Economic Modelling 28 (2011) 741753

Contents lists available at ScienceDirect

Economic Modelling
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e c m o d

Long-term macroeconometric models The case of Poland

Wladyslaw Welfe
Institute of Econometrics, University of Lodz, Poland 41 Rewolucji 1905r., 90-214 Lodz, Poland

a r t i c l e

i n f o

a b s t r a c t
Long-term forecasts and scenario analysis should be based on macroeconometric models. The core of longterm models is extended by introducing production functions generating potential output. Their specication includes total factor productivity (TFP) being representative of technological progress. It depends on knowledge capital, i.e. human capital, domestic and foreign R&D. Several channels of transfer of foreign R&D can be distinguished. The potential output differs from the effective output, representing nal demand, underlying business cycle uctuations. To study potential disequilibria a system of equations explaining nal demand must be established. Thus the long-term macroeconometric model must be a complete model. Its use may cover long-term forecasts and scenario analysis based on model simulations. The paper outlines the above specications of the long-term model using as example a new model of the Polish economy. The model is medium-sized. It covers demand and supply side, including prices and nancial ows. The results of multiplier analysis are shown revealing model feedbacks, including generation of business cycles. The results of its application are shown: long-term forecasts up to the year 2030 as well as scenarios of development of the Polish economy, including recession scenario. 2010 Elsevier B.V. All rights reserved.

Article history: Accepted 18 May 2010 Keywords: Macromodels Extended production functions Knowledge capital Long-term forecasts Scenarios

1. Introduction In the last decades the growth of new modern market economics was associated with the rapid development of knowledge capital. Its impact became the major factor of economic growth, outperforming the investment in xed capital and labour increase. Hence, the economic community has come to a common conclusion that the contemporary economies gravitate to a structure known as a knowledge-based economy. This concept has been formulated in contrast to an industrial economy system that prevailed in the last centuries (Smith, 2002). Even though the earlier economic systems also took advantage of knowledge that determined their technological progress, at the turn of the 20th century the role of knowledge capital started to dominate in the functioning of economies as a result of automation of manufacturing processes, speedy distribution of management information (recently via the Internet), and in economic growth related to endogenisation of technological progress, mainly due to the development of the Information and Communication Technology (ICT). There is vast literature based on new theories of endogenous growth aimed to explain at the world-wide level the differences in the rates of growth of particular countries and the issues in their convergence. It is
Tel.: + 48 42 6355172. E-mail address: emws@uni.lodz.pl. 0264-9993/$ see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.econmod.2010.05.002

based on an analysis of international cross-section samples and emphasizes the role of particular factors of growth. The explanation of total factor productivity (TFP) growth including the impact of domestic and foreign R&D expenditures and human capital was studied from many different points of view (see Welfe, 2007b, 2009c). The generation, absorption and use of the many forms of knowledge capital, were the subject of many empirical studies. Their non-technical excellent summary can be found in Helpman (2004). The results of this research are rather exceptionally applied to investigations into economic growth of single economies, except for the US economy. The in depth studies led by Professor Jorgenson (Jorgenson et al., 2003, see also Richards, 2000) are worth mentioning. It applies also for Poland (see Welfe, 2001). This has a practical aspect, too. The authorities and the scientic community of a country need to have an instrument that will help construct scenarios of long-term economic growth for 2030 years ahead. All these studies were concentrated on the supply side of the economy. Of course, the knowledge of a long-term, extended production function with endogenous TFP being dependent of knowledge capital embodied it xed capital and labour is a prerequisite of such studies. However, this instrument allows to generate potential output only. It may considerably diverge from effective output which represents realizations of nal demand that underlie business cycle uctuations. Hence, to be used in empirical analysis and simulation exercises we need to construct a complete model that contains both the

742

W. Welfe / Economic Modelling 28 (2011) 741753 Table 1 Major characteristics of the model W8D-2007. Characteristics Variables total - Excluding dummies - Exogenous (E) - Excluding dummies - Endogenous Equations total - Stochastic (B) - Identities (I) Lags, leads - Maximal lag - Maximal lead -Lags total (L) -Leads total (W) Endogenous variables - Presimultaneous - Jointly determined - Post simulataneous Feedback variables Equations by blocks Final demand total - Domestic - Foreign - Macroaggregates Production factors Technical progress Potential output Average wages and incomes Prices, deators Financial ows - Money markets - State budget; - Balance of payment Macrocharacteristics The number of variables/equations 393 258 157 22 235 235 111 124 8 0 165 0 24 80 131 7 64 34 19 11 17 40 7 12 36 48 16 18 14 11

