An empirical Stock-Flow Consistent regional model of Campania

We develop an innovative Stock-Flow Consistent macroeconometric regional model with five sectors, exploiting economic and financial statistics for Campania, covering the period 1995 – 2018, and propose a methodology to close the financial account of the private sector when financial data are lacking. The model is then used to perform medium term Economic Policy Scenario Analysis. We find that a debt-funded fiscal expansion has permanent positive effects on growth, with an impact multiplier above one and a medium-run multiplier of 0.71. In the case of a balanced-budget rule the same increase in government spending has still positive effects on growth – with a medium-run multiplier of 0.6 – but adverse ones on the private corporate sector.


| INTRODUCTION
Since unification, the Italian economy has been characterized by strong regional disparities.Over the last 160 years, the development of the Italian economic system has been characterised by a dual pattern of growth, emphasised by significant regional inequalities between the Centre-North and the South (the so-called Mezzogiorno).Some convergence took place during the "economic miracle" (from 1950s to 1970s) fostered and led by a particularly incisive set of public policy programmes, mainly carried out through the Cassa per il Mezzogiorno (Graziani, 1975(Graziani, , 1979(Graziani, , 2000;;Iuzzolino et al., 2013).From the 1980s onwards, and at a higher pace since the introduction of the common currency, the process inverted.In fact, the regional divide has progressively increased and even more so since the outbreak of the global financial crisis of 2008, as shown in Figure 1.
Drawing on the Italian economic divide, Krugman (1991) coined the concept of Mezzogiornification of Europe in order to identify the deep changes in the production structure and the growing concentration of industries set in the central areas of the European Monetary Union (EMU) at the expenses of the peripheral ones, with significant implications for countries' trade balance positions.Mezzogiornification mainly implies that economic dualism which characterised the Italian economy is re-emerging, shaping the relationships between central and peripheral countries of the EMU, far away from the convergence path theorised by the European Commission (1990).Therefore, the debated Questione Meridionale offers important insights for stressing the role of the State in leading the catching-up process at the national and European level, through a systemic industrial and innovation policy able to rebalance the economy towards a new path of economic growth and social development (Realfonzo, 2008).
In this framework, the Italian government authorities promoted in 2017 a plan for large investments 1 aimed at reducing the chronic economic divide and at promoting the convergence process among the main regions of the nation.Central to the investment plan was the realisation of Special Economic Zones in southern regions, and Campaniawhich is the main contributor to GDP in Mezzogiornowas the first region to apply. 2   Economic policies mayand this is usually the casehave idiosyncratic effects within the same country, due to disparities between regions (Aiello et al., 2019;Bachtler & Begg, 2018;Laurent & Mignolet, 2009;Pellegrini, Terribile et al., 2013).Therefore, a useful tool for policymakers, economists, and practitioners for tracking how policy changes affect particular areas is represented by regional models.Indeed, there is a large body of literature of General Computable Equilibrium (CGE) and Input-Output (I-O) models analysing regional economic systems.As we will see, in most of these models the financial sector is left aside: banks and money, credit and debt, financial assets and F I G U R E 1 Real GDP Source: ISTAT.Notes: percent change in real GDP.2007 = 100 portfolio allocation are usually absent, notwithstanding their importance and centrality in shaping economic growth in highly financialized systems.
Pioneered by Wynne Godley, the Stock Flow Consistent (SFC) approach integrates a post-Keynesian analysis of real markets with flow-of-funds analysis of balance sheets and a tobinesque approach to portfolio choice, providing the appropriate structure to study contemporary economies, with all the links between a complex "real" and "financial" side explicitly addressed.However, there has not been any attempt to develop a regional model adopting the SFC approach yet.
The aim of the present work is to fill this gap, by developing the first Stock-Flow Consistent macroeconometric regional model, exploiting regional data for Campania from the Italian National Statistics Institute (ISTAT) and Bank of Italy (BoI), covering the period 1995-2018.We discuss how to build a consistent system of accounting identities to accurately describe the economic system at hand, and propose a methodology to "close" the financial accounts when complete data at the regional level are lacking.We expand the three-balances New Cambridge model originally developed by Godley into a five institutional sector structure, to better disentangle the different roles played by each in the determination of demand and output, investment in real assets and the portfolio allocation of financial assets and liabilities, across sectors and within their balance sheets.
The model consists of 66 equations, twelve of which are behavioural equations estimated over annual data for the period 1995-2018, aimed at performing medium-term economic policy Scenario analysis.
The rest of the paper is structured as follows.Section 2 presents the literature review, focusing on the SFC approach (Section 2.1) and on existing empirical regional models for Italy (Section 2.2).In Section 3 we introduce our database and discuss the accounting structure of the model-describing the Transaction and Balance sheet matrices for Campania.Section 4 presents the main equations of the model, while in Section 5 we simulate the model, performing different shocks, and discuss its properties.Section 6 concludes.

| LITERATURE REVIEW
In recent decades, two different classes of macroeconomic models have emerged as the main tools of analysis for governments, central banks, and major international institutions, each with its pros and cons (Pescatori & Zaman, 2011).The first type is represented by structural models, 3 which are built starting from the fundamental principles of economic theory, often at the expense of the descriptive capacity of the model.The main ones are currently constituted by dynamic stochastic general equilibrium models (DSGE). 4The second typology is represented by largescale models, which are a hybrid between structural modelsas they make extensive use of economic theory for the definition of the relationships between variablesand purely econometric models (such as VARs)as they are built starting from a large block of equations directly derived from national accounts data. 5

