Economic Modelling 20 Ž2002. 69᎐92
MANEGE: a small macro-econometric model of the French economy Nicolas CarnotU Direction de la Pre¨ ision, Ministry of Economy, Finance and Industry, 139 rue de Bercy, 75572 Paris Cedex 12, France Accepted 12 June 2001
Abstract This paper describes MANEGE, a small quarterly macro-econometric model of the French economy. The model mingles short-run Keynesian dynamics with a consistent, partly estimated, neo-classical supply side. A well-defined steady-state growth path is fully derived analytically. Standard simulations display plausible short to long-run responses to both demand and supply shocks. The model is helpful in a variety of policy analysis. To illustrate, we use our framework, first to gauge the role of factor costs in the evolution of French structural unemployment, then to evaluate relevant weights for building a Monetary Conditions Index. 䊚 2002 Elsevier Science B.V. All rights reserved. JEL classifications: C51; E17 Keywords: Macro-econometric model; French economy; Model simulations; Structural unemployment; Monetary Conditions Index
1. Introduction MANEGE ŽModele Agrege National pour l’Etude Generale de l’Economie. is a simple quarterly macro-econometric model of the French economy designed for policy analysis and forecasting. Building on common views over the working of the economy, it considers a one-good open economy with short-run Keynesian dynamics and a neo-classical steady state specification driven by supply-side determinants. Error correction models on French national accounts data are estimated U
Tel.: q33-1-53-18-55-20; fax: q33-1-53-18-36-28. E-mail address:
[email protected] ŽN. Carnot..
0264-9993r02r$ - see front matter 䊚 2002 Elsevier Science B.V. All rights reserved. PII: S 0 2 6 4 - 9 9 9 3 Ž 0 1 . 0 0 0 8 6 - 4
N. Carnot r Economic Modelling 20 (2002) 69᎐92
70
over the last two decades to deliver key behavioral parameters and adjustment dynamics. Overall simulations of the model display plausible short to long-run responses to standard supply and demand shocks and the model has already been used in a number of policy applications. Our work aims at filling a relative gap in the French modeling tradition. The heavily-centralized French modeling industry has tended to focus on the construction of very large disaggregated models involving a host of subtle mechanisms and technical assumptions Že.g. the METRIC model at the Ministry of Finance.. Despite a move toward smaller-scale models and efforts to disseminate the knowledge in the 1990s Žsee, in particular, the special issue of Economie et Prevision, 1998., models as a rule remain hard to comprehend outside the re´ stricted teams of day-to-day practitioners and model-users themselves often find it cumbersome to work with very large systems. By contrast a relatively compact model providing a core of behavioral equations allows both for realism and a flexible approach to policy analysis. With a dozen econometric equations and approximately twice as more technical and accounting relationships, MANEGE provides this kind of synthetic yet pragmatic tool to handle a variety of practical model-based analyses.1 A critical ingredient of MANEGE is a well specified and partly estimated long-run supply block. Firms decisions on prices and factor demands Žlabor and capital. are derived consistently from profit maximization, assuming a monopolistic competition framework, constant returns to scale with a CES technology, and the ‘right to manage’ hypothesis. We impose the relevant restrictions on the long-run coefficients in the price, employment and capital equations. In addition, the wage equation describes the outcome of a bargaining process whereby unemployment adversely affects the level of real wages. With this framework price setting by firms and wage setting by bargainers jointly determine a single rate of equilibrium unemployment. The latter is shown to depend on the tax burden, the real cost of capital and the real exchange rate. It dictates both the threshold under which inflationary pressures stem from the labor market, and the rate at which actual unemployment ultimately stabilizes in overall simulations. The demand side of the model is fairly traditional, with households spending driven by real disposable income, interest rates and inflation. Trade volumes depend on demand as well as relative prices and capacity utilization. Production is demand-determined, which, due to significant inflation inertia and adjustment costs, entails a short-run trade-off between inflation and unemployment. Adjustment between supply and demand is secured in the long-run by the impact of goods and labor markets pressures on wage and prices as well as by the feedback effects of higher prices on private spending and net exports. Overall, simulations indicate that full crowding out of demand shocks occur in up to 5 years. Convergence to the new equilibrium may, however, take a longer time, typically 10᎐20 years, notably
1
In this regard, an earlier competitor of MANEGE was the maquette Micro-DMS presented in Brillet Ž1993..
N. Carnot r Economic Modelling 20 (2002) 69᎐92
71
because of the slow adjustment of the capital stock. We derive analytically the steady-state solution of the log-linearized version of the model and show the effects of exogenous technological and policy assumptions on endogenous variables. Notwithstanding these standard features, we have taken a number of shortcuts relative to current modeling practice. In particular, we retain a backward looking setting and include lagged terms to proxy for expected variables, e.g. in wage and price equations. Consistently with this approach, we treat the interest rates and the nominal exchange rate as exogenous variables shaped by policy decisions and off-model developments. We do not include a monetary and financial block and limit the description of the public sector to a few expenditure and revenue categories. As a rule we also abstract from modeling stocks-flows interactions, thereby potentially missing important feedback effects, though the model can be run with an additional public sustainability constraint with the deficit ratio as target. The rest of this paper presents first MANEGE’s main features, then its properties in standard simulations and finally a few policy applications. Section 2 introduces the long-run supply-side specification and the corresponding dynamic equations. Section 3 briefly describes demand behaviors. In Section 4 we use an aggregate supplyraggregate demand framework to derive the full-model long-run solution. Section 5 examines MANEGE’s properties in standard simulations by reporting the effects of a demand shock and a supply shock. Finally, two illustrative policy applications are discussed in Section 6. Section 7 concludes.
2. The supply side The core of the supply block comprises decisions from firms over prices and factor demands as well as the wage equation. We focus first on the long run specification, then on empirical estimates and dynamic properties. 2.1. Long-run theoretical structure Aggregate firm behavior describes the symmetric outcome of individual decisions from a continuum of monopolistic producers. Each producer specializes in the manufacturing of a particular good and sets his price while facing a downward-sloping demand curve with a constant relative price elasticity. Input prices are taken as given at the firm level and a similar, constant returns to scale, CES technology in labor and capital is assumed throughout the sectors. Technological progress takes the form of an exogenous Harrod-neutral labor augmenting trend. Under these conditions it can be shown that individual producers maximize their profits by selecting a constant markup of prices over unit costs. In fact, as a common technology is assumed for all firms, the emerging equilibrium is symmetric and all producers choose the same price. The level of Žindividual and aggregate. production is then fixed by aggregate expenditure and factor demands follow so as to minimize costs.
