2006/3 Februar 2006 Documents Statistics Norway Research Department Documents Ådne Cappelen, Robin Choudhury and Torfinn Harding A small macroeconomic model for Malawi Table of content 1. Introduction......................................................................................................................................3 2. Main features of the model..............................................................................................................3 3. Detailed specification of sub-models...............................................................................................5 3.1. Introduction................................................................................................................................5 3.2. Government sector......................................................................................................................5 3.3. Household sector........................................................................................................................8 3.4. The external sector....................................................................................................................10 3.5. The business sector...................................................................................................................11 3.6. National accounts identities and adding up..............................................................................14 3.7. Financial programming.............................................................................................................16 4. The baseline scenario.....................................................................................................................17 5. Model performance........................................................................................................................19 6. Multiplier analyses.........................................................................................................................21 6.1. Increase in government spending.............................................................................................21 6.2. Increase in import price............................................................................................................24 References............................................................................................................................................26 Appendix 1. List of variables........................................................................................................27 Appendix 2. Some data issues.......................................................................................................32 Appendix 2.1. Capital stocks........................................................................................................32 Appendix 2.2. Employment..........................................................................................................32 Appendix 3. The model.................................................................................................................33 Appendix 4. Estimation results....................................................................................................36 Appendix 4.1. Introduction...........................................................................................................36 Appendix 4.2. Private consumption..............................................................................................37 Appendix 4.3. Price of export.......................................................................................................38 Appendix 4.4. Volume of exports.................................................................................................40 Appendix 4.5. Private sector employment....................................................................................41 Appendix 4.6. Factor price............................................................................................................43 Appendix 4.7. Consumer price.....................................................................................................44 Appendix 4.8. Price of investments..............................................................................................46 1 List of figures Figure 1 Schematic outline of the model...................................................................................4 Figure 2 Historical and simulated values: GDP private sector................................................19 Figure 3 Historical and simulated values: Private consumption..............................................19 Figure 4 Historical and simulated values: Wage earners private sector..................................20 Figure 5 Historical and simulated values: GDP deflator factor costs private sector...............20 Figure 6 Historical and simulated values: Deflator private consumption...............................20 Figure 7 Historical and simulated values: Deflator private investments.................................20 Figure 8 GDP, consumption, imports......................................................................................21 Figure 9 Employment and labour costs...................................................................................21 Figure 10 Government revenues and expenditures.................................................................22 Figure 11 Current accounts and government savings..............................................................22 Figure 12 Price deflators..........................................................................................................22 Figure 13 Simplified flow chart of the model.........................................................................23 Figure 14 GDP, consumption, imports....................................................................................24 Figure 15 Employment and labour costs.................................................................................24 Figure 16 Government revenues and expenditures.................................................................25 Figure 17 Current account and government savings...............................................................25 Figure 18 Price