nal demand and the total supplies. That makes possible to estimate the likely disequilibria: output gap, unemployment, foreign trade decits, etc. The model should be closed by introduction price, wage and nancial ow sector. All this justies the need for construction of extended, long term macroeconometric models for single economies (W. Welfe, 2008a). We tried to show their suggested structure using as an example the annual long-term macroeconometric models W8D built for the Polish economy. Their characteristic is provided in the following section. Next sections of this paper contain discussions of properties of many alternative measures of major determinants of economic growth, the alternative approaches to their explanation with an attempt to show the interdependencies within the whole economic system. The role of investment in xed capital and knowledge capital is discussed in the light of multiplier analysis. The applications in scenario analyses based on model simulations are provided at the end of the paper. 2. The macroeconometric models of a knowledge-based economy The quantitative mechanisms that underlie the growth of a knowledge-based economy can be described empirically by means of adequately expanded macroeconometric models. Such models should draw on economic growth theory which has been enjoying its renaissance, and especially on endogenous growth theory (see Grossman and Helpman, 1991; Barro and Sala-i-Martin, 1995; Aghion and Howitt, 1999, and more recently Nahuis, 2003; Tokarski, 2001, 2007). The long-term macroeconometric models built along these lines, extended to include processes in which knowledge capital is generated and used, seem to be the most relevant tools of long-term economic analysis. Their structure may follow the framework of the mainstream models outlined by Klein et al. (1999). See also Bodkin et al., 1991 and Whitley, 1994. These models specify the nal demand equations along the neoKeynesian lines, but the potential output and demand for the factors of production, as well as impacts of technological progress they generate referring to the neoclassical theory of production (Solow, 1957). This approach draws on the early theories of growth developed by Harrod and Domar and on the concept of models of production possibility frontier that have recently been developed by Jorgenson (2000). The stylized empirical model of growth by W. Welfe (2005) follows a similar approach. In Poland empirical investigations referring to the above developments build on the concept of an empirical model of economic growth developed by Welfe (2000). This concept gave rise to the construction of the long-term macroeconometric models of the Polish economy W8D (see Welfe, 2001, 2004 and recently W8D2007 in 2008a, 2009a). The W8D models were built to encourage studies of the evolution of Polish economy towards a knowledge-based economy and elaboration of long-term scenarios reaching the years 20202030 (Welfe, 2009b). More specically, the models enabled a thorough analysis of the impacts of endogenous technological progress, or rather of the changes in knowledge capital the progress induces, while allowing for relevant feedbacks. The new, long-term model W8D 2007 is a complete structure. Its quantitative description is shown in Table 1. It is one sectoral, mediumsized model (see W. Welfe, 2008b, 2009a). The core of its simulation version comprises several blocks of equations, traditionally following the familiar classication of economic activities. The blocks explain: - nal demand, including exports and imports; - the supply side, including potential output, and the primary factors of production; - impact of technological progress (TFP); and - prices and wages and nancial ows.

The nal demand block explains private and government consumption, investments and foreign trade. In the next blocks, the long-run potential output is generated, depending on xed capital, labour and TFP determinants, i.e. human capital per employee and cumulative R&D expenditures, both domestic and foreign. The direct and indirect channels of the transfer of foreign knowledge capital via imports are distinguished. The indirect channels comprise knowledge capital embodied in imports of investment goods and/or imports of high-tech and low-tech products. The role of investment as a factor determining an increase in potential output as well as in nal demand is emphasized. It offers the possibility of studying potential disequilibria in the commodity markets in the long run. The last blocks of the model are prices and nancial ows. Prices react to disequilibria and changes in costs, while wages are formed in the course of negotiations. The nancial ows are explained within particular institutional sectors. The economic mechanisms represented in the system of models' equations can be summarized in the following manner. There are familiar feedbacks identifying relationships that occur in the production sector among: a) consumption, production and employment (the consumption multiplier), b) investment and production (the accelerator), c) production and the nancial sector; the relationships are transmitted, inter alia, via the tax system and the budget (the scal multiplier), and d) prices and wages (the inationary spiral). The relationships between the production sector and the nancial sector are mainly taken account of by price and wage adjustments that in turn affect the intensity of the quantitative adjustments (such as changes in demand). Schematic relationships between particular blocks are illustrated in Fig. 1. When the supply side is extended by introducing equations generating potential GDP, additional mechanisms can be observed in

W. Welfe / Economic Modelling 28 (2011) 741753

743

Fig. 1. Structure of model W8.

the real sector. On one hand, investments in xed capital that depend on effective GDP, increase the stocks of xed capital and potential output. It changes the rate of utilisation of production potential and affects the investment expenditures and ultimately the effective GDP. This specic feedback plays a major role in generating the business cycle uctuations, as shown by the outcome of a multiplier analysis to follow. On the other hand, there are additional relationships related to the generation of knowledge capital. The stock of domestic R&D shaped by real domestic expenditures on R&D, and thus by domestic GDP, increases TFP and potential GDP. Likewise, the foreign stock of R&D and additions to foreign R&D transferred with imported machinery inuence domestic TFP and GDP. Similarly, human capital per employee that depends on investment in human capital inuences TFP and GDP due to the increasing real expenditures on education (especially on the tertiary education improving the structure and educational attainment of the graduates). If potential GDP equalled effective GDP, i.e. in the case of a long-run supply driven regime, the model would be provided with new feedbacks. 3. Modelling nal demand Total nal demand comprises domestic nal demand Ht and net foreign demand EMt. It determines demand for the domestic output Xt Xt = Ht + EMt : 1

Domestic nal demand can be decomposed into household (consumer) demand Ct, demand of public institutions Gt, demand for investments in xed capital Jt, and demand for inventory increase Rt: Ht = Ct + Gt + Jt + Rt : 2

Proposals have been formulated in the recent years to extend the notion of demand for investment goods to include investments in knowledge capital, i.e. R&D and educational expenditures, in addition to investments in xed capital.1 If such an extension were accepted, a pertinent modication of the SNA would be required, reducing the demand for consumer goods of public institutions and increasing total investments.2 The demand of households and public institutions for consumer goods can be modelled in traditional manner. For developed and stable economies the F. Modigliani's life-cycle hypothesis should be the point of departure. In many countries, however, for instance Poland, this applies to the behaviour of a rather small fraction of the population. Most households living in unstable conditions are constrained by either real disposable income or credit. Because of the specic, important role played by the investment demand we shall concentrate on the analysis of its specication in more detail. The specication of private enterprises' demand for investment goods starts with an accelerator. To put it simply, producer capacities expand in the long run following expected increase in the demand for products the producers can provide. This increase creates, allowing for restitution demand the potential demand for investment goods. It has to be adjusted for a likely change of the level of utilization of the available equipment. The perception of investment risks, changes in the protability of investment projects and substitution between labour and capital also has an effect on the effective demand. The expected future output is typically represented by output generated in the past. As for machinery and equipment, the Koyck transformation leading to a reduced form of the investment demand function can be used, where the explanatory variables are being

1 Zienkowski (2003) suggests to introduce a new macro category called investment in development. 2 Jorgenon et al. (2002) add the value of services provided by dwellings and consumer durables to the private sector's GDP.