N. Carnot r Economic Modelling 20 (2002) 69᎐92
72
In MANEGE, the log-linearized version of this framework is used to tie down the long-run properties of the producer price and factor demands equations. Leaving aside constant terms, we obtain a set of four equations characterizing the aggregate behavior of firms Žin log form.: y s ␣ Ž l q e . q Ž1 y ␣ . k
Ž1.
p s ␣ Ž w y e ) . q Ž1 y ␣ . c k
Ž2.
k s y y Ž ck y p .
Ž3.
l s y y e) y Ž w y p y e) .
Ž4.
where y is output, l employment, k the capital stock, e labor efficiency, eU trend labor efficiency, ␣ the labor parameter, p the value-added price, w labor cost Žincluding social charges., c k the user cost of capital and the elasticity of substitution between labor and capital. This specification bears some resemblance with that presented by Turner et al. Ž1996. for the INTERLINK model. The following interpretations hold. Eq. Ž1. constitutes the first order approximation to the aggregate production function. This equation does not determine output, which will be demanddetermined through the GDP accounting identity. Instead, it is used to calculate the technology residual e, which is just set at each date so that Eq. Ž1. holds. Over the sample period, e is filtered by an 8-year centered moving average to yield a smoother trend denoted eU . In turn, variable eU is interpreted as an index of Harrod-neutral, labor augmenting, technical progress. It is treated as an exogenous determinant in model simulations. Eq. Ž2. is the price setting equation. Aggregate price is set with a Žconstant, hence not appearing in the equation. markup over marginal unit cost. The latter, which with constant returns to scale is equal to average unit cost, comprises both labor and investment costs. A useful alternative way of writing Eq. Ž2. is the following: Ž w y e) y p. s y
1y␣
␣
Ž ck y p .
Ž5.
This has the form of a factor price frontier. It displays a link between the real wage minus labor efficiency and the real cost of capital. In MANEGE the latter is simply calculated as the sum of the depreciation rate and the 10-year real interest rate. Hence the price decision implicitly connects the long-run real wage to the value of the real interest rate ᎏ a property that proves important in analyzing full model simulations. Eqs. Ž3. and Ž4. then specify the long-run targets for employment and the capital stock. Factor demands depend on the level of production and the real cost of
N. Carnot r Economic Modelling 20 (2002) 69᎐92
73
inputs. Thus, according to Eq. Ž3., the capitalroutput ratio increases when the real cost of capital falls. Desired employment as described by Eq. Ž4. shows the following nice feature: if the real wage rises as labor efficiency Žthat is, if w y p y eU s 0, as will be the case on a steady-state path., the desired level of employment simply equals output minus technical progress Ž l s y y eU .. Wage setting takes place in a bargaining framework. Here we directly postulate a real wage target in the spirit of Layard et al. Ž1991., in which higher unemployment lowers the bargaining power of employees: w y p s e ) q y U
Ž6.
with the wedge, U the unemployment rate and a positive parameter capturing the sensitivity of real wages to unemployment and usually regarded as a rough measure of labor market flexibility. The wedge we construct collects in an aggregate variable all the factors that create a gap between real labor cost and the real after-tax consumption wage. Hence it includes the rates of employers and employees social contributions, the income tax rate and the ratio of consumer prices to producer prices ᎏ which, in turn, includes the rate of value-added tax as well as the relative price of imports Žall tax rates are averages.. The assumption in Eq. Ž6. is that wage setters fully adjust their claims to the level of labor efficiency. A decline in trend technical progress would therefore fully translate into lower wage demands in the long run. Changes in the level of the wedge will also be incorporated into wage claims. That is, wage setters try to compensate entirely for increases in the level of taxes or adverse terms of trade shocks which lower the purchasing power of employees. As shown below, this Žstrong. hypothesis entails an effect of the wedge on structural unemployment. With the additional assumption of an exogenous labor force, the set of Eqs. Ž1. ᎐ Ž6. completes the long-run specification of the supply block. To see the mechanics involved, we first combine the price setting relationship wEq. Ž2.x with the wage setting equation wEq. Ž6.x to get the following expression of equilibrium unemployment: U) s
1
q
1y␣
␣
Ž ck y p .
Ž7.
Long-run unemployment depends on the level of taxes and import prices Žthrough ‘wedge effects’. and on the real long-term interest rate Žthrough its effect on the cost of capital.. Equilibrium employment follows from the assumption of a constant labor force. Next, note that given the real interest rate, real labor cost is fully fixed by the factor price frontier wEq. Ž5.x. Labor demand wEq. Ž4.x then implicitly determines output, while investment demand wEq. Ž3.x yields the equilibrium level of the capital stock.2 The production function wEq. Ž1.x then determines e, which rejoins eU in the long-run as demand adjusts to supply. 2
N. Carnot r Economic Modelling 20 (2002) 69᎐92
74
Hence, the long-run specification of the supply side fully determines the levels of equilibrium unemployment, output and factor quantities ᎏ conditional, however, on the values of the real interest rate, tax rates and the real exchange rate Žthe latter because of the impact of import prices on wage claims.. 2.2. Empirical estimates: real wage flexibility, factor substitution and adjustment speed The empirical translation of the above framework takes the form of error correction models estimated on quarterly data over the past 20 years. We impose the long-run specification above but estimate the coefficients on error correction terms Žwhich turn out to be significant ., the crucial parameters and , as well as dynamic coefficients. We summarize thereafter the main results. 2.2.1. The effect of unemployment on wages The sensitivity of real wages to unemployment, denoted above as , is a key parameter. First, it influences the response of wages to labor market conditions, acting as one of the main channels of adjustment to market equilibrium, along with dynamic effects of excess demand measures in the price equation. Most importantly, it determines the quantitative impact of changes in the wedge or the cost of capital on equilibrium unemployment wsee Eq. Ž7.x. Our estimate stands at s 4.2%, implying that a one percentage point permanent increase in the unemployment rate would ultimately diminish wages by over 4 percentage points. As a reminder of the uncertainties involved here, however, Table 1 sums up some other available empirical estimates. 3 Our evaluation appears average among the estimates of the wage-settingrprice setting model on French macro-economic data. It nevertheless points to a somewhat higher long-run real wage flexibility than what would suggest a direct macro-transposition of the often referred to results of Blanchflower and Oswald Ž1995., based on panel data. 2.2.2. Factor substitution We estimate a value of 0.38 for , the elasticity of substitution between capital and labor. The parameter appears in both the labor and capital equations. Unconstrained estimates stand at 0.38 in the former, 0.30 for the latter. We favor the labor demand estimate and impose in the capital equation to ensure consistency. Our choice is motivated by the high uncertainties associated with our measure of the user-cost of capital. Specifically, we rely on a basic measure of the user-cost Žthe price of investment times the real interest rate plus the depreciation rate., which eschews important but hardly captured influences such as fiscal schemes and adjustment costs. In practice these factors are likely to dampen the very large swings of interest rates and thus to smooth the effective user-cost of capital. We feel our volatile proxy for the user cost downwardly biases the freely
3
One should note, nevertheless, that these estimates are not always easily comparable, since the variables included in the wage equation may differ between studies.