deflators..........................................................................................................25 Figure 19 Actual and fitted values: Non-smallholders consumption......................................38 Figure 20 Actual and fitted values: Export price.....................................................................39 Figure 21 Actual and fitted values: Volume of exports...........................................................41 Figure 22 Actual and fitted values: Number of wage earners, private sector..........................42 Figure 23 Actual and fitted values: Factor price deflator, GDP private sector........................44 Figure 24 Actual and fitted values: Deflator non-smallholders consumption.........................45 Figure 25 Actual and fitted values: Deflator private investment.............................................46 List of tables Table 1 Baseline scenario: Main macroeconomic figures.......................................................17 Table 2 Baseline scenario: Central government operations.....................................................18 Table 3 Baseline scenario: Prices and costs............................................................................18 Table 4 Baseline scenario: Balance of payments....................................................................18 Table 5 Statistics from estimation in TROLL.........................................................................37 2 1. Introduction In meetings between representatives of the Malawian government and Statistics Norway in May 2004, we agreed upon developing a small and aggregated model of the Malawian economy as a first step in efforts to construct a disaggregated model based on new national accounts data for Malawi expected to become available sometime during 2006. A model project consists of many components of which one is knowledge of the software used for programming and solving the model. In order to gain some experience with the TROLL software, we agreed to start with a small model as a testing ground for later work. Also there may be delays in creating a database for a large-scale model so the project would benefit from proceeding along several independent tracks to avoid being potentially held up by delays in other parts of the overall project. A third reason is that many of the issues in econometrics and economic theory that one encounters in modelling are more or less the same irrespective of the level of aggregation. So there is a general learning potential for modelling, even by starting off by developing an aggregated model. Ending up with a small model of some hundred equations is rather on the big side, and perhaps not the best starting point. A much smaller model would serve the purpose as a training ground for learning the TROLL software. However, as applied economists we like to have some feeling of relevance when modelling. Thus the model easily becomes larger than originally contemplated even though issues relating fiscal policy and its impact on monetary conditions as well as interest rates etc. are not modelled in a satisfactory way in the present small model. On the other hand and referring back to what was said earlier, the present model of the real side of the economy (but clearly not the monetary side) is perhaps not far from how one will model various sectors later on. Hopefully many of the adding up equations and identities will be useful for a disaggregated model as well. In this document we first present the main features of model, before the various blocks of the model is more thoroughly described in Chapter 3. In Chapter 4 we present the main variables in the baseline scenario for the period 2005-2011. In Chapter 5 we study the models tracking performance by means of a historical simulation. In Chapter 6 we have used the model to simulate a fiscal shift. In Appendix 1 there is a list of all the variables in the model, and some data issues are discussed in Appendix 2. In Appendix 3 all the equations in the model is presented, and in Appendix 4 we present detailed information from estimating the econometric equations. 2. Main features of the model The structure of the small macroeconomic model can be illustrated schematically as in Figure 1. The model user will have to make assumptions with respect to a number of fiscal policy variables such as government employment and purchases, the average government wage rate, direct and indirect tax rates incl. import duties. Among the monetary variables the model user will have to fix interest rates and some financial assets such as government net lending abroad. Some variables are determined outside the Malawian economy such as demand for Malawian export produce, world market prices, foreign grants etc. Finally some other variables such as the Malawian labour force will have to be fixed. Also private investments are an exogenous variable in this model but should clearly be made endogenous at a later stage. Given these assumptions, the model will calculate a number of national account variables such as value added, price indices for value added as well as for final demand components. Financial balances in addition to the government budget balance will be determined along with some very aggregated monetary variables, like money supply and total domestic lending. 3 Figure 1 Schematic outline of the model Fiscal policy Government expenditures and revenues Monetary policy Prices, incomes, value The macro added/GDP, economic model employment, exports imports, World economy consumption Other exogenous Current account, variables Financial balances Exogenous variables System of equations Endogenous variables The model works like a simple Keynesian model if taken very literally. This means that a change in demand such as government purchases will increase incomes, employment and private consumption and thereby also GDP in the private sector. Tax revenues will increase as a consequence but not sufficient to eliminate the expenditure increase so that the government budget balance will deteriorate and increase government domestic lending. Increased demand will increase imports and increase the current account deficit. This will be financed by an increase in private net lending abroad because government lending is exogenous. If the model user does not think this