| The stock-flow consistent approach
The basic principles of the Stock-Flow Consistent approach 6 can be dated back to the 1970s and 1980s with the independent works of Wynne Godley (and the New Cambridge School) 7 on one side, and of James Tobin (and the New Haven School) 8 on the other.Originally developed to provide a robust policy tool based on actual data, its consistency requirements, as we will see shortly, are the same as those embedded in the System of National Accounts (SNA, European Commission et al., 2009) and Flow of Funds tables, and so allows for a systematic treatment of whole economies.SFC macroeconomic models focus on the financial side of the economic system, and on the interdependencies that connect the balance sheets of the various institutional sectors to their real transactions in a monetary production economy, with an explicit role for banks and financial institutions.
Interest in the SFC methodology surged in recent times, driven by two main reasons.The first one, which is more academic, was that the major contribution of Godley and Lavoie (2007) which provided a systematic account on the construction of increasingly complex theoretical modelsclarified that this class of models are flexible enough to host various theoretical views and to discuss how modern economies work. 9The second, related this time to policy, is that econometric models following the approach have been effective in providing timely warnings of coming recessions and of increasing financial fragility and instability, 10 as well as producing more accurate macroeconomic projections relative to traditional models. 11  Referring the interested reader to Godley and Lavoie (2007) and the mentioned surveys, Zezza and Zezza (2019) summarize the requirements of Stock-Flow Consistent models: • Horizontal consistency -"everything comes from somewhere and goes somewhere" (Godley & Lavoie, 2007, p.6), i.e., there cannot be any black hole.This implies that any source of funds for a sector is a use of funds of one or more other sectors, that any surplus matches a deficit or that the imports of a country embody the exports of others, possibly identifying who-to-whom relations.
• Vertical consistency -each transaction should be recorded once in the current account of the sector involved (payments/receipts), and at least once more as a change in the assets/liabilities of that sector (credit/debit).This means that wages received by household are recorded both in the current account as income receipts, and in the balance sheet as a change in its deposits (assuming it is not used to extinguish debts or acquire financial assets).
The two preceding principles imply that every transaction needs a quadruple entry in the accounting structure, as in Copeland (1947), assuring the consistency of the system.
• Flows-to-stocks consistencythe end-of-period value of any real or financial stock at current prices is given by the accumulation of the relevant flows during the period, plus the possible capital gains due to changes in asset prices.This introduces path-dependence, which plays an important role in SFC dynamics.
• Balance sheet consistency -the liabilities of a sector are the asset of another (possibly matching creditors and debtors).This means that the overall net wealth of the system sums up to zero.This principle applies both to changes in the balance sheets (flow of funds) and to the end-of-period stocks.
• Stock-to-flows feedback -increases in the stocks of financial liabilities imply higher future payments from one sector (debtor) to another (creditor).Taking these flows properly into account reinforces the path-dependence characteristic of SFC models.
As in the System of National Accounts, the accounting of SFC models rely on the same sequence of matrices for describing an economy (Zezza, 2015): going from flow accounting (production; distribution of income; use of income) to flow of funds (which give details of changes in real and financial assets for each sector), to the revaluation account (which measures changes in the value of stocks due to fluctuation in market prices), to balance sheets (which measure end-of-period net real and financial wealth for each sector).
The accounting framework of SFC models is represented by the Transactions Matrix (TM) and the Balance Sheet Matrix (BSM).The Transaction Matrix represents all the main monetary flows that take place in each period, for the entire economic system.Sources of funds are conventionally registered using a positive sign, while uses of funds have a negative sign.The net lending of the sectori.e., the end-of-period financial positionis given by the difference between sources and uses of fundsi.e., the difference between saving and investment.Recalling the first accounting principle, horizontal consistency implies that flow and uses of funds for each transaction sum up to zero, while vertical consistency involves that the sum of each column of the matrix is zero.The Balance Sheet Matrix describes the allocation of real capital and financial assets across institutional sectors.Assets are denoted using a positive sign, whereas liabilities and net worth are given a negative sign.The stock consistency requires that the sum of each row equals to zero.
It should be noted that these accounting consistency requirements should be respected by any macro modelirrespective of the theory behind its construction.Moreover, next to the previous five accounting principles, other requirements come out of logical consistency: • Flows of capital incomes (i.e., interest payments and receipts, dividends etc.) should be endogenously determined from the accumulated stocks of financial assets/liabilities.Thus, the flow of interest payments on debt S at time t would be given by the relative interest rate r over the opening stocki.e., on the end-of-previous period stock.
These quasi-identities usually introduce non-linearities, affecting the trajectories of the model during simulations and reinforcing path-dependency.
• Stocks of financial assets/liabilities should feed-back on the behaviour of at least one sector.For example, positive saving implies accumulation of real and financial wealth, while the values of the stocks must in turn be relevant for consumption/saving decisions, introducing stock-flow norms 12 that constrain the dynamics of the model.
While accounting consistency is important per se for building a sound macroeconomic model, since it reduces the degrees of freedom and provides some important insights about the constraints faced by any economic system, it is not enough.As shown long ago by Taylor (2004) and Taylor and Lysy (1979), the conclusions that can be drawn from a model are primarily led by the direction of causality the author imposes over the variables, in other words, its closures.From this standpoint, the SFC literature has always grounded itself within the boundaries of post-Keynesian economics. 13Thus, it is effective demand that drives economic growth both in the short and in the long-run, with output driving the adjustment and inflation being the result of wage-bargaining processes, as in Kalecki (2013).
Given the k accounting identities that come out of the Transactions and Balance Sheets matrices, 14 if we want to determine n endogenous variables we need n À k additional equations.These are given by specifying how agents in the system determine and finance their expenditures and net borrowing positions, and how they allocate their wealth.Finally, another set of behavioural equations are needed to model productivity growth, wages, and inflation, as well as to the define the behaviour of the Public Sector and the Monetary authority, when needed.
In theoretical models, the short-run equilibrium of the system is driven by price adjustments in financial markets while changes in output ensure that savings are equal to investments.Changes in variables that rule expenditure or portfolio decisions and/or the departure of stocks or other variables from their target levels at the end of the period will determine further adjustments in following periods.The long-run equilibrium steady state is reached when stock-flow ratios are stable: thus, sectors' reactions to changes in stocks and stock-flow ratios in the short-run drive the system towards the long-run steady state.

| Review of models of Italian regions
The SFC literature has so far focused on either whole-country models, or theoretical multi-country models, showing that the integration of the real and financial markets of two (or more) open economies will generate different results than other traditional models.What is missing, however, are the applications of the Stock-Flow Consistent approach to the analysis of a single regional economyand this is the goal we set for ourselves in the present work.Moreover, since Campania is smallwhen compared to the rest of the countryusing a multi-regional structure would have greatly increased the complexity of the model, without altering our main conclusions.
The same cannot be said for the neoclassical and New-Keynesian schools, where theoretical and empirical contributions dealing with the modelling of oneor moreregional systems abound.
Although characterised by different degrees of sophistication, all the dynamic models presented in Table 1 suffer from the same two shortcomings.The first is related to the closure of the models, as long-term growth is ultimately determined only by supply.Despite the introduction of increasingly realistic adjustment mechanisms, which slow down the movements of relative input prices, these are only present in the short term, thus eliminating the possibility of hysteresis effects and, more generally, of accounting for the impact of demand developments on long-term growth.
The second is that these models only deal with real markets.There is no mention to banks or financial sector, credit, money, or financial assets.Moreover, the equations for the accumulation of real capital stocks are usually absent. 16  Alongside these shortcomings, which apply tout court to most I-O models, there is the one specifically inherent to static analysis.It is surely important to assess how economic policies can have idiosyncratic effects on the industrial structure, or how these policies affect the value chains in which firms are embedded or the terms of trade.However, the limit of static simulations is that although they can show the "ending point"i.e., the long-run equilibrium positionsnothing can tell us about the "traverse".This assumes that the parameters or calibrations are stableas well as the behaviours of the economic agents consideredand that the adjustment to long-run positions is rapid. 17