N. Carnot r Economic Modelling 20 (2002) 69᎐92
75
Table 1 Wage setting equation of the form w y p s U q Zw : a few empirical estimates of Cotis et al. Ž1996. Bonnet and Mahfouz Ž1996. L’Horty and Rault Ž1999. Turner et al. Ž1996. Interlink model MANEGE Cotis and Loufir Ž1990. L’Horty and Sobzack Ž1996. Blanchflower and Oswald Ž1995., on panel data
5.8 5.5 5.0 4.2 4.2 Between 3 and 5 depending on the equation 1.1᎐2.3 1 on average Ž0.1 to log U, implying 1 for U s 10%.
estimated elasticity of the capital stock to its real cost ᎏ a conjecture that is consistent with the result we get with the employment equation. That said, some skepticism is in order regarding the robustness of the elasticity of labor demand to real wage cost as well. To give an idea of the range of plausible values, Table 2 reviews some estimates from the literature. Our result of an elasticity of substitution close to 0.5 matches the consensual view of a moderate degree of substitution between capital and labor. 2.2.3. Dynamic adjustment of wage, price, and factors of production Table 3 gives estimates of the adjustment speed of factor quantities to the level of production, as well as of wage and prices relative to each other. Short-term adjustment of employment to output is fairly rapid, with 70% of the adjustment completed within a year and more than 95% over 2 years. The response of the capital stock is somewhat slower with approximately 75% of the adjustment achieved in 4 years. This, however, implies a much stronger response of investment, with an accelerator-type effect peaking at approximately 2.5 times the change in output after two quarters. Wage and price indexation occurs with relatively short time lags. Wages move in
Table 2 Some empirical estimates of the elasticity of labor demand to real labor cost Doisy and Duchene ˆ Ž2000. non-farm business sector METRIC model MANEGE Ž2000. Hamermesh Ž1993. Turner et al. Ž1996., Interlink model
0.2 0.3 0.38 0.6 ŽAverage from a range of studies. 1
76
Response after...
One quarter
Two quarters
1 year
2 years
5 years
10 years
Long-run
Employment response to output Capital stock response to output Investment response to output Value-added price response to unit labor cost Value-added price response to usercost of capital Wage response to consumer price
0.09
0.27
0.71
0.98
1.02
1.00
1.00
0.04
0.11
0.23
0.45
0.87
1.13
1.00
1.62
2.57
2.46
2.27
1.79
1.32
1.00
0.51
0.79
0.90
0.95
0.84
0.75
0.70
0.00
0.02
0.04
0.09
0.18
0.26
0.30
0.43
0.79
1.04
1.04
1.02
1.01
1.00
N. Carnot r Economic Modelling 20 (2002) 69᎐92
Table 3 Adjustment speed of factor demands and wagerprice: single equation response to a 1% permanent shock from baseline
N. Carnot r Economic Modelling 20 (2002) 69᎐92
77
line with increases in consumer prices in less than a year and producer prices catch up with labor costs in similar delays. By contrast adjustment of the value-added price to a higher user-cost of capital is quite long Žwith a mean adjustment delay of just over 5 years., reflecting the likely time period for covering investment costs. Dynamic homogeneity is reasonably well accepted in the wage and producer price equations and is therefore imposed. This rules out a long run trade off between inflation and unemployment. At the same time, it ensures that the equilibrium unemployment rate might also be seen as a NAIRU ᎏ i.e. a rate of unemployment which stabilizes inflation.
3. Demand behaviors The specification of the demand side of the economy is fairly traditional. It includes a set of equations determining the components of aggregate demand with the following main features: 䢇
䢇
䢇
䢇
Household consumption depends on real disposable income, real interest rate and inflation. The latter stands as a proxy for real balance effects. In the long-run, the income elasticity of consumption is constrained to unity. Hence the saving rate ultimately returns to baseline in simulation, except when there is a permanent change either in the real interest rate or in the rate of inflation. Residential investment is a function of real disposable income and the real short-term interest rate, the latter standing for the cost of credit or for the relative return of real estate assets. Inventory investment by firms responds to the fluctuations of real activity. Real household disposable income is itself broken down into three components: the wage bill net of social charges, household profits, and other revenues ᎏ the bulk of which corresponds to welfare transfers and financial revenues indexed to nominal GDP. Taxes and social charges result from the application of average tax rates. Imports and exports of goods and services rely on the usual model of imperfect competition with differentiated national and foreign products. Imports depend on total final expenditure Ždomestic demand plus exports. while exports respond to foreign demand Ža weighted sum of the imports from France’s main trading partners.. Both are sensitive to a measure of relative prices based on ratios between trade deflators and foreign prices. Time trends are also included in order to capture the growing penetration of imports and non-price competitiveness effects. In addition, imports are sensitive in the short-run to the condition of supply through the degree of capacity utilization.
Production is demand-determined. That is, the different components of aggregate expenditure add up to set real activity. As wage and prices adjust sluggishly, this leaves some room for the traditional accelerator᎐multiplier mechanisms in the short run. Table 4 summarizes estimates for the major demand channels. Notice-
78
Response after...