744

W. Welfe / Economic Modelling 28 (2011) 741753

conned to lagged investment (Jt 1) and current output Xt (see W. Welfe and A. Welfe, 2004). The rate of capacity utilization (WXt) may be obtained using several approximations (see the next section). The investment risk RJt can be approximated using various indicators, such as government decit, government debt service, or a rate of ination. The protability of investment projects is usually expressed by a ratio of producer prices (PXt) and user costs KIt, the latter predominantly depending on the long-term interest rate RLt. The rate of wages (WBPt) and investment deator (PJt) stands for the effects of substitution. Hence, a typical specication of the investment demand function is as follows:
1 Jt = AJt1 Xt 2 WXt 3 RJt 4 PXt =KIt 5 WBPt =PJVt 6 e

cycle uctuations. The rate of capacity utilization is dened as WXt = Xt = Xt :

P

The rate can be decomposed further into the rates of utilization of the xed capital WKt and total time worked WNt. Owing to these partial rates of utilization, we can redene the production function (4) into:
1 t

Xt = BAt Kt WKt Nt WNt or

1 t

where KIt = PJt(RLt + ), PJt is an investment deator, RLt is the real long-term interest rate, WXt is the capacity utilisation rate, and is the rate of depreciation. In the W8D model the following elasticities were obtained for investments in machinery equipment: with respect to the output long-run 1, but short-run 2, with respect to the capacity utilization: long run 1.1, with respect to the user costs: long-run 0.1, short-run 0.06, and with respect to the wage/investment deator: long-run 0.5 and short-run 0.1. The above investment function can be enlarged even further, for instance to reect changes in the tax rates. Equations explaining foreign trade can be constructed in a traditional manner. Imports and exports depend on the levels of economic activity and relative prices. 4. Modelling the process of production: the extended production function Domestic products are generated via the process of production that transforms the factors of production xed capital and labour by means of a specic technology and organization of labour. Analysis of the production process requires the quantication of its specic relationships. An important instrument used in such analysis is the production function. Among its various forms, a double-log function and a CES function with constant elasticity of substitution are frequently used. Following many authors, we recommend the double-log production function, i.e. a CobbDouglas function with constant returns to scale in its extended version: Xt = BAt Kt Nt where XP t At Kt Nt is the potential output (GDP at the macroscale) in constant prices, is the total factor productivity (TFP), is the xed capital in constant prices, and is employment.
P 1 t

Xt = BWXt At Kt Nt

where WXt = WKWN(1 ). t t To estimate the parameters of the above function, we need information on the rate of factor utilization and total factor productivity. Seeking a relevant approximation of the utilization rates of particular factors is not a trivial task, because the rates differ in terms of their accuracy and statistical coverage. Some methods take advantage of industrial surveys that ask direct questions about the level of utilization of machinery and equipment, as well as time worked (Grzda-Latocha, 2005). Central banks use techniques involving analysis of the residuals (deviations from a GDP trend). Other procedures use the aforementioned decomposition of the rate of utilization of xed capital and employment. Because the rst category fails to provide empirical estimates, the characteristics of utilization of time worked by employees are usually used. Several countries use total time worked by employees as the explanatory variable, which can be decomposed into the number of shifts worked and the length of the working time. These characteristics can serve as the approximate measures of time worked by machinery and equipment (see W. Welfe, 1992). 5. The dynamics of total factor productivity: problems of measurement Following the concept of Solow residual, total factor productivity is commonly used to represent the effects of knowledge capital absorption and widely applied to international comparisons (Florczak and Welfe, 2000; W. Welfe, 2001). However, some measurement problems that constrain its use have not been solved yet (W. Welfe, 2002, 2007a,b, 2009c; Cornwall and Cornwall, 2002). Firstly, TFP is computed using effective (i.e. observable) output and not potential output. Secondly, several studies, mainly sectoral, use the concept of gross output instead of value added. Further, output elasticities with respect to xed capital are frequently calibrated and not estimated. All these issues require additional discussions so that empirically acceptable solutions can be reached. The spectrum of research should be broadened towards sectoral and regional decomposition and it should distinguish between the high and low-tech industries. It may additionally include the multisectoral models, such as the inputoutput ones. The above questions will be analyzed in some detail. Because total factor productivity is not directly observable, Solow (1957) residual has been commonly accepted as a starting point for TFP computation. This residual is derived from the production function where the only explanatory variables are the primary factors of production, i.e. xed capital and employment. Using the CobbDouglas production function with constant returns to scale (4), we get: Xt = BKt Nt
1 t

Estimating the function's parameters for a market economy is not a straightforward exercise, because observations represent effective output, i.e. the realization of the nal demand Xt, and not potential output, and most frequently Xt b XP. Therefore, changes in the rate of t utilization of production potential must be addressed. In most cases, they originate from changes in nal demand, i.e. from the business

e ;

W. Welfe / Economic Modelling 28 (2011) 741753

745

where Xt is potential output obtained under the assumption that technological progress exerts no impact, i.e. A = 1. The rate of TFP growth is assumed to follow the rate of growth of the residual. It is calculated by subtracting the rate of output growth, which was obtained assuming that only the rates of growth of Kt and Nt count, from the rate of growth of output Xt :   h i residualt = X t Xt = Xt a K t + 1 Nt 8