| DATA AND ACCOUNTING STRUCTURE
To develop a macroeconomic model that ensures all the criteria of the Stock-Flow Consistent approach are met, and uses national accounts data correctly, the first step to take is setting up the Transaction and Balance Sheet Matrices.
For Italy, regional statistics are compiled by multiple sources.ISTAT collects the statistics for the Territorial Economic Accounts, which form the core of our database, covering the period 1995-2018.Financial data are instead collected by the Bank of Italy, with longer time series for the stock of loans   3.1 | The transaction and balance sheet matrices In models for an entire country, the SFC approach 18 suggests starting from the description of the Balance Sheet Matrixi.e., how real and financial asset and liability stocks are distributed among the institutional sectors.This is not possible in a regional model, as the Bank of Italy does not produce complete statistics on sectoral balance sheets with territorial detail.The strategy adopted for the construction of the model was thus to start by defining the structure for the Transaction Matrix (Table 2)using data from Istat territorial accounts.
As we said, the representation of the accounting structure of SFC models resembles the national accounts and Flow of Funds tables.Indeed, the TM is organized exactly in the same way: in the columns we distinguish five sectors (households, firms, government, other regions, and rest of the world), with current and capital account, and in the rows the transactions between sectors, with all variables measured at current prices.
The first row of the TM records 19 the component of GDP, namely consumption (C), investment in fixed capital (I) and inventories (DINV), government expenditures (G), net imports from other regions (NM or ), and imports (M w ) and exports (X w ) to the rest of the world.
The next three rows, together with the first column, detail the sectoral distribution of incomes generated in production between wages (W), profits (P), and net indirect taxation (NINDT).We assume that part of the profits is distributed to households.
The second block of the matrix records the distribution of primary incomes, detailing the flows related to capital incomes (payments/receipts of interest, dividends, and other capital incomes).As the table shows, we have assumed that all financial relationships between households, firms, banks, and the public authority are held with financial entities located in other Italian regions.While this may seem like a rather risky hypothesis, it should be realistic in the Italian context.When a customer in Campania, whether it is a family or a business, turns to a bank to ask for a loan or deposit her savings (or invest them), in most cases this is done through a financial institution based outside the Region, particularly in the Centre-North.
The third block records the distribution of secondary incomes, i.e., the redistribution following the payments of direct taxes (TAX)on households' incomes (INCTAX), and firms' profits (DTAX)social contributions (SOCCON), pensions (PENS), and other transfer (OT).
Next, we have the components of expenditures (consumption, investment, and government expenditures) and trade.The difference between sources and uses of funds for each sector determines its Net Lending/Borrowing (NL) position.Finally, the lower part of the TM records the Flow of Funds, i.e., how changes in net lending translate into changes in the stocks of deposits (D), loans (L), and other financial assets (OFA), inside the balance sheet of the relative sector.
For the Balance Sheet Matrix, the choice on the level of detail has been reduced by the limited amount of data available.As mentioned above, the only historical series long enough to enter directly into the model is that on the stocks of loans to the different sectors (L), while the one for monetary deposits (D) is only available from 2011 to 2020.The most important aspect, however, is the absence of data on the rest of the financial portfolio and of its allocation between the institutional sectors.Following Albareto, Bronzini, Carmignani, and Venturini (2008), we used the information present in the Financial Accounts of the Institutional Sectors (FAIS), 20 together with the available territorial historical series, to build quotas in national aggregates, and use them to project regional series backwards using the appropriate techniques.The Balance Sheet Matrix for Campania is presented in Table 3.
We split the stock of real capital between the private and public sectors, differentiating between real estate (KH), machinery (KM) and non-residential construction (KNR). 21With respect to financial stocks, we assume that the share of mortgages on the total loans of households is the same as for the rest of the country.As we do not have territorial statistics on the rest of financial assets and liabilities, how can we build a complete set of accounts for our financial sector?To do so, we build a residual category of "Other Financial Assets" (OFA), using the principles behind Notes: (+) sign stands for "sources of funds", (À) sign stands for "uses of funds".
i.e., once the changes in assets and liabilities (in our case, loans and deposits Δ L ð Þ and Δ D ð Þ, respectively) have been determined.In symbols: Once created this flow variable and having assigned an initial value for the stock, 22 one only needs to cumulate the flows to compute the stock.This also ensures that the accounting consistency of the model is respected.
Changes in the net financial positions of the sectors, therefore, result in changes in their stocks of assets and liabilities, which generate higher or lower future flows of capital incomes in the following periods, adding a further interaction between real and financial markets.
Although simpler in its structure than other existing empirical SFC models, 23 this work is a useful simulation tool for analysing economic trends in the region and ascertain the effects of exogenous shocks, sophisticated enough to answer multiple policy questions and to perform different scenario analysis.The model consists of 66 equations, twelve of which are determined by stochastic equations estimated on annual data for the period .The model tracks the evolution and dynamics of the main components of GDP, financial balances, and portfolio allocation, the impacts of public expenditure on private sector productivity and investment, the effects of economic policies on the regional production structure and consumption behaviour, as well as the performance in trade.
In the next section we will briefly introduce the equations of the model and describe the main transmission channels.For the econometric specification, we adopted a general-to-specific approach, taking care of the order of integration of each variable in the model: starting from a general model in term of dynamics, we tested the restrictions on nonsignificant parameters to find the final parsimonious model.All stochastic equations have been tested for autocorrelation, heteroskedasticity and normality of residuals, and parameter stability.We must stress, nevertheless, that the short size of our sample at annual frequency prevented us from the adoption of more advanced estimation techniques, from which model dynamics would certainly benefit.The size of our sample also prevented the adoption of proper tests for parameter stability.With longer time series, of course, these problems may be partly overcome.
Large-scale models are typically used to perform macroeconomic policy forecasts and to conduct scenario analysis on the possible effects of such policies.The model presented here is indeed developed to analyse the effects on the