One quarter
Two quarters
1 year
2 years
5 years
10 years
Long-run
Household consumption response to real disposable income Residential investment response to real disposable income Investment response to output Exports response to foreign demand Imports response to total final expenditure
0.13
0.24
0.53
0.75
0.96
0.99
1.00
0.00
0.36
0.52
0.68
0.89
0.98
1.00
1.62
2.57
2.46
2.27
1.79
1.32
1.00
0.44
0.79
0.90
0.98
1.00
1.00
1.00
2.04
1.98
1.74
1.44
1.09
1.01
1.00
N. Carnot r Economic Modelling 20 (2002) 69᎐92
Table 4 Demand channels ᎏ selected elasticities: single equation response to a 1% permanent shock from baseline
N. Carnot r Economic Modelling 20 (2002) 69᎐92
79
Table 5 Degree of price eviction: model response to a 1% permanent increase in domestic prices Response in...
First year
Second year
Third year
Fifth year
Long-run
Real GDP Private consumption Imports Exports
y0.20 y0.24
y0.27 y0.23
y0.30 y0.26
y0.35 y0.31
y0.44 y0.44
y0.38 y0.15
y0.32 y0.24
y0.24 y0.30
y0.19 y0.34
y0.05 y0.37
Note: The simulation is carried out with exogenous prices, except trade deflators, constant real interest rates and constant nominal exchange rate.
able here is the relatively high marginal propensity to import. This entails a degree of crowding out in the wake of demand shocks which roughly compensates for the increases in consumption and investment. The gradual adjustment between demand and supply stems from the combination of two mechanisms: first, the rise in wage and prices produced by excess demand in goods and labor markets; and second, the Žnegative. feedback effects of these growing prices on aggregate demand. Higher inflation thus generates an increase in the households saving rate and a decline in domestic competitiveness that brings about a fall in net exports. These feedback mechanisms determine in practice the speed of adjustment towards long-run equilibrium. To give an idea of their magnitude, we simulated a 1% increase in domestic prices and checked its dynamic effects on the demand components. Results are reported in Table 5. In the first year, aggregate demand is reduced by 0.2 percentage point through a combination of real balance and competitiveness effects. In the middle to long-run the degree of price eviction is more than twice larger but only gradually attained. The price elasticity of demand in the long-run Žy0.44. mainly reflects the estimated price competitiveness elasticities in trade volumes equations, which are close to 0.75 for both exports and imports. Error correction models are also estimated for the consumer price and the trade deflators. These various prices are modeled as weighted combinations of domestic and foreign prices with static homogeneity imposed in the long-run. Import prices are also affected by the price of oil. The model is then completed by a series of accounting identities, e.g. to calculate the trade balance or the deficit ratio. Government spending and the short-term interest rate are regarded as exogenous policy instruments and, by hypothesis, the long-term interest rate moves in line with the short-term rate. We do not model the financial and monetary sectors of the economy. In normal use, we assume a constant real interest rate to hold throughout the simulations so that monetary policy remains neutral. In addition, we consider the fluctuations of the euro against other currencies as off-model developments. Since France is part of the euro zone, this amounts in practice to treating the French nominal effective exchange rate as an exogenous variable.
N. Carnot r Economic Modelling 20 (2002) 69᎐92
80
4. Aggregate supply and aggregate demand in the long-run MANEGE benefits from a well-defined long-run equilibrium. Building on this feature, quasi-infinite steady-state paths have been produced by projecting the various exogenous variables in a consistent way. Simulations carried out over a very long time span show that convergence indeed occurs in response to shocks, taking up to one or two decades in practice Žsee Section 5.. In this regard, an advantage of MANEGE is that its reduced size actually makes it possible to fully derive the long-run of the model in an analytical form. Using the obtained framework, the long-run effects of exogenous shocks can be quickly calculated, with the underlying mechanisms readily identified, so that the relevance of the model in depicting practical situations can be easily assessed.4 To proceed, an intuitive method consists in putting together the steady state versions of, first the supply equations, then the demand behaviors introduced above. In this way, two reduced-form steady state equations are obtained which may be loosely interpreted as an aggregate supply curve ŽAS. and an aggregate demand curve ŽAD.. Along with exogenous determinants, only two endogenous variables appear in these equations, namely real output and the producer price. Both curves intersect at the long-run equilibrium for output and the producer price. Changes in exogenous variables then translate into shifts of either ŽAS. or ŽAD., or both at the same time. The long-run supply curve essentially proceeds from the expression of equilibrium unemployment wEq. Ž7.x, combined with the labor demand curve and the long-run specification of consumer prices and import prices to eliminate other endogenous prices than the producer price. The long-run aggregate demand curve is obtained by collecting the long-run versions of the various demand components and real disposable income. After replacing parameters by their numerical values, calculations lead to the following expressions: Ž AS.
y f e q pop y 0.24 Ž te q ts q ti q tt¨ a. y 2.09r y 0.0026 Ž poil y p ) . q 0.026 Ž p y p ) .
Ž AD.
Ž8.
y f 1.87g q 0.55wd y 0.56 Ž te q ts . y 0.83tt¨ a y 1.24ti y 3.88r y 0.0016 Ž poil y p ) . y 0.39 Ž p y p ) .
Ž9.
where y is real output, p the value-added price, pU foreign prices in domestic currency Žhence p y pU is the real exchange rate., poil the price of oil Žhence poil y pU is the real world price of oil., e labor efficiency, pop the labor force, te and ts employers and employees tax rates, respectively, ti the income tax rate, tt¨ a 4 Our approach builds on Malgrange Ž1991. and chapters 3 and 5 in Whitley Ž1994.. As a matter of fact, only the long-run of the log-linearized version of MANEGE can be analytically derived and is shown hereafter. This is amply satisfactory, however, for common simulations of relatively small shocks. Full details of the calculations are given in Carnot Ž2000., though not reported here in order to keep space.