Therefore, the dynamics of the residual reects the impact of technological progress only for a constant rate of capacity utilization, i.e. WXt = 0. This observation is especially meaningful in the research into countries or periods with changing rates of capacity utilization, mainly following the business cycle. Let us note that the bias could be partly reduced, if we managed to insert some indicators of individual factor utilization, e.g. to swap the number of employees for time worked. 6. Specication of the function explaining TFP dynamics: the role of innovations, research and development The tradition of modelling the TFP growth factors is quite short and many questions still await their answers. A frequently suggested way of explaining TFP dynamics is a decomposition of TFP changes into the effects of the readily available capital of knowledge (W), the impacts of expanding knowledge t capital embodied in xed capital (AK) and increasing human capital t (N). Taking the production function (4), we have: t

where Xt is an effective output. Using this denition of the rate of TFP growth takes us back to the already mentioned, several questions that are solved in different ways (W.Welfe, 2002). Firstly, the prevailing view is that output is represented at the macro scale by value added (GDP for the national economy). Hence, the only explanatory variables that enter the production function are the primary factors of production xed capital and employment. At the micro scale, and often mezzo scale (sectors, branches of industry), the most frequent measures of output are gross output or sales in real terms, mainly because these indicators are more readily available and their calculations are more precise. However, this approach involves an appropriate extension of the production function (4). Inputs of energy and intermediate goods must be added as the explanatory variables, i.e. the KLEM production function must be used. Fullling this requirement is not an easy task, as information on the use of intermediate inputs is scarce and hardly available for the time series. Secondly, using the Eq. (8) implies that the value of output elasticity with respect to xed capital is known. This value can be arrived at by either estimating parameters of function (4) or calibration. In the rst case, the function explaining TFP, i.e. At, must be specied in detail. In the other case, the results obtained under the neoclassical theory of production (Barro, 1999) are employed; they imply that a value of parameter can be approximated by the share of surplus in total value added, i.e. by subtracting the share of labour costs from 1. Even though this approach is widely used in empirical research, its application stirs some doubts, because the long-run share of labour costs varies in time and its denitions are not uniform. For instance, different research projects place the share's values for Poland within a range from 0.25 to 0.5 (the latter being close to SNA estimates). For computations performed within the W8D 2007 model the value 0.5 was accepted. This spread of parameters seriously affects evaluation of the rates of TFP growth. For many years the slow rates of growth of xed capital were associated in Poland with declining employment. After the labour costs were given high weights, the rates of TFP growth showed an upward bias (see W.Welfe, 2002). However, the most difcult problem arises when the rates of TFP growth (Eq. (8)) are computed with the observed rates of growth of effective GDP (Xt) instead of the unobserved rates of growth of p potential GDP (Xt ). This approach is likely to yield biased estimates of TFP dynamics, because the residual derived from Eq. (4) reects not only the impact of technological progress, but also uctuations in the rate of capacity utilization WXt. From the denition of WXt it follows that Xt = WXtX p, hence Xt = WXt + Xp. t t In fact, the rates of TFP growth should be computed using the production function (4), which assumes full utilization of the production factors. We then have: h i p At = Xt K t + 1 Nt : 9

At = At

+ At + 1 At ;

11

If we transform the expression dening Solow residual by replacing Xt with the term WXt + Xp, we obtain: t   p Residualt = WXt + Xt a K t + 1 Nt = WXt + At : 10

where () is the rate of growth. The effects of generally available knowledge capital (AW) are either t treated as exogenous (usually as an exponential function of time) or attributed to the growth of knowledge capital associated with improving quality of employment. In the past, the effects of expanding knowledge capital embodied in xed capital were treated as functions of time; successive generations of xed capital were assumed to represent rising levels of technology. A simplication was occasionally applied, distinguishing only the share of new equipment. The stock of machinery and equipment was frequently used in lieu of total xed capital. Recently, following the mass computerization of production and management processes, decomposition of xed capital into computers, computer programs and teleequipment has been suggested. For instance, this approach was used in studies exploring the growth of the US economy (Jorgenson et al., 2002; Jorgenson et al., 2003), and then in research on the OECD countries (Colecchia and Schreyer, 2002) and the Netherlands (Leeuwen and Wiel, 2003). It means, however, the removal of computerization effects from the notion of TFP. When the starting point is decomposition of TFP dynamics as given by Eq. (11), the impact of TFP embodied in xed assets (AK) can be t related, both directly and indirectly, to the effects of innovations and R&D expenditures. In the rst case, patent citations, publications, and technical proximity are used as the approximate variables explaining this impact (Peretto and Smulder, 2002). Viewed in broader terms, however, this impact depends on the anticipated results of R&D expenditures. The cumulative R&D expenditures both domestic (BRKK) and foreign, transferred from abroad (BRKM) are assumed to t t represent the capital of technical and organizational knowledge (Coe and Helpman, 1995). Further efforts are necessary to answer many specic questions in this area, especially these concerning the transfer of foreign R&D. The direct and indirect channels through which R&D is transferred are distinguished. The expanding systems of telephone lines, closer technological proximity, more frequent use of patents, etc., stimulate the direct transfer of R&D (Lee, 2005), whereas imports of commodities, i.e. of intermediate and investment goods, represent indirect transfer (Xu and Wang, 1999). The indirect transfer of knowledge can be summarized by computing the weighted sum of R&D expenditures incurred by the distinguished country j: BRKi = wj BRKij ;
j M K

12

746

W. Welfe / Economic Modelling 28 (2011) 741753

Fig. 2. 10% increase in investment outlays.

where wj represents a weight assigned to R&D expenditures of the country j. The weights can stand for particular countries' shares in the total imports of the analyzed country, or rather for the ratios of imports

from these countries (i.e. their exports) to their GDPs (Lichtenberg and Van Pottelsberghe de la Potterie, 1998). The weights can be linked with the imports of intermediate goods (transfer of technology), with imports of investment goods (transfer of new machines,

W. Welfe / Economic Modelling 28 (2011) 741753

747

Fig. 3. 10% increase in the world trade.

etc.), which seems a better approach (Bayoumi et al., 1999), or with total imports, when their decomposition is not possible. This approach was used in earlier versions of W8D models for the Polish economy. However, in the most recent version of the model imports were decomposed into groups of commodities distinguished by the level of technological advancement.