| Real transactions
As in the first row of the Transaction Matrix, GDP (Equation 1) is given by the sum of demand components.These are households' consumption and their housing investment (C and I h ), firms investment (in machinery and warehouses, and inventories, I f and DINV f ), Government expenditures and investment (G and I g , respectively), and exports to the rest of the world (X w ), minus net imports from other Italian regions (NM or ) and the imports from the rest of the world (M w ).Since we do not have the data for all demand components at constant prices, we will determine all model variables at current prices and use the GDP deflator 25 to compute real values (Equation 2).
The GDP deflator (p gdp , Equation 3) is estimated through a stochastic equation which links inflation to wage dynamics, including a shift variable for 2001, which is statistically significant, and shows that the ratio collapsed when the country fully joined the monetary union.For the average wage (WAGEU, Equation 4) we find a wage-curve where the rate of growth in average wages is positively driven by the growth in inflation (proxied by the GDP deflator, with a coefficient of 1) and the level of participation rate in the labour force (PARTRATE), and negatively by the growth in the unemployment rate (UR).Finally, we find evidence for a sort of Kaldor-Verdoorn effect, as the productivity level (PROD, Equation 5) positively depends on the share of private sector value added in GDP VApvt GDP and the level of the average wage in real terms WAGEU p gdp .Given the short size of our sample, which prevents the adoption of more complex dynamic specifications, and given the contemporaneous feedbacks among prices, wages, and productivity, we chose to estimate Equations 3 to 5 with an IV estimator.Estimation results are shown below.Production is undertaken by private and public enterprises.Regional value added (VA, Equation 6) is computed by subtracting from GDP the tax revenues on production and imports, net of subsidies to firms (NETTAX, Equation 7), which are computed by applying an ex-post tax rate (τ nettax ) to GDP.Private sector production (VA pvt , Equation 8) is determined by subtracting the public sector share (VA g ) from total value added.
We then pass to functional distributioni.e., how incomes generated in production are split between the different institutional sectors.Profits (P, Equation 9) are computed as the difference between GDP and the sum of wages (W) and net indirect taxes NINDT gt ð Þ , with firms' profits (P f , Equation 10) computed as a residual, after subtracting households' share P h ð Þ. Wages (Equation 11) are determined by employment (EMP) and the average wage, while employment (Equation 12) is determined by the relation between real GDP and average labour productivity.Finally, net indirect taxes are computed through an average tax rate on production (Equation 13).Obviously, the sum of the components is equal to GDP, as in the first column of the Transaction Matrix.
The third block of Table 2 records the flows of capital incomes of households and firms.Households' outlays (KYP h , Equation 14) are the sum of the interest they paid on their existing loans (INTP h , Equation 15) 26 and of the other transactions in capital account (TRKP h ).Capital incomes received (KYP h , Equation 16), are in turn the sum of the interest incomes receivedon their stocks of deposits and other financial assets 27 (INTR h , Equation 17)dividends (DIVR h ) and other capital incomes (TRKR h ). 28Since we do not have the relative time series in the territorial accounts, interest paid by firms on their loans are computed by the model (INTP f , Equation 18) as well as the receipts on their deposits (INTR f , Equation 19), based on the BSM.
Households pay direct taxes on their primary income 29 (INCTAX, Equation 20) and social contributions on their income from production (SOCCON, Equation 21)both computed from implicit average tax rates.They also receive pension payments (PENS, Equation 22)based on the number of retired people 30 and the average wageand other net current transfers (OCTN), which are instead left exogenous.Firms pay direct taxes on their profits, determined as well through an implicit average tax rate (DTAX, Equation 23).
Reading top-down each column of the TM, we reconstructed the disposable income of the various sectors.For households, disposable income is the sum of incomes from production, capital incomes and pension received, net of taxes paid on their incomes, social contributions, and other transfers in current account (YD h , Equation 24).For firms, it is the sum of profits and capital incomes, net of taxes paid (YD f , Equation 25).For the Public sector, finally, is given by the sum of tax revenues from direct and indirect taxation and social contributions, net of pension payments and of the transfers made to households (YD g , Equation 26).
Following Godley and Lavoie (2007), Muellbauer (2016), and Zezza and Zezza (2020), we model consumption (Equation 27) as a function of disposable income 31 and the opening stock of wealth Notice that this is coherent with a dynamic process of adjustment toward a stable stock-flow norm between household income and wealth.In this way we add another feedback channel in the modelif household see their wealth increasing (in housing and financial assets), consumption will adjust accordingly until reaching a new, stable, stock-flow ratio.We found the presence of a structural break relative to the financial crisis, which we took care of through a shift variable, while we included a dummy for 2002 to eliminate autocorrelation among residuals.Government consumption and investment are left exogenous, as autonomous decision of the public authority.
As we said in the introduction, we do not have time series relative to interregional trade, but only net (total) imports.However, since COEWEB releases data relative to imports and exports of goods from/to the rest of the world, we can decompose net imports from other Italian regions (NM or ).As in Aiello and Pupo (2003) and Buran et al. (2006), net imports from other regions are estimated as a function of private sector (non-residential) final demand and the level of regional GDP, relative to that of other regions GDP GDPor , Equation 28).With respect to international trade, the growth in imports from the rest of the world (M w , Equation 29) is estimated again as a function of private sector (non-residential) final demand, while exports to the rest of the world (X w , Equation 30) are estimated with an Error Correction Model as a function of world demand, with a short-run elasticity of 0.87 and a long-run one of 0.32.For the first two equations, we found the presence of a structural break related to the financial crisis, which we took care of through a dummy variable, which is always statistically significant.We now turn to the determination of gross fixed investment of the various sectors.In the model, we have three different real assets: homes (KH), machineries (KM), warehouses/infrastructures (KNR).Following Byrialsen and Raza (2020), Godley and Lavoie (2007), and Zezza and Zezza (2020), investment in housing is determined by a stochastic equation which links the growth rate in investmentrelative to the existing stock of homes Ih KHtÀ1 , Equation 31)to disposable income and the interest rate on mortgages (r LMO ).Firm's investment is determined by a stochastic equation which links its growth raterelative to the existing capital stock If KMtÀ1þKNRtÀ1 , Equation 32)to their disposable income and the changes in loans

KMtÀ1þKNRtÀ1
. Notice that in both cases the specifications are consistent with a process of adjustment to a stable stock-flow ratio of capital to output.Once netted out sectoral investment from saving, we finally get to the net lending/borrowing of the various sectorsi.e., their financial balances.These are represented by the column totals, and must respect the fundamental identity, 32 i.e., that the sum of net lending for all sectors is zero.In Godley's terminology, these are labelled as "Net acquisition of Financial Assets" (NAFA, of households and firms, , the Public Sector Borrowing Requirement (PSBR, usually called Government Deficit, Equation 35), and the Current Account Balance (CAB, vs other Italian regions and the rest of the world, .
Figures 2 and 3 display, respectively, the net lending/borrowing position of each sector and the aggregate financial balances.A couple of things stands out.
In the years prior to the financial crisis, households were getting into debt and firms were reducing their accumulation of assets.At the same time, the public sector was trying to reduce (nation-wide) the debt-GDP ratio by cutting spendingthus reducing its deficit.This implied a higher exposition with the foreign sector, in this case represented by the other Italian regionssince the CAB with the rest of the world has been balanced until 2009.In 2009, private sector incomes collapsed, and the households' NAFA turned positive.After the Sovereign debt crisis, the private sector started again to accumulate financial assets, with firms leading the waywith their NAFA increasing up to 10 percent of GDP in 2014.As the public sector deficit did not increase substantiallyso that the adjustment took place in the private sectorthe aggregate CAB against the foreign sector narrowed, from À7% of GDP in 2012 to À2% in 2017.However, to the surge in firms NAFA corresponds a drop in their investment and the diversion of more funds through the financial sectori.e., towards Centre-North regions.