N. Carnot r Economic Modelling 20 (2002) 69᎐92
81
Fig. 1. Long-run aggregate supply and demand. Effects of illustrative shocks.
the value-added tax rate, r the real long-term interest rate, g public spending Žin percent of GDP. and wd foreign demand. Only y and p are endogenous variables. Fig. 1 shows ŽAS. and ŽAD., along with the consequences of two illustrative exogenous shocks. A few comments are in order here. First, we verify that aggregate supply is upward-sloping and aggregate demand downward sloping, as expected. Note that ŽAS. is not entirely vertical, however, the reason being the effect of the real exchange rate on equilibrium unemployment ŽAS would be entirely vertical for a closed economy.. Indeed, since wage setters adjust their claims to consumer prices, a real appreciation of the currency reduces the wedge between labor cost at producer price and the purchasing power of wages, hence easing wage bargaining and lowering structural unemployment. The effect remains quantitatively small, however. We find that aggregate demand is approximately 15 times flatter than aggregate supply in the long-run, thanks mainly to the effect of competitiveness on external trade. Next, we see that both ŽAS. and ŽAD. are homogenous with regard to prices. This entails that foreign prices Žexpressed in national currency. constitute a nominal anchor for the economy ᎏ as one would expect given our assumption of an exogenous nominal exchange rate. Hence a permanent increase in foreign prices with an unchanged real world price of oil eventually translates into a proportional rise in domestic prices, leaving real output unaffected in the long-run. Eqs. Ž9. and Ž10. also show that exogenous shocks can be classified into three groups according to their long-run implications: first, purely supply shocks, e.g. changes in trend technical progress and the labor force. Not surprisingly, these technological determinants enter the supply curve with unit elasticity. Second, we find pure demand shocks such as increases in public spending and foreign demand. These shocks only move ŽAD.. Finally, we have a number of shocks which simultaneously hit demand and supply, e.g. variations in tax rates, real interest rates or foreign and oil prices. Such shocks show up with a negative sign in ŽAS. because they alter equilibrium unemployment wEq. Ž7.x but adversely affect ŽAD. too by depressing various expenditure components.
N. Carnot r Economic Modelling 20 (2002) 69᎐92
82
Here a notable property is that tax rates appear in ŽAS. with the same elasticity: indeed, whatever the tax, a 1 percentage point marginal increase in the rate augments the wedge on labor income by the same proportion. Taxes do not have uniform effects on ŽAD. however, because they modify household income differently. The smallest coefficient is observed for social contributions, which only reduce net wages while leaving the other sources of income unaffected. The value-added tax reduces a larger fraction of household revenues, with the exception of consumer-price indexed transfers, however. Finally, the income tax rate reduces the whole of gross household income, which is why it comes up with the highest coefficient in ŽAD.. Counterbalancing this feature, however, income tax is also the most lucrative in terms of fiscal revenues: hence the various changes in tax rates may be combined with contrasted hypotheses on exogenous public spending g which may eventually broadly offset their differentiated impacts on ŽAD..
5. Full model properties In this section, we examine analytical simulations to bring light on the full model response to shocks. On the whole we find plausible results for the short-run responses to shocks and the long-run adjustment toward equilibrium. To be concise, we focus on two standard simulations: 5 a rise in public investment Žpositive demand shock. and an increase in wages Žnegative supply shock.. The simulations are carried out under the assumption of a constant nominal exchange rate and constant real interest rates. For the public expenditure simulation, we also consider the consequences of adding to the model a constraint for public finance sustainability. 5.1. Public expenditure shock The simulation consists in a permanent increase in public investment equivalent to 1% of GDP. Table 6a summarizes the main consequences. Table 6b reports the results obtained with a slightly modified version of the model in which a public sustainability constraint has been added. The latter ensures that the deficit ratio ultimately returns to baseline, using the income tax rate as an instrument: when public deficit increases relative to baseline, the rate is gradually raised in subsequent periods.6 For the first 2 years both simulations are in fact quite close and may be confounded in analysis. With sticky wage and price adjustment, expenditure mechanisms tend to predominate. The injection of public spending spurs an increase in 5
A wider battery of analytical simulations is commented in Carnot Ž2000.. The increase in the income tax rate is calibrated to ensure that, all else equal, the return to baseline public deficit occurs in approximately 3 years. Choosing another rule Že.g. targeting the debt ratio. could naturally alter the results. Conditions of equivalence between different rules may, however, be derived ŽMitchell et al., 2000.. 6
N. Carnot r Economic Modelling 20 (2002) 69᎐92
83
Table 6 Permanent increase in public expenditure equivalent to 1% of baseline GDP Year Ža. No public sustainability constraint GDP Private consumption Business investment Exports Imports Consumer price Value-added price Public deficit Žpercent of GDP. Trade balance Žpercent of GDP. Unemployment rate Žpercentage points. Žb. With public sustainability constraint GDP Private consumption Business in vestment Exports Imports Consumer price Value-added price Public deficit Žpercent of GDP. Trade balance Žpercent of GDP. Unemployment ate Ž percentage points. Income tax rate Ž percentage points.
1
2
3
5
10
Long-run
0.84 0.24
0.97 0.51
0.65 0.47
0.08 0.49
0.18 0.86
0.13 0.47
2.37
2.59
1.30
y0.54
0.29
0.13
y0.01 2.86 0.08
y0.18 2.95 0.87
y0.58 2.29 2.32
y1.56 2.34 4.72
y1.92 3.28 4.20
y1.66 2.60 4.11
0.12
1.26
3.18
6.15
4.93
4.64
0.80
0.68
0.75
0.94
0.92
0.94
y0.77
y0.84
y0.79
y1.07
y1.39
y1.17
y0.22
y0.71
y0.80
y0.42
y0.03
y0.12
0.82 0.20
0.86 0.27
0.41 0.00
y0.36 y0.50
y0.42 y1.32
y0.47 y2.22
2.33
2.26
0.62
y1.61
y0.68
y0.47
y0.01 2.80 0.08
y0.18 2.57 0.89
y0.58 1.58 2.30
y1.45 1.04 4.29
y1.00 0.67 1.58
y0.12 y0.38 0.30
0.12
1.28
3.16
5.55
1.57
0.33
0.71
0.43
0.41
0.42
0.08
0.00
y0.76
y0.75
y0.60
y0.71
y0.44
0.08
y0.22
y0.67
y0.68
y0.09
0.57
0.41
0.12
0.35
0.52
0.86
1.38
1.57
Constant real interest rates and nominal exchange rate; percent deviations from baseline.