In the last 15 years, the discussed research has been given international dimension and its scope includes now not only the industrialized countries, but also the developing ones (Engelbrecht, 1997; Bayoumi et al., 1999). However, the role of FDI in stimulating TFP dynamics continues to be debatable. It is also necessary to explore further the impact of domestic R&D expenditures and human capital on

748

W. Welfe / Economic Modelling 28 (2011) 741753

the rate of absorption of the transferred foreign capital of knowledge (Cincera and Pottelsberghe de la Potterie van, 2001). An extension of research programes to cover TFP development in sectors or branches involves the introduction of intersectoral transfers of knowledge capital. The above relationships are multiplicative. Hence, the rst approximation we can write is: ln At = o ln BRKt + 1 ln BRKt ;
K K M

13

where is a weight representing the share of imports, i.e. the degree of openness of the economy. In the model W8D 2007 for Poland the elasticities with respect to domestic R&D capital were close to 0.3, which is in line with the results reported by other authors for industrialized countries, whereas with respect to foreign R&D capital they exceeded 0.6, which was found characteristic for less developed countries. Because knowledge capital depreciates in time, its value represented by cumulative R&D expenditures will be derived from the balance equation; in the case of domestic capital it reads: BRKt = BRKt11 BRKt1 + BRt where - BRK is current R&D expenditures (constant prices), and t - is the rate of depreciation of knowledge capital embodied in xed capital that ranges, depending on the author, from 0.05 to 0.15. It is worth noting that Eq. (14) provides a link to R&D expenditures nanced by the government budget and the enterprise sector. 7. The effects of increasing human capital
K K K K

The weights are standardized and the value of the lowest educational level serves as a base. In most cases, the weights represent the number of schooling years. However, the third approach, where the weights reect the market efciency of the level of education, seems to be the most appropriate treatment of human capital functioning as a factor of production (W. Welfe et al., 2002). It was used in all W8D models of the Polish economy. It is worthwhile to note that the elasticity of TFP w.r. to human capital per employee exceeded 1. Human capital per employee is obtained by dividing total human capital by the total number of employees: ht = Ht = Nt : 16

The dynamics of total human capital can be derived from the standard balance equation: Ht = Ht1 Ht1 + HJt ; 17

14

where HJt is investments in human capital (constant prices); the weights represent either educational costs or wages, and is the rate of knowledge depreciation. Seeking relationships between investments in human capital and educational expenditure is a difcult task. To make this search effective, a submodel describing the educational process and related expenditures has been built (see Welfe et al., 2002). The broad denition of human capital can be applied following an adequate extension of the above measures. The additional characteristics must be related to the central measure. They must capture employees' experience (resulting from learning by doing), which is frequently represented by employee age, access to free knowledge (libraries), health status as indicated by life expectancy, etc. (see Benabou, 2002). 8. A note on estimation of the macroeconometric models

Earlier investigations evaluating the impacts of expanding human capital on economic growth produced inconclusive results. The main reason was that they used inconsistent data on the schooling years. Let us note that many researchers treat human capital as a separate factor of growth and eliminate it from the notion of TFP, which does not seem appropriate. In general, the scope (i.e. denition) of human capital varies in terms of its coverage. The narrow denition, most frequently used in empirical research, accentuates the differences between the levels of employees' education. The broad version allows for the impacts of learning by doing, health status, etc. In either case, the measurement problems need an adequate solution. Unfortunately, many international projects use simplied measures of human capital, i.e. shares of employees with tertiary education, shares of employees with secondary and tertiary education, or sometimes numbers of students leaving school or simply being educated. This situation has improved in the last years, as more adequate information on the number of schooling years has become available (Fuente de la, 2004). Notwithstanding, only few researchers take advantage of the newly developed summary characteristics of human capital per employee. These characteristics of human capital (Ht) are designed as the weighted sums of employees with different educational levels i: Ht = i Nit ;
i

The parameters of the equations in the W8D 2007 model for the Polish economy were estimated using mainly updated annual time series extended to the year 2005. Most sample data spanned the years 19702005 and the nancial variables were generally derived from the period 19802005. Variables tested for the level of integration were mainly I (1), but in exceptional cases, e.g. for prices, they were I (2). The estimation of equation parameters may pose several problems that can be dealt with by means of a procedure that basically comprises two steps. First, following Engle and Granger, the long-run parameters were estimated from the data. Then the ECM procedure was applied, to allow for the short-term adjustments. The results obtained for both the segments were analyzed separately. Application of cointegration techniques to a medium-sized model has not been possible so far. The suggestions by Brdsen et al. (2005) and Garratt et al. (2006) to decompose large models wait for their practical implementation. 9. The validation of the model: the multipliers The simulation version of the model W8D 2007 is used to support the medium and long-term forecasting and scenario analyses extended to the year 2030, or even beyond, until 2040, if the model solutions and the multiplier analysis demonstrate satisfactory stability. The properties of the long-term macroeconometric models can be analyzed by means of multiplier analysis. We shall concentrate on exante multipliers analyzing the impact of 10% shocks affecting two most important variables: world exports (Fig. 2) and investment in xed capital (Fig. 3). The gures will be shown respectively for the major macro variables: household consumption (C), investment in xed capital (J), exports (E), imports (M), GDP (X), and potential GDP (XNK).3
3

15

where i is the level of education and i is a weight. The weights may represent: a) an average number of schooling years, thus Ht stands for total schooling years, or b) average unit costs of schooling; in this case Ho denotes total educational expenditure, and c) average wages per employee with educational level i; here Ht is an estimate of the total wage bill.

For detailed information see W. Welfe ed. (2008b).