| Stocks and flows
Among the peculiarities of the SFC approach, a prominent spot is occupied by the ability to integrate the spending and investment choices of the different sectors with the accumulation of real and financial capital stocks, consistently with the BSM presented above.The stocks of real capital (K i,j )for sector i and stock j, at time tevolve depending on the relative flows of investments (I i,j ), given the cost of replacing capital.In symbols: Net financial wealth (NW i , Equations 39-40) for sector i is instead given by the end-of-previous-period stock of wealth plus the current net lending/borrowing position.If a sector is in a net lending (borrowing) position, it is either accumulating (selling) financial assets or extinguishing (increasing) debts.Recall that we assume that all financial intermediation is done through banks located in other Italian regionsthus we add the net lending position of all our extended foreign sector to the opening stock of wealth.
While the variation in wealth ultimately depends on the end-of-period financial position of each sector, the decision regarding how to allocate their financial investment between the different assets, or the possibility of taking on new loans, depends on sectoral portfolio choices.
In SFC models à la Godley-Lavoie, the coefficients for portfolio choices are usually determined through a relative rate of returns matrix, as in Tobin (1969).When dealing with real world statistics, however, it is difficult to estimate from the data (given their structure, the available time span, the presence of structural breaks, etc.) the appropriate relations-if they exist-between the relative rate of returns and the demand and supply for different assets and liabilities.In their model of the Italian economy, for example, Zezza and Zezza (2020) only find statistical evidence for a negative relationship between the rate of return on Public debt and foreign assets, and between banks' obligations and banks' sharesinside households' portfolio structure.In our case, given that financial statistics at the territorial level are lacking and that our model does not have that many assets to determine, we choose to adopt some of the principles of the SFC approach to model asset allocation, consistently with the Balance Sheet Matrix presented in Table 3 and post-Keynesian theory.
On the asset side of households' balance sheet, we find monetary deposits at banks (D h ) and other financial assets (OFA h ), while they take on mortgages (LMO) and loans for consumption expenditures (LC).The demand 33 for deposits is determined as a fixed share of savings (Equation 41).As mortgages represent the largest source for funding investment, households' demand for mortgagesrelative to income -is estimated as a function of investment in housing, the changes in assets value and the relative interest rate (Equation 42).The demand for consumer credit (in real terms) depends instead on real consumption, along with the changes in real assets Δ DhþOFAh p gdp used here as a proxy for collateral availabilityand the real interest rate on short term credit (r LC Á Δp gdp , Equation 43). 34Finally, the stock of other financial assets is determined as a residuali.e., as the difference between the end-of-period stock of wealth and the other components of their balance sheet (OFA h , Equation 44).On firm's asset side we have deposits (D f , Equation 45), determined as a fixed share of wages paid.Firms' demand for loansrelative to GDP Δ LF ð Þ GDP , Equation 46)is determined as a function of their net lending and the outstanding stock-to-GDP ratio LF

À Á
, meaning that firms will finance their investment through credit if their own funds are not sufficient, assuming banks accommodate their demand for credit.Again, the stock of other financial assets is determined as a residual (OFA f , Equation 47).

| SIMULATIONS AND MODEL PROPERTIES
In this Section we will make use of the model developed above to perform economic policy Scenario Analysis.Before running the simulations, we shall underline once more that for this class of models, the validation is given not only by the ability to replicate the data, 35 but also by the realism of out-of-sample dynamics.This implies that, although it may be better to use simpler models for short-term forecasting (such as VARs), these dynamic structural models are the only one able to describe the traverse toward medium-run equilibrium positions, thus providing a completely different tool for Policy Analysis.Moreover, the SFC approach enables to ascertain the effects of these policies on sectoral financial behaviour and balance sheet dynamics, which are usually not considered in standard SEM at the regional level.
The model is solved for the period 2001-2017, and we then project our exogenous variables up to 2034 in the future, assuming a stable growth path. 36  When building a baseline scenario, different strategies can be chosen.Notice that, in our case, the primary objective of the simulations presented here is not to forecast the economy by pinpointing the future short-term growth rates, but rather to explore the structural linkages between financial and real side of the economy in the short to the medium run for the region under analysis.This choice was primarily led by the fact that in 2020 Campania has been hitas the rest of the world -by the Covid19-Pandemic.This most certainly has affected aggregate behaviour, and time series will show structural breaks as more observations become available.As we do not have yet statistical information at the regional level related to the Pandemic period, we found it sensible to run simulation exercises not including the Covid-shock.
We simulate four different Scenarios, comparing the effects of distinct exogenous shocks on the regional economy over the medium run.To begin with, we test the effects of different fiscal shocks.In Scenario I, we simulate the impact of a deficitfinanced expansion of Government expenditure equal to 1% of GDP.In Scenario II, instead, to ascertain the effects of a balanced-budget rule for the Government sector, the same increase in expenditure is accompanied by an increase in the indirect tax rate on private sector production, so that the Net Lending of the Government (e.g., its deficit) goes to zero.
We then pass to standard external shocks.In Scenario III, we increase the growth in the level of world demand, while, in Scenario IV we simulate the effect of a credit crunch, i.e., a 1% increase in interest rates on private sector borrowing.Table 4 summarizes the overall results, alongside the baseline.Figures 4 to 7 detail the outcomes of the simulations.More specifically, Figure 4 shows the effects of the four exogenous shocks on the components of demand; Figure 5 displays the effects of the shocks on Financial Balances, i.e., the aggregate net lending/borrowing positionsrelative to GDP; Figure 6 presents the evolution of financial stocks in private sector balance sheet; Figure 7 shows the dynamics in labour market, for the four simulation scenarios.