84
N. Carnot r Economic Modelling 20 (2002) 69᎐92
private expenditure with investment rising through accelerator effects and household consumption gradually stimulated by higher revenues. Private employment adjusts with a lag. However, some crowding out occurs in the short run through a marked trade leakage. Hence the overall expenditure multiplier, which peaks after a year, hardly reaches unity ᎏ an order of magnitude which seems sensible for a relatively open economy like France. The medium-run effects suggest weak expansionary power for the fiscal stimulus, if not actually negative. Full crowding out of the demand shock is observed in a range of 4᎐6 years. It is somewhat quicker with the public sustainability constraint as, in that case, the increase in taxes starts reversing private income and consumption after less than 2 years. In both simulations prices are gradually responsive to pressures in the goods and labor markets. The impact of a lower unemployment rate on wages appears as the main inflationary channel, accounting for approximately three-quarters of the rise in prices. Higher inflation in turn restrains the initial demand boost by competitiveness and real balance effects, while private investment falls with the decline in demand and the weakening of profitability. Long-run effects are quite different depending on the imposition of a public sustainability constraint or not. With no constraint the long-run multiplier remains slightly positive at less than 15% of the initial stimulus. This residual effect is attributable to an appreciation of the real exchange rate which lowers equilibrium unemployment. Clearly, however, this is not a sustainable outcome, for the deficit ratio remains nearly 1 percentage point higher than in the baseline scenario. In addition, the weakened domestic competitiveness creates a trade deficit of more than 1% of GDP. With public sustainability the long-run impact turns out to be negative and close to 0.5 percentage points of GDP. The main reason for this adverse impact is that the marked rise in the income tax rate Žmore than 1.5 percentage points. generates an increase in the wedge on labor income. This hampers wage setting and therefore requires a higher rate of unemployment to equilibrate the supply and demand of labor. On the demand side, the burden of adjustment falls mainly on private consumption as there is a fall in after-tax incomes.7 5.2. Wage shock We simulate the effects of an ex ante increase in the level of wages. Wages are affected by a 1% residual shock and assumed to remain endogenous throughout the simulation. The purpose of the exercise is primarily to expound model properties, though it could be viewed as depicting the effects either of an increase in the minimum wage or of a rise in the reservation wage. Table 7 reports the outcome of the wage shock. The increase in wages acts both as a negative supply shock Žraising the cost of an input. and a positive demand 7 One ancillary assumption of the model is that public expenditure has no stimulating effect on the supply side. To the extent that public investment brings about positive spillovers on productivity, the result of a negative long-run impact on GDP could be nullified, or even reversed.
N. Carnot r Economic Modelling 20 (2002) 69᎐92
85
Table 7 A 1% ex ante increase in wages Year
1
2
3
5
10
Long-run
GDP Private consumption Business investment Exports Imports Consumer price Value-added price Nominal wage Real consumer wage Public deficit Ž% of GDP. Trade balance Ž% of GDP. Unemployment rate Žpercentage points.
y0.16 y0.14 y0.60 y0.15 y0.24 0.77 1.12 1.50 0.73 y0.01 0.02 0.05
y0.41 y0.27 y1.23 y0.41 y0.39 1.56 2.10 2.31 0.75 0.11 y0.03 0.15
y0.54 y0.28 y1.44 y0.62 y0.21 1.84 2.38 2.53 0.69 0.17 y0.13 0.32
y0.45 y0.13 y0.81 y0.62 0.34 1.24 1.45 1.57 0.33 0.16 y0.26 0.45
y0.16 y0.19 0.02 y0.10 0.04 0.21 0.18 0.18 y0.03 0.07 y0.03 0.17
y0.25 y0.20 y0.25 y0.24 0.05 0.58 0.66 0.66 0.07 0.10 y0.08 0.22
Constant real interest rates and nominal exchange rate. Percent deviations from baseline.
shock Žincreasing wage earnings.. However, according to our estimates the adverse supply effect dominates, even in the short-run. The shock has inflationary and recessive consequences. Inflation rises sharply as firms facing an increase in labor costs try to preserve their margins by raising prices. Wages in turn adjust to higher prices in an inflationary spiral. The ex post rise in the real wage in fact does not fully match the ex ante shock: over a 2-year horizon, the real consumer wage rises by just 0.75 of a percentage point. Not surprisingly, investment and exports fall significantly as competitiveness and profitability diminish. The negative impact on consumption may seem more puzzling, at least in the short-run. To be sure, the rise in the real wage does stimulate real household disposable income and favor consumption. Yet three counter-mechanisms combine to slightly reverse this channel. First, the decline in employment quickly reduces the growth of the wage bill. Second, the other components of household revenues such as dividends and small enterprise profits drop sharply. In addition, the model assumes that government transfers are proportional to real activity and therefore these components of household revenues decrease. Finally, the surge of inflation causes a rise in the saving rate through a real balance effect. Private consumption thus decreases from the outset, though more moderately than real output. To understand the middle to long-run response, one must get back to the implications of the factor price frontier wEq. Ž5.x. Remember that this equation entails a negative relationship between the real cost of capital and the real cost of labor Žat producer price. in the long-run. The underlying rationale is that an increase in the real cost of an input has to be compensated by a fall in the real cost of the other input if a constant level of profitability is to be preserved. Since the simulation assumes constant real interest rates Žand therefore, a constant real user cost of capital., the real wage ultimately returns to baseline in equilibrium. Countering the ex ante attempt at raising the real wage therefore requires a rise in
N. Carnot r Economic Modelling 20 (2002) 69᎐92
86
the equilibrium level of unemployment which lowers the bargaining power of employees in wage setting and restores equilibrium in the labor market. Hence the ex ante rise in the desired level of real wages is eventually neutralized ex post by an increase in equilibrium unemployment. Employment, real GDP and the capital stock then stabilize at a quarter of a percentage point lower than baseline.
6. Tools for policy analysis To illustrate the variety of potential applications, we now put forward two by-products of MANEGE which may serve as relevant indicators in the policy debate. The first is the series for structural unemployment that is implicit in the model. The other is a Monetary Conditions Index based on interest rate and exchange rate simulations. 6.1. Taxes, real interest rates and equilibrium unemployment One of the key features of MANEGE is a well-defined level of equilibrium unemployment. The latter depends on the aggregate wedge and the real cost of capital, as shown by Eq. Ž7. that we repeat here for convenience: U) s
1
q
1y␣
␣
Ž ck y p .
Ž7.
with the sensitivity of the real wage to unemployment, ␣ the labor share parameter, the wedge between real labor cost and the purchasing power of net labor income and c k y p the real user cost of capital. We have c k y p s logŽ r q ␦ . where r is the real long-term interest rate and ␦ the depreciation rate. A first order decomposition of the wedge is s te q ts q ti q Ž pc y p ., with te and ts the employers and employees tax rates, ti the income tax rate and pc y p the gap between consumer prices and producer prices Žincluding the value-added tax.. The implied historical series for French equilibrium unemployment has been computed and is shown in Fig. 2. In order to clear the remaining volatility, the resulting series has been smoothed by a 10-year moving average. The result may be viewed as an update of the seminal work from Cotis et al. Ž1996., who used a related framework. According to our estimate, equilibrium unemployment rose by 5.5 percentage points over the 1980s owing to both the rise in real interest rates at the beginning of the decade Ž2.5 percentage points. and the continuous increase in the wedge Ž3 percentage points.. It reached a high in the early 1990s at approximately 9%. Since then it has decreased by 1.5 percentage point, thanks mainly to the fall in the cost of capital. The cuts in employers’ contributions at the lower end of the wage scale, which were gradually implemented by French governments since the mid-1990s, have also helped reduced the wedge on labor cost but, to some extent, these measures have been mitigated by other increases in the tax wedge over the 1990s.
