labor growth as exogenous and capital accumulation as the result of savings and investment decisions. The vintage structure for capital allows for putty/semi-putty assumptions with sluggish mobility of installed capital. The model grows on the ground of a regionalized SAM that we described account by account in Section 3. The Italian regional SAM represents our input data, and its main strength is the easy procedure we implemented to update data and calibrate them. In this solid framework, it is possible to create a picture of the Italian economic and fiscal system so that we can run simulations. The actual debates in Italy are directed at improving the regional economic outcomes through a possible modification of IRAP (the business tax). This framework justifies our proposed application. In Section 4, we simulated three different scenarios on IRAP: abolition, tax shift with IRES (corporate income tax), and tax shift with VAT (value added tax). This exercise represents an example of how the IRENCGE-DF model can be used and, more in general, how CGE models can help simulate economic-fiscal shocks in a specific regionalized economy. In the first scenario (IRAP abolition), we allow the government to compensate for its loss with a higher public deficit. The impact on GDP is positive (+0.35% in 2018) and the highest if we compare it with the other scenarios. In regional terms, the abolition of IRAP causes adverse effects on regional value added for Sardegna, Valle D’Aosta, Basilicata, and Trentino Alto Adige. On the contrary, Lazio, Lombardia, and Marche would benefit most from this scenario. Looking at the welfare effects in terms of households’ monetary utility functions, the abolition of IRAP would positively affect all regions (i.e., the households would benefit from IRAP abolition because of the higher disposable income). In the second scenario (tax shift IRAP-IRES), we endogenized the IRES tax schedule using a uniform shift in the marginal tax rate. In addition, we applied a multiplicative tax adjuster to all activities. The effect on GDP looks constant over time and is not significant. Regarding regional effects, this scenario exacerbates the adverse effects related to the first one, and the production falls for 13 out of 20 regions. The same effects are observed regarding welfare in almost all regions. Finally, in the third scenario (tax shift IRAP-VAT), the same procedure as the second scenario was followed, with the unique difference that here we treated the VAT instead of IRES. The impact on GDP is initially negative (in 2018), but after a couple of years, it has a positive and increasing impact. Regarding regional welfare, the effects look 25 mitigated with respect to the previous scenarios. For Lombardia and Puglia the impact on production has a reversed sign. In general, five regions show negative effects on production. Among these regions, Lombardia and Trentino Alto Adige would experience a positive effect in terms of welfare. Piemonte and Umbria, two regions with boosted production, would be slightly worse off than the others in the same category. Based on the needs of governments and policy-makers, this model can represent a starting point to evaluate fiscal and economic shocks in Italy, with time specification and regional/environmental data. In future research steps, the authors will work on improving the quality of data used as input by regionalizing environmental data through statistical or heuristic procedures. In addition, future versions of the model will be directed in testing closures other than the classic Walrasian one as sensitivity analysis for the Italian regions study case. Finally, the inclusion of information on the regional tax gap in the Italian regional SAM and in the IRENCGE-DF model would make a great contribution to a better representation of the Italian economic system and more complete policy analyses. References Anderson, K. and Martin, W. (2005). Agricultural trade reform and the doha development agenda. World Economy, 28(9):1301–1327. Armington, P. S. (1969). A theory of demand for products distinguished by place of production. Staff Papers, 16(1):159–178. Bigano, A., Bosello, F., Roson, R., and Tol, R. S. (2008). Economy-wide impacts of climate change: a joint analysis for sea level rise and tourism. Mitigation and Adaptation Strategies for Global Change, 13(8):765–791. Bouët, A., Bureau, J.