is feasible he/she may have to reduce the foreign exchange reserves in order to finance the increased deficit on the current account. If world demand for Malawian exports increases for a given terms of trade, exports will increase and so will GDP. Increased incomes will again increase employment and household incomes and thereby private consumption. Tax revenues will now increase while expenditures remain unchanged so the government budget balance improves. Thus net domestic lending is reduced. Export revenues will increase but so will imports because of higher domestic demand, but the current account will also improve. This may either reduce private lending or be counteracted by increasing foreign exchange reserves. A worsening term of trade, say as a result of higher import prices, will worsen the current account and increase prices also on domestic components of demand, reduce real household incomes and private consumption. Lower real incomes will on the one hand reduce some direct tax revenues but may increase some indirect taxes due to the increase in nominal values (but again this may be counteracted due to lower volumes of consumption and imports). Our choice of how the model is closed, that is how we have chosen which variables are exogenous and which are endogenous (and is solved by the equations of the model), can easily be changed if the model user thinks another closure or solution is more realistic. We have simply made a very traditional choice in order to have a working model. If one chooses to have a fixed value of the government budget balance due to say financing restrictions or targets on domestic borrowing, parts of government spending can be made endogenous in order meet that target. Similarly, if one chooses to impose a current account target, the exchange rate or volume of imports will have to be determined so that this is achieved. If imports are to be controlled, some components of domestic demand must be adjusted for this to take place. This can be achieved through spending limits or tax rate hikes. The macro model has more than one hundred equations but most of these equations are not part of a simultaneous system. In fact the largest block constitute of 29 simultaneous equations. This should be considered the core of the model. 4 Here we explain the structure of the model in some detail to give an overview of the structure. First of all the model is solved recursively through time. This means that for a given set of exogenous variables that the model user will have to determine before the computer can solve anything, the set of endogenous variables is solved for each year chronologically. Given these input data for exogenous variables, the model first (for each year) solves an equation for the wage rate in the private sector (WP). This rate is simply proportional to the exogenous wage rate in the government sector (WG). Then government consumption in constant prices (CG) is determined given exogenous input values for government employment (LWP) and purchases of goods and services (MG). Total investments are determined as the sum of various fixed investments as well as stockbuilding. Value added in the government sector (YG) is assumed to be proportional to the input of labour in the government sector. Based on the solution of the variables in these four equations, the software then solves the largest simultaneous block of 28 variables. This block contains most of the main macroeconomic variables such as demand components, employment, and price indices. Given values for these macro variables a number of other variables such as government revenues and expenditures are determined. Then government saving, lending and interest payments are determined in a small block consisting of four equations. Later a similar small block containing the current account and non-factor services and private foreign borrowing and interest payments are determined. Finally, the macroeconomic variables are used to determine the industry structure in a "top down" system of recursive equations. If one should chose another way of closing the model, the block structure will change, so the structure above pertains to this special version of the model. We illustrate numerically the workings of the model in Chapter 6 by carrying out a simple multiplier study. 3. Detailed specification of sub-models 3.1. Introduction In what follows, we have tried to separate the presentation into equation for the various model blocks. The financial programming part is clearly the most undeveloped of these in the present model. For some equations we suggest alternatives that would result in additional equations. These equations are not implemented in the first model version. 3.2. Government sector We assume the government has no assets abroad. Further, government owned enterprises are assumed to be part of the business sector implying that the government sector is considered mainly as a functional unit rather than an institutional sector. Government revenue (GREV) is disaggregated into tax revenue (TREV) and other non-tax revenue (ONTR) Eq. 1 GREV = TREV+ONTR Total government revenue (GRTOT) includes further transfers to the government from abroad or foreign grants (TRFG) Eq. 2 GRTOT = GREV+TRFG Gross tax revenue is decomposed into direct taxes (TD), indirect taxes (TIT), and miscellaneous taxes (TOT) comprised of miscellaneous duties, export levy, other taxes, tax refunds, and collection of arrears Eq. 3 TREV = TD+TIT+TOT Then total direct tax revenue (TD) is the sum of taxes from households (TDH) and from companies (TDB) Eq. 4 TD = TDH+TDB Indirect taxes (TIT) is modelled as the sum of import duties (TITID) and an aggregate of value added tax and excise duty (TIVATED) 5 Eq. 5 TIT = TITID+TIVATED while import duties is linked to the value of import by a rate Eq. 6 TITID = TIDRATE*VI Value added tax is determined by a rate to the value of private consumption (PCPO*CPO) Eq. 7 TIVATED = TIVATRATE*PCPO*CPO Total government expenditure (GEXP) is divided into government consumption and gross investment (VCG and VJG respectively), transfers to households (TRGH), government interest payment (INTG), and other government expenditures (OEG). Eq. 8 GEXP = VCG+VJG+TRGH+INTG+OEG It is relevant to try to specify equations determining each of these variables