N. Carnot r Economic Modelling 20 (2002) 69᎐92
87
Fig. 2. Estimated equilibrium unemployment and contributions.
At the end of 1999, our estimate of structural unemployment stood just under 8 percentage points, which was significantly lower than the actual rate of unemployment Žapprox. 10.5 at the time. and also in the lower range of the available estimates of equilibrium unemployment ŽPisani-Ferry, 2000, collects a bunch of empirical evaluations ranging between 8 and 10 percentage points for 1999᎐2000.. Yet the continuous decline in unemployment observed throughout 2000 with very little signs of wage inflation suggest that these estimations might indeed have overestimated structural unemployment. Such an evaluation should naturally be viewed with caution, as it is derived under rough hypotheses. Specific priors are assumed on firms decisions and the wage bargaining process. Of notable concern too is whether Eq. Ž7. truly exhibits exogenous determinants of structural unemployment. One may argue, for instance, that the rise in the wedge in the late 1970s and 1980s was actually not an exogenous policy move but rather a consequence of slow growth and large unemployment which made it hard to balance the budget. Yet we feel our framework remains useful in practice. Its major virtue is to highlight the critical role played by input prices in the rise Žand fall. of French unemployment. Our quantitative results, in particular, appear consistent with broad stories of European unemployment such as Blanchard Ž2000.. 6.2. The building of a Monetary Conditions Index Monetary Conditions Index ŽMCI. have been used for some time in various institutions, notably by the central banks of open economies such as Canada ŽFreedman, 1994. or New Zealand ŽMayes and Razzak, 1998. and by private banks. The basic idea behind MCIs is to collect in a single indicator the various components of monetary conditions, in particular the interest rates and the exchange rate. Meanwhile some confusion has arisen over what MCIs exactly intend to measure and, correspondingly, over the way MCIs should be calculated
N. Carnot r Economic Modelling 20 (2002) 69᎐92
88
ŽEika et al., 1996.. Here we restrict ourselves to a well-defined concept and provide estimates.8 We construct our MCI as a weighted sum of changes in the real interest rate and the real exchange rate: I s ␥r Ž r y r 0 . q ␥q Ž q y q0 .
Ž 10 .
with r the real interest-rate, q the real effective exchange rate and r 0 and q0 baseline values chosen as historical averages. The interest rate is itself a combination of short- and long-term real interest rates, with a 70% weight attributed to the latter, corresponding to the share of long-term credits in the economy. ␥r and ␥q are weighing coefficients. One of the main tasks in building an MCI consists in evaluating these weights. Our specification is based on real rather than nominal variables. Nominal variables can lead to misleading interpretations in practice. For instance, an increase in nominal interest rates that is generated by a rise in inflationary expectations should not be viewed as a tightening of monetary conditions. On the contrary a looser monetary stance may well have brought about the rise in price expectations at first, as happened in the 1970s. What is more, inflationary tensions in the middle run stem from supply and demand imbalances which are best captured by the output gap. Real monetary determinants rather than nominal ones contribute to the build up of the output gap and should therefore be preferred. According to this view, the MCI should reflect the contribution of monetary conditions to the business cycle by evaluating their effects on real aggregate demand. The weights ␥r and ␥q should therefore measure the impact of monetary variables on final expenditure. We have evaluated these by standard simulations of MANEGE. Hence ␥r is the impact on real aggregate demand when simulating a permanent 100 basis points reduction in short and long real interest rates. Similarly ␥q is estimated by simulating a permanent 1% real effective depreciation, assuming constant real interest rates. Table 8 reports the results. The mechanisms at work include the effects of real interest rates on private spending Žconsumption and residential investment. and its effect on productive investment via the cost of capital. For the real exchange rate, the main channel involves competitiveness and the trade balance. Because of adjustment lags, the effects rise with time and the relative weight given to the exchange rate ␥q r ␥r also varies. We consider a 2᎐5-year horizon to be the most relevant, since central banks try to stabilize the economy over the short- to medium-run. This entails a relative weight between 0.3 and 0.5 for the exchange 8
Since we base our estimates on MANEGE simulations, we limit ourselves to giving a MCI for France Žrather than the euro area.. The adoption of the euro does not confine to the past the relevance of our indicator however. It is still useful for analyzing current French monetary conditions. Relative to the pre-euro period, the one change that could be expected would be a smoother effective exchange rate ᎏ though even that fact has not clearly shown up till now given the large swings undergone by the euro since its inception.
N. Carnot r Economic Modelling 20 (2002) 69᎐92
89
Table 8 Effects of monetary conditions on real aggregate demand At the end of year...
1
2
3
5
20
␥r : response of real aggregate demand to a 100 basis point decrease in real interest rates ␥q : response of real aggregate demand to a 1% real effective depreciation ␥qr␥r : relative weight of exchange raterinterest rates
0.24
0.56
0.84
1.25
1.57
0.20
0.28
0.32
0.36
0.45
0.8
0.5
0.4
0.3
0.3
Note: Figures report the impact of monetary conditions on real aggregate demand, assuming no price reaction from the supply-side.
rate, a value which is close to the ratios commonly used by practitioners. Fig. 3 plots the MCIs obtained when using the 2, 3 and 5-year horizon weights. It appears that the broad conclusions in fact do not depend on the choice of the time horizon. Fig. 3 also shows the respective contributions of interest rates and exchange rate movements to the evolution of the 3-year-weights MCI. Our indicator allows us to revisit the evolution of monetary conditions throughout the 1990s. Up to the end of 1996, monetary conditions as a whole showed a tendency toward tightening. This in fact reflected two opposite movements in the evolution of interest rates and the exchange rate. The French real effective exchange rate appreciated markedly until the mid-1990s, thanks in particular to the exit of the sterling, the lira and the peseta from the European exchange rate mechanism in 1992᎐1993. Meanwhile, real interest rates started to decline by 1993 from the highs reached in the early 1990s. Overall monetary conditions remained fairly tight in the 1994᎐1996 period though, probably contributing to very sluggish growth at the time and delaying the subsequent recovery. From 1997 on, monetary conditions have markedly eased with a continuous decline of real interest rates until the launch of the euro. Since then, the sharp depreciation of the common currency has by far outweighed the modest rebound in real interest rates. At the end of 2000, monetary conditions appeared rather accommodative compared to their average over the 1990s, yet a rebound of the euro in 2001 could quickly reverse this diagnosis.