-C., Decreux, Y., and Jean, S. (2005). Multilateral agricultural trade liberalisation: The contrasting fortunes of developing countries in the doha round. World Economy, 28(9):1329–1354. Brandsma, A., Kancs, d., Monfort, P., and Rillaers, A. (2015). Rhomolo: A dynamic spatial general equilibrium model for assessing the impact of cohesion policy. Papers in Regional Science, 94:S197–S221. Britz, W. (2020). Comparing three penalty functions-cross entropy, quadratic and linear loss–in sam splitting applications. Canning, P. N. and Tsigas, M. E. (2000). Regionalism, federalism, and taxation: A food and farm perspective. Technical report. Cherubini, L. and Los, B. (2016). Regional employment patterns in a globalizing world: A tale of four italies. Global value chains: new evidence and implications, pages 213–235. Cicowiez, M. and Lofgren, H. (2017). Building macro sams from cross-country databases: method and matrices for 133 countries. World Bank Policy Research Working Paper, (8273). Cirillo, C., Imperioli, L., Manzo, M., et al. (2021). The value added tax simulation model: Vatsim-df (ii). Technical report. 26 Darwin, R. F. and Tol, R. S. (2001). Estimates of the economic effects of sea level rise. Environmental and Resource Economics, 19(2):113–129. Dixit, A. K. and Stiglitz, J. E. (1977). Monopolistic competition and optimum product diversity. The American economic review, 67(3):297–308. Dixon, P. B. and Rimmer, M. T. (2001). A wage–tax policy to increase employment. Australian Economic Review, 34(1):64–80. Dubé, J. and Lemelin, A. (2005). Estimation expérimentale des flux d’échanges interrégionaux par la méthode de minimisation de l’entropie croisée. Canadian Journal of Regional Science, 28(3):513–536. Eurobarometer, S. (2008). Attitudes of european citizens towards the environment. European Commission, 295. Flegg, A. T. and Webber, C. (1997). On the appropriate use of location quotients in generating regional input-output tables: reply. Regional studies, 31(8):795–805. Flegg, A. T. and Webber, C. (2000). Regional size, regional specialization and the flq formula. Regional Studies, 34(6):563–569. Flegg, A. T., Webber, C., and Elliott, M. (1995). On the appropriate use of location quotients in generating regional input–output tables. Regional studies, 29(6):547–561. Gesualdo, M., Giesecke, J. A., Tran, N. H., and Felici, F. (2019). Building a computable general equilibrium tax model for italy. Applied Economics, 51(56):6009–6020. Gesualdo, M. and Rosignoli, S. (2013). Building a computable general equilibrium model on a regional social accounting matrix: the case of tuscany. Hertel, T. W. (1997). Global trade analysis: modeling and applications. Cambridge university press. Jean, S., Laborde, D., et al. (2004). The impact of multilateral liberalisation on European regions: a CGE assessment. CEPII Paris. Jorgenson, D. W. and Wilcoxen, P. J. (1997). The long-run dynamics of fundamental tax reform. The American Economic Review, 87(2):126–132. Lofgren, H., Harris, R. L., and Robinson, S. (2002). A standard computable general equilibrium (CGE) model in GAMS, volume 5. Intl Food Policy Res Inst. Mankiw, N. G., Weinzierl, M., and Yagan, D. (2009). Optimal taxation in theory and practice. Journal of Economic Perspectives, 23(4):147–74. Mercenier, J., Martinez, M. T. A., Brandsma, A., Di Comite, F., Diukanova, O., Lecca, P., Lopez-Cobo, M., Monfort, P., Persyn, D., Rillaers, A., et al. (2016). Rhomolo-v2 model description: A spatial computable general equilibrium model for eu regions and sectors. Technical report, Joint Research Centre (Seville site). Naqvi, F. and Peter, M. W. (1996). A multiregional, multisectoral model of the australian economy with an illustrative application. Australian Economic Papers, 35(66):94–113. Paniccià, R. and Rosignoli, S. (2018). A methodology for building multiregional supply and use tables for italy. Studi e Approfondimenti. Powell, A. A. and Snape, R. H. (1993). The contribution of applied general equilibrium analysis to policy reform in australia. Journal of Policy Modeling, 15(4):393–414. Pyatt, G. and Round, J. I. (1977). Social accounting matrices for development planning 1. Review of Income and Wealth, 23(4):339–364. Reinert, K. A. and Roland-Holst, D. W. (1997). Social accounting matrices. Applied methods for trade policy analysis: A handbook, pages 94–121. 27 Robinson, S. (2006). Macro Models and Multipliers: Leontief, Stone, Keynes, and CGE Models, pages 205–232. Robinson, S. and El-Said, M. (2000). Gams code for estimating a social accounting matrix (sam) using cross entropy (ce) methods. Technical report. Severini, F., Felici, F., Ferracuti, N., Pretaroli, R., and Socci, C. (2019). Gender policy and female employment: a cge model for italy. Economic Systems Research, 31(1):92–113. Shoven, J. B. and Whalley, J. (1984). Applied general-equilibrium models of taxation and international trade: an introduction and survey. Journal of Economic literature, 22(3):1007–1051. Skjærseth, J. B. and Wettestad, J. (2010). Making the eu emissions trading system: the european commission as an entrepreneurial epistemic leader. Global Environmental Change, 20(2):314–321. Slemrod, J. (1990). Optimal taxation and optimal tax systems. Journal of economic Perspectives, 4(1):157–178. Standardi, G., Bosello, F., and Eboli, F. (2014). A sub-national cge model for italy. 28