W. Welfe / Economic Modelling 28 (2011) 741753 Table 2 Assumptions of for the scenarios of economic development up to the year 2030. Macrovariable scenarios 2010 2015 2020 38.0 32.9 18.0 4.0 2.1 1.3 3.0 3.0 3.0 9.0 5.2 4.5 4.7 4.2 3.1 2.60 2.00 0.65 80 40 0 2025 35.0 31.3 18.0 3.5 1.5 0.8 2.8 2.8 2.8 9.0 7.3 4.0 4.7 4.0 2.7 2.80 2.00 0.60 100 50 0 2030 35.0 31.0 18.0 3.0 1.1 0.7 2.8 2.8 2.8 9.0 4.0 4.0 4.8 4.0 2.5 3.2 2.00 0.50 100 60 0

749

Investment GDP rate in % c.p. Optimistic 26.5 35.0 Forecast 23.9 33.4 Pessimistic 22.0 20.0 FDI GDP rate in % c.p. Optimistic 4.2 4.0 Forecast 3.1 2.8 Pessimistic 0.4 1.4 Transfers net from UE in % GDP c.p. Optimistic 3.5 3.2 Forecast 3.5 3.2 Pessimistic 3.5 3.2 Exports SNA rate of growth in % Optimistic 10.0 12.0 Forecast 7.5 11.2 Pessimistic 5.0 5.0 Expenditures on education GDP ratio in % c.p. Optimistic 5.0 4.8 Forecast 4.5 4.3 Pessimistic 3.8 3.5 R&D expenditures GDP ratio in % c.p. Optimistic 1.00 2.00 Forecast 0.79 1.15 Pessimistic 0.76 0.75 Increase of elasticity of absorption of foreign R&D Optimistic 40 60 Forecast 20 30 Pessimistic 0 0

Firstly, the long-term forecast was constructed that extends to the year 2030. Next, the multipliers were calculated by means of models' simulation, using this baseline forecast. The impulse and sustained multipliers were computed. The external shock is dying in 2 years except for potential output that only starts declining after 5 years. The sustained impact is characterized by a declining tendency in exports (to 7% in the long-run), because of the negative impact of rising exchange rates. The major macrovariables show a lagged increase, the highest for investments (up to 17%), declining smoothly. The elasticity of GDP with respect to exports goes down but in the middle of the period it is close to 0.6. The analysis of the impact of 10% shock in investment generated by foreign nancing (EU transfers, FDI) is somewhat incomplete, as the exports are not affected unlike the rising imports. The impulse multipliers show a decline, which approaches zero in 58 years only, except for potential output that starts declining in 5 6 years. The sustained multipliers present an interesting picture. Initially, because of the accelerator the investments grow from 15% to 35% after 6 years. Then they decline because of declining capacity utilization and stabilize around 20%. Consumption follows this pattern with a considerable delay. The substantial growth of imports delays a GDP increase, which attains its maximum in 10 years. For the midpoint of the period the elasticity of GDP with respect to investment is close to 1. Notice that the potential GDP increases with a longer lag, but stabilizes at nearly 20%.

Fig. 4. GDP, % rates of growth (left axis) and % deviations from the baseline (right axis).

Fig. 5. Household consumption, % rates of growth (left axis) and % deviations from the baseline (right axis).

750

W. Welfe / Economic Modelling 28 (2011) 741753

Fig. 6. Investment outlays, % rates of growth (left axis) and % deviations from the baseline (right axis).

Fig. 7. Net exports/GDP (% ratio).

Hence, there is a long-run tendency for a decline in capacity utilization, which adversely affects investment growth. However, it would not be justied to conclude, that the model predicts overinvestment, as we did not allow for any relevant increase in exports. 10. Long-term scenario analyses based on model simulation The construction of the new macroeconometric model of a knowledge-based economy helped to launch a new series of long-term forecasts and scenario analyses underpinned by the model-based simulations. Two basic scenarios were distinguished: optimistic and pessimistic. In the optimistic variant, the research took into account the impacts of long-run increases in investments, in domestic R&D expenditures, better absorption of transferred foreign expenditures on R&D, and growth of human capital stimulated by larger nancial allocations to the tertiary and postgraduate education. In the pessimistic variant, it revealed a deterioration of these factors.4 It has opened the way for evaluating the likely effects of realization of particular scenarios that the administrative authorities and scientic community might wish to draw up. Table 2 shows the specic assumptions in for the consecutive 5year periods. The optimistic and pessimistic scenarios assume the still realistic upper and bottom values of exogenous variables, leaving room for alternative less radical scenarios.
4 These exercises were performed before the world crises started. Most recently the model was used in simulations to show how Poland escaped recession in 2009.

In optimistic variant, the investment/GDP rate was assumed to reach very high values (35%) but in pessimistic version, they stagnate. The expenditures on knowledge capital substantially grow in optimistic variant, but only for R&D, while in the pessimistic variant they decline. The rates of growth of exports do not exceed the longrun values (9%) in the optimistic scenario; they decline to 4% in the pessimistic one. Figs. 411 demonstrate the outcome of the simulations for the period extended to the year 2030. They show the impacts of the above assumptions on the major macrovariables.5 The optimistic scenario generates very high rates of growth of GDP initially reaching 8% and then, going down to 56% at the end of simulation period. They are cyclical (Fig. 4). The deviations from the baseline forecast are substantial: from 12% in 2013 to more than 50% in 2030. The main source of these high rates of growth is mainly the assumed increase in investment expenditures (Fig. 6.) They reveal a cyclical behaviour. They initially grow up to 17%, then they decline to 5% in 2025, to increase once again to 7% at the end of period. These expenditures will be higher than forecasted by 22% already in 2013, and twice as high in 2030. The indirect effect of investment growth on the consumption (Fig. 5) shows an initial increase by 45%, declining to 1.52.5% at

5 Because of the limited space, the tables presenting the results of all simulations are available only in the monograph W. Welfe (ed.) (2009b).