7. Conclusion In this paper we described the main features of MANEGE, a small quarterly macro-econometric model of the French economy. Particular attention has been paid to the specification of a consistent supply-side block which is partly imposed, partly estimated. The model exhibits a number of desirable properties. Sensible elasticities are obtained for key behavioral parameters and adjustment dynamics. Full model simulations display plausible responses to demand and supply shocks
N. Carnot r Economic Modelling 20 (2002) 69᎐92
Fig. 3. French Monetary Conditions Index.
90
N. Carnot r Economic Modelling 20 (2002) 69᎐92
91
with gradual convergence toward a well-defined long-run equilibrium, which we fully characterize in an aggregate supplyraggregate demand framework. The model can serve as a benchmark in a variety of contexts. First it gives orders of magnitude for macro-economic behaviors recurrently referred to in forecasting processes and counterfactual reasoning. Then it helps analyze the domestic effects of exogenous shocks, e.g. foreign developments, monetary and fiscal moves or technology shocks. The reduced size of the model has some advantages here. It permits a quick understanding of the results, facilitates sensitivity analyses and allows one to rapidly modify mechanisms at the margin when judged necessary. Our recent experience using MANEGE suggests that this kind of tool is very efficient in shaping the policy debate, by setting a frame of reference with clearly identified and easily adjustable hypotheses. It is also useful in model comparison exercises, especially for cross-checking the properties of larger macro-models. Correlative to its simplicity however, the model presents a number of limitations. We implicitly assume adaptive expectations to prevail and make no attempt at modeling forward looking behaviors. We also largely abstract from feedback of stocks over flows, e.g. in the form of wealth effects. Finally, our representation of the supply-side remains highly aggregate. We neglect, in particular, potential heterogeneity among the producers and, perhaps more importantly, segmentation of the labor market between distinct levels of skills. As various categories of workers seem to exhibit strong behavioral differences, this shortcut could indeed prove misleading when analyzing the supply-side effects of specific shocks, such as tax measures targeted on low-skilled workers or job-rich producing sectors.
Acknowledgements This maquette was built as part of the macroeconomic modelling research at the Direction de la Prevision. I would like to thank for helpful exchanges and comments Celine Allard, Ludovic Aubert, Sandrine Duchene and Fabrice Pesin.
References Blanchard, O., 2000. The Economics of Unemployment. Lionel Robbins Lectures. London School of Economics. Blanchflower, D., Oswald, A., 1995. An introduction to the wage curve. J. Econ. Perspect. 9 Ž3., 153᎐167. Bonnet, X., Mahfouz, S., 1996. The Influence of Different Specifications of Wage Price Spirals on the Measure of the Nairu: The Case of France. Working Paper INSEE G 96 11. Brillet, J-L., 1993. Micro-DMS: Presentation et Proprietes. ´ ´ ´ Working Paper INSEE G 93 06. Carnot, N., 2000. Minimod: une maquette macro-econometrique de l’economie franc¸aise, note A2-00-68, ´ ´ ´ Direction de la Prevision. ´ Cotis, J-Ph., Loufir, A., 1990. Formation des salaires, chomage d’equilibre et incidence des cotisations ˆ ´ sociales sur le cout 92-93, 97᎐110. ˆ du travail. Econ. Prevision ´ Cotis, J-Ph., Meary, R., Sobzack N., 1996. Le chomage d’equilibre: une ´ evaluation. Working Paper ˆ ´ Direction de la Prevision 96-14. ´
92
N. Carnot r Economic Modelling 20 (2002) 69᎐92
Doisy, S., Duchene, ˆ S., 2000. Un examen du lien entre croissance emploi et de l’impact du cout ˆ du travail, note A2-00-55, Direction de la Prevision. ´ Economie et Prevision, 1998. Structure et proprietes franc¸ais, ´ ´ ´ de cinq modeles ` macro-economiques ´ special issue of Economie et Prevision 134 Ž3. 1᎐96. ´ Eika, K., Ericsson, N., Nymoen, R., 1996. Hazards in Implementing a Monetary Conditions Index, Board of Governors of the Federal Reserve System International Finance Discussion Paper 568. Freedman, C., 1994. The use of indicators of monetary conditions in Canada. In: Balino, T., Cottarelli, C. ŽEds.., Frameworks for Monetary Stability: Policy Issues and Country Experiences. International Monetary Fund, Washington DC, pp. 458᎐476. Hamermesh, D., 1993. Labor Demand. Princeton University Press. L’Horty, Y., Sobzack, N., 1996. Estimations d’un modele ` wage setting-price setting sur donnees ´ trimestrielles franc¸aises, Working Paper Direction de la Prevision 96-8. ´ L’Horty, Y., Rault, C., 1999. Les Causes du chomage en France: une reestimation du modele ˆ ´ ` WS-PS. Working Paper CSERC. Layard, R., Nickell, S.J., Jackman, R., 1991. Unemployment: Macroeconomic Performance and the Labor Market. Oxford University Press, Oxford. Malgrange, P., 1991. Multinational Models for Europe: An Overview of their Long-run Structure. Working Paper Cepremap 9132. Mayes, D.G., Razzak, W.A., 1998. Transparency and accountability: Empirical models and policy making at the Reserve Bank of New Zealand. Econ. Model. 15 Ž3., 377᎐394. Mitchell, P., Sault, J., Wallis, K., 2000. Fiscal policy rules in macroeconomic models: principles and practice. Econ. Model. 17, 171᎐193. Pisani-Ferry, J., 2000. Plein emploi, Rapport du Conseil d’Analyse Economique no 30, Paris. Turner, D., Richardson, P., Rauffet, S., 1996. Estimating the Supply Side of the Seven Major OECD Economies. OECD Economics Department Working Paper no 167. Whitley, J., 1994. A Course in Macroeconomic Modelling and Forecasting. Harvester Wheatshaf.