| Deficit-financed expansion
In Scenario I we permanently increase government expenditures by 1% of GDP, equal to 107 million euro.The increase in spending leads to an immediate increase in nominal GDP of 108 millionwith a short-run multiplier just above 1which translates in an increase in growth rate equal to 0.1%, compared to baseline, driven by rising private sector output.As functional distribution does not change, and government does not increase tax rates on production, the increased spending immediately pushes private sector wages and profits, with revenue from production increasing in the following periods due to rising economic activity.
In the first year following the shock, the increase in household income leads to higher consumption, while firms accumulate all their extra-profits.The private sector financial surplus initially increases household wealth, which is also assumed to be a collateral for mortgages, so that the demand for mortgages also increases, while firms initially reduce their debt burden (i.e., extinguish their loans).
The current account balance towards both other regions and the rest of the world immediately deteriorates, as the demand for exports coming from the rest of the world does not change, while importsfrom both other regions and the rest of the worldincrease, due to the higher private sector final demand.
Notice also that the increase in spending generates a twin deficit: as the regional government deficit is financed by the central authority, the current account balance against other regions deteriorates, with the private sector returning to balance over the medium-run.
When households and firms start to invest the CAB further deteriorates, pushing GDP down in the following years.As household investment increase, their financial position turns negative, as they continue to accumulate mortgage debts, which further increases their interest burden.Notice that to every increase in net interest paid corresponds a higher current account deficit towards other regions, which also contributes to reduce domestic output.
By the end of our simulation sample, however, household net wealth is higher than in the baseline, due to the robust increase in monetary holdings and other financial assets, which more than compensate the increase in liabilities.
Firms, in turn, start to accumulate financial assets, increasing their real financial wealth with respect to the baseline.The higher private sector holdings of financial assets imply higher payments from other regions to domestic sectors, so that the CAB starts to recover somewhat after 3 periods.As private sector incomes increase, moreover, the government deficit starts to fall, due to higher revenues from taxes on wages and profit.
The increase in economic activity boosts wages, and productivity, while the effect on employment is relatively small 38 (Figure 7, upper-left quadrant).
By the end of the simulation sample, the increase in spending led to an increase in nominal GDP of 76.6 million eurowith a medium-run multiplier of 0.71.
Notice that, for this class of demand-led models, when simulating increases in single components of demand results are symmetric, meaning that carrying on the same exercisedecreasing instead government spendingwould generate the same results with inverted signs.Thus, decreasing spending will lead to a short-run decrease in GDP equal to the change in spending (as the impact multiplier is just above 1), and a 0.71 negative change in the medium-run.

| Balanced-budget rule
In Scenario II we ascertain the effects of a balanced-budget rule for the Government sector.Contrary to Scenario I, in this case the same increase in expenditure of 1% of GDP is accompanied by an increase in the indirect tax rate on private sector production, so that the Net Lending of the Government (e.g., its deficit) goes to zero in the year of the shock.The increase in spending leads here to an immediate increase in nominal GDP of 86 millionwith a short-run multiplier of 0.8which translates in an increase in the growth rate equal to 0.08%, compared to the baseline, driven again by rising private sector output and government expenditures.
Contrary to Scenario I, however, as the government levies higher taxes on production, this reduces profits of households and firms, while wages increasewith respect to the baselinebut less than in Scenario I.As household wages increase more than the fall in their profits, consumption increases in the first period, followed by housing investment.In subsequent years, both consumption and housing investment fall with respect to baseline, because of the higher tax rate, and recover only after 6 and 15 periods, respectively.To protect their purchasing power, households initially sell financial assets and use their monetary depositswhose stock slightly fall relative to baseline.
Demand for mortgage loans follows the dynamics of housing investment.
As for the following periods, the effects of the shock on the household sector and on the labour market are similar to Scenario I, although everything is scaled down, as the impact on incomes is lower due to the fall in profits.
Households' net financial wealth increases, as the lower investmentrelative to the baselineimplies a lower demand for loans, which indeed turns negative five periods after the shock.The positive net lending translates into higher demand for deposits and financial assets, whose stocks rise, leading to higher interest receipts, that further reduce the decline in income in the following periods.
Firms face a permanent drop in their profits.In the first year following the shock, part of the fall in profits generates a permanent reduction of their investmentwhich stays below the baseline level throughout the simulation horizon.The rest of the fall translates into a financial deficit: firms sell their financial assets, to reduce the impact on incomes, and increase their demand for loans.The increase in debt continues in the following periods, implying higher interest payments on their loans which are not counterbalanced by receipts on assets.
The increase in economic activity leads to a deterioration of the current account balance toward other regions, pumped also by higher financial payments from firms, while the CAB against the rest of the world improves, as imports fall due to the lower firms' investment (which drastically reduce non-housing private sector final demand).
As shown in Figures 4-7 and Table 4, in Scenario II the "second twin" is represented by the private (corporate) sector: as the government balances its budget, the CAB deficit is matched by a decline in the financial position of the private (corporate) sector, which increases its indebtedness towards other regions and reduces its reliance on foreign products.By the end of the simulation sample, nominal GDP is 65.1 million above the baseline, with a medium-run multiplier of 0.6.

| Scenario III: Increasing world demand
In Scenario III we increase the growth rate of world demand by one percent with respect to the baseline.This implies an increase in exports equal to 90.4 million on impact, which generates an increase in nominal GDP relative to baseline of 91.9 million, with a multiplier of 1.02.By the end of the simulation sample, exports and GDP are 54.1 and 40.2 million above the baseline respectively, implying a medium-run export multiplier of 0.73.The rate of growth in GDP eventually stabilises at 0.07% more than in the baseline by the end of the simulation.
The increased domestic output generates higher incomes for the private sector, so that household consumption rises, followed by housing and firms' investment in following periods.Imports from other regions respond to rising economic activity, but the net effect on the overall current account balance is positive, due to the fall in net imports from the rest of the world.Private sector net wealth rises significantly, as the increase in loans is more than offset by the surge in monetary holdings and the accumulation of other financial assets.Finally, the increase in output generates a sustained increase in productivity and a timid one in wages, which brings a small reduction in the unemployment rate.

| Scenario IV: Interest rate shock
As we said, in our last simulation exercise we check how the regional economy responds to an exogenous shock in the cost of borrowing for households and firms.To do so we impose a 100b.p. increase in the interest rates on mortgages, consumer credit, and loans to firms.
Figure 7 reports the effect of the shocks on the labour market, for our four Scenarios.
As expected, the interest rate hike generates a recession in the short run.
The immediate drop in consumption (of 26.7 million) brings down imports from both other regions (À10.1) and the foreign sector (À2.9), so that the reduction in nominal GDP is only of 13.6 million.
In the following periods also the other components of demand start to fall, especially housing investment, which peaks (downwards) three years after the shock and recovers somewhat afterwards.As interest rates rise, the private sector diverts its funds away from investment and asset accumulation towards consumption expenditures, while using the remaining funds to reimburse debts.In particular, the large drop in housing investment strongly reduces the demand for mortgages, whose stock collapse.The recession causes a strong and permanent reduction in wages and productivity, and a slight increase in the unemployment rate.By the end of the sample, GDP is still growing 0.01% less than in the baseline.
This suggests that in the case of exogenous shocks to the cost of borrowing, the public authority should intervene to sustain investment, providing transfers/tax reductions for the private sector.