W. Welfe / Economic Modelling 28 (2011) 741753

751

Fig. 8. Potential GDP, % rates of growth (left axis) and % deviations from the baseline (right axis).

the end of period. Altogether the domestic nal demand would exceed the baseline forecast by 14% in 2013 and by 40% in 2030. The dynamics of the net exports (Fig. 7) reects the changes in assumed exports and calculated imports. After initial decline it is systematically growing, thus positively affecting the current account.

On the supply side, the high rates of growth of potential GDP are noticed, considerably exceeding the forecast (Fig. 8). They oscillate around 8% and are higher than those for effective GDP. Hence, the capacity utilization has a declining tendency, negatively affecting the growth of investment. In total, potential GDP is higher by 50% than the forecasted one as of the end of period.

Fig. 9. Total factor productivity (TFP), % rates of growth (left axis) and % deviations from the baseline (right axis).

Fig. 10. Rate of unemployment.

752

W. Welfe / Economic Modelling 28 (2011) 741753

Fig. 11. GDP deator, % rates of growth (left axis) and % deviations from the baseline (right axis).

Despite the impact of rising investment the TFP growth plays a signicant role (Fig. 9). Its rates of growth rise up from 2.8% to 3.4% at the end of period. This is due to assumed rising of domestic R&D capital and more efcient absorption of foreign R&D capital. The rate of growth of employment is declining because of the high increase in labour productivity. The unemployment rates go down to 6% (Fig. 10). The initially stable rates of ination grow in the last years of forecast to 48% (Fig. 11). Turning now to the pessimistic scenario, that assumes low levels of investment in xed capital and knowledge capital the rates of growth would be initially negative (2%) (Fig. 4). The recession would be over in 7 years. The rates of growth would reach 2% as late as in the year 2025, and 4% in 2030 owing to the results of recovery. Hence, the GDP level would be lower by ca 50% than the forecast by the end of simulation period. This outcome is mainly due to the assumed decline in investment activities, initially by 5% (Fig. 6). In the middle of the period they would rise to 2%, reaching 5% only in the last 5 years. Nevertheless, in the period in question they would be lower than forecasted by 60 70%. The level of consumption stagnates (Fig. 5). Hence, the domestic nal demand after an initial decline by 2%, would show an increase by the end of period. The net exports exhibit a rising tendency that mitigates to some extent the negative impact of domestic demand (Fig. 7). Potential GDP shows all the time positive rates of growth, however not exceeding 1% (Fig. 8). This is mainly due to the declining rates of growth of xed capital. The labour productivity rates of growth are low, but sufcient to sustain a systematic decline of employment. As a result, the unemployment rates rise dramatically from 10% to nearly 20% in the last years of simulation period (Fig. 10). The TFP rates of growth decline due to low R&D expenditures (Fig. 9). The rates of ination in the depressed economy become low; lower than in the forecast (Fig. 11). 11. Final comments The simulation analyses open the oor for constructing many alternative scenarios that would take into account factors of development other than investment in xed capital and knowledge capital. However, the advantage the presented framework has is that it allows drawing a clear distinction between the factors of growth and the results of the exercise. One of the important applications of this exercise is the possibility of calculating the time in which a particular economy might reach the GDP per capita levels of developed, industrial economies. For instance,

these calculations show that Poland may have a chance to reach the average level of UE-15 countries in 2030 only in the optimistic scenario, while in the pessimistic scenario it would remain at the 46% level. There are many other possible applications that support the construction of the visions of future growth.

References
Aghion, Ph., Howitt, P., 1999. Endogenous Growth Theory. MIT Press, Cambridge MA. London. Brdsen, G., Eitriheim, O., Jansen, E.S., Nymeoen, R., 2005. The Econometrics of Macroeconometric Modelling. Oxford University Press, Oxford. Barro, R.J., 1999. Notes on growth accounting. Journal of Economic Growth 4, 119137. Barro, R.J., Sala-i-Martin, X., 1995. Economic Growth. McGraw-Hill Inc., New York. Bayoumi, T., Coe, D.T., Helpman, E., 1999. R&D spillovers and global growth. Journal of International Economics 47, 399428. Benabou, R., 2002. Human Capital. Technical Change and the Welfare State. Temi di discussione del Servizio Studi. Banca d'Italia, Roma, p. 465. Bodkin, R., Klein, L.R., Marwah, K. (Eds.), 1991. A History of Macroeconometric Model Building, B. Elgar, Aldershot. Cincera, M., Pottelsberghe de la Potterie van, B., 2001. International R&D spillovers, a survey. Cahiers Economiques de Bruxelles 169, 120. Coe, D.T., Helpman, E., 1995. International R&D spillovers. European Economic Review 39, 859887. Colecchia, A., Schreyer, P., 2002. ICT investment and economic growth in the 1990s. Is the United States a unique case? A comparative study of nine OECD countries. Review of Economic Dynamics 3, 408442. Cornwall, J., Cornwall, W., 2002. A demand and supply analysis of productivity growth. Structural Change and Economic Dynamics 13, 203229. Engelbrecht, H.J., 1997. International spillovers. human capital and productivity in OECD economies. An empirical investigation. European Economic Review 41, 14791488. Florczak, W., Welfe, W., 2000. Wyznaczanie potencjalnego PKB, cznej produktywnoci czynnikw produkcji (Potential GDP and TFP). Gospodarka Narodowa 1112, 1055. Fuente de la, A., 2004. Human Capital and Growth: Some Results for the OECD. Current Issues of Economic Growth. Proceedings of OeNB Workshops, 2, Vienna. Garratt, A., Lee, K., Pesaran, M.H., Shin, Y., 2006. Global and National Macroeconometric Modeling: A Long-Run Structural Approach. Oxford University Press, Oxford. Grossman, G.M., Helpman, E., 1991. Innovation and Growth in the Global Economy. MIT Press, Cambridge. Grzda-Latocha, R., 2005. Ekonometryczna analiza koniunktury gospodarczej w krajach strefy Euro (Econometric Analysis of Business Cycles in Euro-area). Ekonomista 5, 621640. Helpman, E., 2004. The Mystery of Economic Growth. Harvard University Press, Cambridge, Mass. Jorgenson, D.W., 2000. Information technology and the US economy. The American Economic Review 91, 132. Jorgenson, D.W., Ho, M.S., Strip, K.J., 2002. Projecting productivity growth. Lessons from the resurgence. Federal Reserve Bank of Atlanta, Economic Review 87, 114. Jorgenson, D.W., Ho, M.S., Stiroh, K.J., 2003. Lessons for Europe from the U.S. growth resurgence. CES IFO Economic Studies 49, 2747. Klein, L.R., Welfe, A., Welfe, W., 1999. Principles of Macroeconometric Modelling. North Holland, Amsterdam.