as functions of relevant explanatory variables. We split the government interest payments into domestic and foreign public debt (INTGF and INTGF respectively) Eq. 9 INTG = INTGF+INTGD We now define government gross savings (GSAV) as Eq. 10 GSAV = GRTOT+VJG-GEXP where GEXP - VJG is current government expenditures. Notice that it is the value of investment and consumption that enter into the budget equations. Government consumption can be divided into labour costs, capital costs and material costs. Wage costs per employee are denoted as WG and the number of employees as LWG. Call the volume of material costs for consumption purposes MG and the corresponding price index PMG. Capital costs are often set equal to depreciation costs (DG) multiplied by the price index for government gross investment (PJG) but are not included in the Malawian national accounts at the present. Thus Eq. 11 VCG = WG*LWG + PMG*MG + RVCG RVCG is a residual term. If the government finances its expenditures partly by using fees, these fees should be subtracted in Eq. 11. At this stage we define material purchases net of fees. Government consumption in fixed prices (CG) is given by Eq. 12 CG = WG.0*LWG + MG + RCG Here WG.0 is the wage rate in the base year of the model (when all price indices are unity) and is equal to 0.01259. RCG is a residual that makes the equation fit the national accounts data outside the base year. Value added produced by the government is determined from the income or supply side and equals labour costs (and in principle also capital costs). The latter consist usually only of depreciation costs (in spite of the fact that cost benefit analysis would often require a positive rate of return on capital). Define the capital stock in the government sector as KG and suppose that there is a constant rate of depreciation DELTAG. We then define the usual capital accumulations formulae as Eq. 13 KG = JG + KG(-1) - DG and the depreciation (DG) is determined by Eq. 14 DG = DELTAG*KG(-1) The volume of government value added (YG) is determined by Eq. 15 YG = LWG*RYG RYG is an exogenous calibration variable used to secure that the equation fits the national accounts data. In current prices government value added is determined by Eq. 16 VYG = WG*LWG*RVYG RVYG is a calibration variable. Gross investment in the government sector in current prices is 6 Eq. 17 VJG = PJG*JG In order to determine fiscal policy sustainability it is useful to make the financing and costs of fiscal deficits endogenous. Let government gross domestic debt (LGD), government gross foreign debt (LGF) and government financial assets (LDG) have corresponding and possibly different interest rates (IRxx). Then interest payments (INTxx) can be determined respectively by Eq. 18 INTGD = IRGD*(LGD + LGD(-1))/2 Eq. 19 INTGF = IRGF*(LGF + LGF(-1))/2 The government budget and financing constraint can be written as Eq. 20 GSAV = VJG + (LDG - LDG(-1)) - (LGD - LGD(-1)) - (LGF - LGF(-1)) This equation states that gross savings plus domestic and foreign borrowing equals the accumulation of physical assets (VJG) or financial assets by lending to domestic residents (again assuming that the government does not accumulate foreign assets). If one considers foreign borrowing as somehow restricted (meaning that LGF is exogenous), this equation can be regarded as determining the government’s net domestic borrowing. Whether to include variables in gross terms as we have done, or in net terms, is not very important at this stage, but it is probably easier to obtain reasonable figures for the interest rates used in Eq. 18 and Eq. 19 if we specify the variables in gross terms. If we consider government lending to domestic residents (LDG) as a policy variable, it is perhaps most relevant to treat LGD as endogenous and determined by Eq. 20, while GSAV is determined by Eq. 10. LGD could also be disaggregated further into lending from the central bank and domestic banks as well as the issuing of bonds bought by the private sector (incl. the domestic financial sector). In the present version of the model, LGD is endogenous. As already mentioned, tax revenues should be described by the model as well. Direct taxes can be decomposed into taxation of household income and companies’ profits. Indirect taxes could be disaggregated into import duties (levied on either import volume or value, or both) and sales taxes (VAT, sales tax, or on certain goods such as petrol, alcohol, tobacco, cars or other goods that governments typically tax heavily). At present we assume that net indirect taxes consist only of a sales tax per volume unit of private sector value added (YP) so that the purchasing price (PYP) is Eq. 21 PYP = (1+TP)*PYPF/(1+TP.0)*RPYP PYPF is the value added price at factor costs and TP is the net indirect tax rate, while RPYP is a calibration variable (that equals one in the base year). In the base year both price indices are one by definition so we must divide by the value of the tax factor in the base year (TP.0), which is equal to 0.1406. Note that import duties have been included in the TP-factor for simplicity. We treat the import price (in Kwacha) as an exogenous variable in the present model but this clearly depends both on foreign prices, the exchange rate as well as import duties. Let us define household taxable income as YH. Assume that the business sector taxation (TDB) is exogenous and that the average direct tax rate on household gross income is TH. We simply postulate Eq. 22 TDH = TH*YH An alternative to simply using an average tax rate in Eq. 22 is to use a linear formulation in order to separate between marginal and average tax effects. Disregarding Eq. 21, this model block consist of 21 equations that determines 21 variables: GREV, GRTOT, TREV, TD, TIT, TITID, TIVATED, GEXP, INTG, GSAV, VCG, CG, KG, DG, YG, VYG, VJG, INTGD, INTGF, LGD and TDH. 7 The exogenous variables pertaining to the government sector are: ONTR, TRFG, TRGH, OEG, LWG, MG, JG, LGF, LDG, TDB, TH, TD and interest rates. In addition, a number of price indices and other variables enter this block. They will be made endogenous in other model blocks presented below. 