| CONCLUSION
The debate on the North-South economic divide has been widely discussed by the literature.Recently, the public debate focused again on the role of Mezzogiorno in shaping the economic recovery, through the implementation of a set of policy interventions aimed at strengthening the economic growth of the Southern regions and at reducing the chronic gap with the Centre-North.
In this line, regional models stand as a noteworthy analytical tool for evaluating the effectiveness and impact of national (or local) policies in enhancing economic growth in a specific area.Even though there is a large literature of empirical works in the neo-classical and New-Keynesian traditioneither SEM or IO models, with varying degrees of sophisticationin these models the financial sector is usually left apart.
The financial sector, and its interconnections with the real economy are instead central to the Stock Flow Consistent (SFC) approach, which integrates a post-Keynesian analysis of real markets with flow-of-funds analysis of balance sheets, and a tobinesque approach to portfolio allocation.However, there has not been any attempt yet to develop a regional model adopting the SFC approach.
In the present work we provided the first SFC macro-econometric regional model, which represents the opening building block of this literature.We developed a five-sector structural SFC model of Campaniathe largest region of Italy's Mezzogiornousing data from ISTAT and the Bank of Italy relative to flows of incomes and expenditures, and stocks of financial assets and liabilities.The model is made of 66 equations, twelve of which are estimated over annual data for the period 1995-2018.
We provided a useful methodology to close the financial accounts for which data is unavailable and keep the accounting consistent.The model links the investment and saving decision of the institutional sectors to their balance sheets, and describes how this wealth is allocated, within each sector and across the different assets.
We performed four simulation exercises, to analyse how the model reacts to different fiscal (deficit-spending and balanced-budget) and exogenous shocks (world demand and interest rates).In Scenario I, we found that increasing government expenditures through a deficit financed expansion has a short-run multiplier of 1 and generates an increase in medium-run real GDP of 0.71, fifteen periods from the shock.Conversely, in Scenario II we checked how a balanced-budget rule for the regional public authority changes the result above, so that in this case, the higher expenditures are accompanied by an increase in the indirect tax rate.While the effect on GDP is only slightly lower than in Scenario 1with a short run multiplier of 0.8 and a medium run one equal to 0.6the distributional impact is completely different.The higher taxes on production generate a fall in the profit share, so that firms' investment permanently fall relative to baseline, while there is a surge in corporate indebtedness, which further deteriorates the current account balance.
In Scenario III, we analysed the effects in changes in world demand on the regional economy.As expected, the increase in exports to the rest of the world leads to an increase in the growth of real GDP in both the short-run and medium-run, which leads to a sustained rise in private sector financial wealth.Finally, in Scenario IV we checked how the model responds to changes in financing conditions for private sector borrowing.We found that increasing interest rates on private sector loans has recessionary short-run effects, largely due to the collapse of housing investment.As aggregate demand recovers so does the growth in real GDP, which however remains lower than the baseline level throughout the simulation horizon.This suggests that the public authority should intervene to sustain investment, providing transfers/tax reductions for the private sector.
Although simpler in its structure relative to other empirical SFC modelswhich however deal with whole countriesthis work is the first example in this literature for analysing economic trends at the regional level, sophisticated enough to address the impact of diverse policy questions and perform different medium-run macroeconomic policy scenario analysis.Moreover, the structure of the model can be easily enlargedor stretchedto include new blocks (the production block, a more sophisticated labour market, but also an "environment" block), provided one keeps the accounting consistent, and respects the principles of the SFC approach.
The model allows to treat the real and the financial sides of the economy in an integrated way -properly accounting for their interdependencies -and assess the impact of demand developments on growth.The approach proposed here has never been adopted at regional level and adds dynamics and an active role for the financial sector to classical large-scale econometric models.The main limitation of the work is represented by the lack of available financial data, which prevent us to build a more complete balance sheet.More accurate data regarding the financial stocks of the private sector at the territorial level, as well as non-financial accounts for every institutional sector, would greatly help to enrich the detail of this class of models.
A promising avenue in this sense would be to integrate the data from Bank of Italy relative to the Household Wealth Survey at the territorial level, which however are not publicly available.
Finally, the model would benefit from longer time-series, which would make possible to adopt more advanced econometric techniques to estimate behavioural equations.