W. Welfe / Economic Modelling 28 (2011) 741753 Lee, G., 2005. Direct versus indirect international R&D spillovers. Information Economics and Policy 17, 334348. Leeuwen, G. van, Wiel, H. van der, 2003. Spillovers effects of ICT. CBP Report 3, 2440. Lichtenberg, F.R., Van Pottelsberghe de la Potterie, B., 1998. International R&D spillovers: a comment. European Economic Review 42, 14831491. Nahuis, R., 2003. Knowledge. Inequality and Growth in the New Economy. E. Elgar, Cheltenham UK. Peretto, P., Smulder, S., 2002. Technological distance. Growth and scale effects. The Economic Journal 112, 603624. Richards, G.R., 2000. Endogenous technological advance in econometric model; implications for productivity and potential output in the United States. Economic Modelling 17, 1334. Smith, K., 2002. What is the Knowledge Economy. Knowledge Intensity and Distributed Knowledge Bases. Discussion Paper, series 2002-6. The UN University, INTECH, Maastricht. Solow, R., 1957. Technical change and aggregate production function. The Review of Economics and Statistics 39, 312320. Tokarski, T., 2001. Modele wzrostu endogenicznego (Models of endogenous growth). In: Welfe, W. (Ed.), Ekonometryczny model wzrostu gospodarczego. Wydawnictwo Uniwersytetu dzkiego, d. Tokarski, T., 2007. Podaowe determinanty wzrostu gospodarczego (Supply Determinants of Economic Growth). In: Welfe, W. (Ed.), Gospodarka oparta na wiedzy. PWE, Warszawa. Welfe, W., 1992. Ekonometryczne modele gospodarki narodowej Polski, (Econometric Models of the Polish Economy). PWE, Warsaw. Welfe, W., 2000. Empiryczne modele wzrostu gospodarczego, (Empirical Models of Economic Growth). Ekonomista 4, 483499. Welfe, W. (Ed.), 2001. Ekonometryczny model wzrostu gospodarczego, (Econometric Model of Economic Growth). Wydawnictwo Uniwersytetu dzkiego, d. Welfe, W., 2002. czna produktywno czynnikw produkcji a postp techniczny, (TFP and Technological Progress). Studia Ekonomiczne 3637, 94115. Welfe, W. (Ed.), 2004. Dugookresowy, makroekonometryczny model W8D-2002 gospodarki polskiej, (Long-term Macroeconometric Model of the Polish Economy

753

W8D-2002). : Acta U, Folia Oeconomica, 172. Wydawnictwo Uniwersytetu dzkiego, d. Welfe, W., 2005. Stylized empirical model of economic growth. In: Welfe, W., Welfe, A. (Eds.), Macromodels 2004. : Acta U, Folia Oeconomica, 190. Wydawnictwo Uniwersytetu dzkiego, d, pp. 109125. Welfe, W., 2007a. Knowledge capital and total factor productivity. Argumenta Oeconomica 19 (12), 211235. Welfe, W. (Ed.), 2007b. Gospodarka oparta na wiedzy, (Knowledge-based economy). PWE, Warszawa. Welfe, W., 2008a. Knowledge based economy. New directions of macromodelling. International Advances in Economic Research 143, 167180. Welfe, W., 2008b. The structure of a new long-term macroeconometric model for Poland. Przegld Statystyczny 55, 523. Welfe, W. (Ed.), 2009a. Makroekonometryczny model gospodarki opartej na wiedzy, (Macroeconometric Model of Knowledge based Economy). : Acta U, Folia Oeconomica. Wydawnictwo Uniwersytetu dzkiego, d. Welfe, W. (Ed.), 2009c. Knowledge based Economies, Models and Methods. P.Lang, Frankfurt /Main. Welfe, W. (Ed.), 2009b. Prognozy i scenariusze dugookresowego rozwoju gospodarczego Polski (Forecasts and Scenarios of Long-term Development of Polish Economy). Wydawnictwo Uniwersytetu dzkiego, d. Welfe, W., Welfe, A., 2004. Ekonometria stosowana, (Applied Econometrics)II edition. PWE, Warszawa. Welfe, W., Florczak, W., Sabanty, L., 2002. Kapita ludzki i jego endogenizacja (Human Capital and its Endogenization). Przegld Statystyczny 50 (2), 736. Whitley, J.D., 1994. Course in Macroeconomic Modelling and Forecasting. Harvester/ Wheatsheaf, New York. Xu, B., Wang, J., 1999. Capital goods trade and R&D spillovers in OECD. Canadian Journal of Economics 32, 12581274. Zienkowski, L. (Ed.), 2003. Wiedza a wzrost gospodarczy, (Knowledge and Economic Growth). Wydawnictwo Naukowe SCHOLAR, Warsaw.