3.3. Household sector The specification of the household sector will to some extent depend on the detailing level in the income accounts of the national accounts. At present the national accounts do not contain income account by institutional sectors, thus we must define household income using the available relevant data. Of importance in this section is the separation of the private sector into a smallholder sector that almost completely produce for own consumption, and the rest of the private sector which we may call the formal, or monetary, part of the private economy. Let us now consider how to model household consumption. In general a typical macro econometric consumption function would include an income term, a wealth term and a real interest term. Due to weak data we exclude the wealth-term, and the real interest term was not found to have significant impact in describing consumption1. As the income term we use disposable income in the household sector adjusted for inflation. The consumption function applies only to “monetary consumption”, i.e. households that are not smallholders, and their consumption is called CPO. Eq. 23 LOG(CPO/CPO(-1)) = 0.0536 + 0.9848*LOG(YHDR/YHDR(-1)) -0.2869*LOG(CPO (-1)/YHDR(-1))-0.6002*DUM94+RCPO The economic interpretation of this specification is that consumption will rise in the short run if real disposable income is increasing, and if the gap between the level of the consumption and real disposable income is negative (i.e. positive saving), and this is adjusted by increasing the consumption level to its long term level. The fact that the coefficient in the long term (the error correction term of LOG(CPO(-1)/ YHDR(-1))) is one, reflects that consumption follow the income level in the long run. We now focus on smallholder consumption. Let us denote this CPS and the corresponding price index PCPS. Smallholder consumption is, as already mentioned, assumed equal to smallholder production. This sector consists of self-employed so the value of production and consumption is all operating surplus. In view of the fact that the production is not marketed, but consumed by the households themselves, the operating surplus is not taxed. We further make the perhaps drastic assumption that output of the smallholder sector is proportional to the total number of self-employed in the economy. Actually there are many self-employed people that are not smallholders but we have no information on how to separate the number of self-employed smallholders and those that are part of the market economy. Note that the total labour force is exogenous also in this model. In a previous version of the model an increase in the number of wage earners either in government or the private economy would reduce the number of self-employed (i.e. they were really regarded as hidden unemployment). In this model version the self-employed will reduce the smallholder production when employed in the market economy. Thus increased employment in the market economy, will partly “crowd out” (or “in” depending on your point of view) employment of the smallholder sector. The consumption function of smallholder household is not estimated, but rather postulated as 1 Musila (2002) pp. 299 estimated a consumption function where wealth and real interest rate were tested as arguments but the equation did not perform satisfactorily. 8 Eq. 24 CPS = LS*RCPS The idea behind this function is that smallholders consume all they produce (as is the case also with the rest of the households in the long term, but not necessarily in the short term). I.e. they have no savings or borrowing. Since smallholder consumption follows smallholder production we can model the consumption level directly as a production function. The overall dominating input factor in the smallholder sector is labour (LS), and we simplify be assuming that only labour is used in their production. The term RCPS is a residual. To implement the assumption that smallholders consume all they produce (YFAGS) we specify Eq. 25 YFAGS = CPS This equation is needed in the breakdown of total GDP into sectors using historical weights. We include an equation summing monetary consumption and smallholders consumption (CPO and CPS), which is equal to total private consumption Eq. 26 CP = CPO + CPS PCP is the price index for private consumption defined as Eq. 27 PCP = VCP/CP We further assume that consumer prices for the smallholders (PCPS) follow consumer prices for total private consumption. The price indices are in fact not very different, in spite of the fact that PCP covers much more than what one would expect enters into the consumption basket of smallholders. We have therefore assumed that PCPS is equal to PCP, but include an error term (RPCPS) in order to reproduce the history of PCPS exactly. Eq. 28 PCPS = PCP*RPCPS We also need private consumption in current value (VCP), defined as Eq. 29 VCP = PCPO*CPO + PCPS*CPS This disaggregation of the private sector is in accordance with the Malawian national accounts, where smallholder production and consumption are specified as separate items in fixed and current prices. The two are almost the same but the smallholder sector has a marginal sale (of maize) to Admarc. In 2000 this sale amounted to little over 3 % of total smallholder production and is ignored here for simplicity. Household disposable income (YHD) is defined as gross income minus direct taxes Eq. 30 YHD = YH – TDH To get real disposable income we divide by the price deflator Eq. 31 YHDR = YHD/PCPO Household gross income (YH) is simply the sum of wages and salaries (WH), Part of operating surplus that accrues to households (OSH), and transfers (pensions) from the government (TRGH) Eq. 32 YH = WH + OSH + TRGH Note that we in the tax equation (Eq. 22) have used YH also as taxable income. This could be discussed. Income such as government transfers (TRGH) may be wholly or partly untaxed. There may be a different tax structure for wage and salary income (WH) compared to capital and property income or income for self-employment (part of OSH). This can be handled by specifying separate tax equations for different (groups of) income components. For now, this very simple specification has been chosen. The terms entering Eq. 32 should be linked to other model variables. We start with wage income. We have earlier specified government wages and employment cf. Eq. 11. By defining similar variables for private sector wage rate (WP) and employment (LWP) we have 9