ACKNOWLEDGMENT
Open Access Funding provided by Universita degli Studi di Roma La Sapienza within the CRUI-CARE Agreement.
[ 2 Special Economic Zone (SEZ) are a widespread instrument adopted by public authorities to boost economic development in particular areas (Akinci & Crittle, 2008;Ambroziak & Hartwell, 2018;Crane et al., 2018;Jensen, 2017;Wang, 2013;Zeng, 2016).The model presented herewhich is a simplified version of the one presented in Canelli et al. (2021) was originally developed to analyse the effects of the implementation of a SEZ in the Campania, and to simulate the effects on the regional private sector of the implementation of economic policies by the Public Administrations.
6 For a comprehensive survey of the SFC literature, see Nikiforos and Zezza (2017) and Caverzasi and Godin (2015).
8 See Backus, Brainard, Smith, andTobin (1969) Brainard andTobin (1968) and Tobin (1969Tobin ( , 1982)). 9The SFC approach has indeed been used to cover a broad variety of theoretical issues in post-Keynesian economics: an early SFC perspective on financialization is in Lavoie (2008) and a recent one is in Gimet, Lagoard-Segot, and Reyes-Ortiz (2019), while Sawyer and Veronese Passarella (2017) focus on monetary circuits; Dos Santos and Zezza (2008)  The NMODS is a New-Keynesian structural model where supply determines long-term growth, while demand effects are exhausted in the short term.The model consists of five distinct blocks of equations, which describe the determinationfor each macro-areaof: prices, supply and value added, components of aggregate demand and distribution of income, wages and employment, and, finally, the trade block.Moreover, to increase the interpretative capacity of its scenario analysis, SVIMEZ has also estimated the Public Sector accounts for all Italian regionsconsistently with the indications coming from the EU for the computation of public sector fiscal balances.
There are also several mono-regional models.The Multi-Sectoral Model (MMS) developed by Prometeia is the most widely used by the Italian regions.The MMS offers a rather complex representation of the functioning of a regional economy and can be applied to carry out different types of analyses.It allows monitoring the regional economic trend, performing medium-long term simulations, and evaluating the impact of public policies.The model has reached the 4.0 version, characterised by higher level of disaggregation, made possible by the availability of more detailed regional information (time series and I-O tables).
Starting from southern regions, the first contribution we find is represented by the Multi-Sectoral Model of the Sicilian region. 42The model, originally developed for the impact assessment of POR funds for the period 2000-2006, is still used for macro-economic forecasts and as an analysis tool to simulate the impacts of policy choices from the Statistical Service of the Regional Administration board.A similar approach was adopted by the Department of Economics and Statistics of the University of Calabria, which developed the MOMACAL model. 43The latter presents a multi-equational and multi-sectoral structure that allows accounting for theoretical elements of both neo-Keynesian and neoclassical inspiration.The model assesses the effects of structural and cohesion funds on the regional economic system.
Following a similar pattern, the Autonomous Province of Trento set up the Econometric Multi-sectoral Model of Trentino (MEMT) that integrates the econometric approach based on time series analysis with the intersectoral approach based on I-O (Podestà, 2010).Traditionally, in these dynamic models, the components of final demand are determined by means of econometric equations estimated on the historical series of national accounts, while the levels of production and added value are calculated by applying an input-output table of the economy to the levels of final demand.The model is made up of more than 100 stochastic equations and over 400 identities, which allow replicating the dynamic characteristics of the local system.It provides a disaggregation of production activities into 19 branches (one for agriculture and construction, eight branches of industry and nine of public and private services).
Household consumption expenditures are divided into six expenditure items, as well as disposable income of household accounts for six components.MEMT is used to outline short-medium term predictive scenarios concerning the main economic variables of the province of Trento, and to develop medium-term forecast scenarios, evaluating the consequences of economic policies implemented at local level.
We also have to mention the Remi-IRPET macroeconometric multi-sectoral model -a dynamic approach based on an input-output core structure developed for assessing the macro-economic effects and the medium-long term impacts of public policies.It was used, among others, by Apulia and Tuscany regions. 44  Another tool is represented by the Italian macroeconometric model developed by the Centro Europa Ricerche 45 Profits of firms Profits of households Wagesprivate sector Total employment Net indirect taxes -government Household capital incomes paid Interest paid by household Capital incomes received by household Interest received households Interest received by firms Interest paid by firms Primary income: household Primary income: firms Primary income: other regions Direct tax on households' primary income Direct tax on firms' profits Social contributions Pension payments Disposable income: households Disposable income: firms Disposable income: government Saving: households Saving: government Investment of firmsmachinery Investment of firmsnon-residential construction Investment of government -non-residential construction Investment of governmentmachinery Net lending position: households Net lending position: firms Net lending position: private sector Net lending government Net lending position: rest of the world Net lending position: other regions Stock of real capital: housing Stock of real capital: non-residential construction Stock of real capital: machinery Stock of real capital: public Net wealth: households Net wealth: firms Net wealth: other regions Change in deposits: households Other financial assets: households Change in deposits: firms Other financial assets: firms Unemployment rate Labour force Stochastic equations Investment in housing Firm's investment Firms' demand for loans Demand for consumer credit Households' demand mortgages Imports from the rest of the world Exports to the rest of the world GDP deflator

=
the SFC methodologyand national accounting.The changes in Other Financial Assets (Δ OFA ð Þ), in fact, are nothing less than the difference between the net financial position in the period and the change in net wealth over the same period T A B L E 2 Campania.Transaction Matrix Rest of the World.
regional private sector of the implementation of economic policies by the Public Administrations aimed at reducing the North-South divide.It was therefore necessary to extend the New Cambridge model by splitting the private sector between households and firms and adding other Italian regions to the foreign sector.This allowed us to obtain a five sectors structurewith households, firms, Public Administration, Other Italian Regions, and the Rest of the Worldwhich reduces to a three-sector modelwith private, public, and foreign sectors.The equations follow the presentation of the accounting structure of the previous section, with the addition of stochastic equations that close the model.24

F
I G U R E 2 Sectoral Financial Balances Source: author's calculations.Notes: households, firms, government, other regions, and rest of world Net Lending/Borrowing position, as percent of GDP F I G U R E 3 Financial Balances Source: author's calculations.Notes: aggregate Financial Balances as percent of GDP

F
I G U R E 4 GDP and demand components Notes: The Figure reports Real GDP (r.h.s.) and demand components in real terms.Million euro, differences with respect to baseline.Scenario I to IV

F
I G U R E 5 Financial Balances Notes: The Figure reports the Financial Balances of private, public, and foreign sector.Percent of GDP, differences with respect to baseline.Scenario I to IV

F
I G U R E 6 Private sector Balance sheets Notes: The Figure reports private sector balance sheet.Million euro, differences with respect to baseline.Scenario I to IV

F
Figure5(in the upper-right quadrant) shows that the net lending of the public sector is nihil in the first year, goes slightly negative in the following four periods and then starts to fall, producing a budget surplus by the end of the sample.

(
Figure AII.2 reports the evolution in the variables related to labour market, namely, the unemployment rate (a), the average wage (b), the productivity level (c) and the GDP deflator, which corresponds to our price variable (d).As Figure (c) shows, the model overshoots in the first third of the sample, and undershoots in the following years up to 2010, where it aligns again to historical data Finally, FigureAII.3shows the evolution of the financial side of the model.Figure (A) display the net lending/ borrowing position of the aggregated sectors, as percent of GDP.Figures (B) and (C) represent private wealth-to-GDP ratio and loans-to-GDP ratio, respectively.

Table 1
summarises the regional models currently in use in Italy 15 at government and academic institutions, highlighting their main features.For brevity, we focus only on empirical works, which are described in more details in Selected large-scale models of Italian regions A B L E 3 Campania.Balance Sheet Matrix All equations pass the standard tests for autocorrelation, normality and heteroskedasticity of residuals.
Notes: the first column of the Table displays the values for major endogenous variables in the Baseline (in real 2015 billion euro, or percent change).The rest of the Table reports the changes -relative to the baseline -in endogenous variables in Scenario I to V at 1, 5, and 15 periods after the shock (in real 2015 million euro, or percent change).
Notes:The equation passes the standard tests for autocorrelation, normality and heteroskedasticity of residuals.T A B L E 4 Baseline and Scenario I-IV: effect of the shocks on endogenous variables Correction added on 18 May 2022, after first online publication: CRUI funding statement has been added.] Caiani et al. (2016) and Galanis (2017)in the context of a Kaleckian growth model whileMandarino and Dos Santos (2020)use instead a Sraffian super-multiplier model;Dafermos, Nikolaidi, and Galanis (2017)and Carnevali, Deleidi, Pariboni, and Veronese Passarella (2020) provide applications for a SFC approach to ecological economics; finally, another rich line of research uses the SFC principles to build Agent-Based simulation models: see, among many,Caiani et al. (2016).
pute the shares of Bonds (ratio B h ) and foreign assets (ratio F h ) in households' portfolio.Notice that in FAIS Bonds and