Analyzing Economic Structural Change in a General Equilibrium Framework: The case of Switzerland from 1990 to 2001

Transcription

1 Analyzng Economc Structural Change n a General Equlbrum Framework: The case of Swtzerland from 1990 to 2001 Laurent Cretegny Ecoplan, Swtzerland May 1, 2005 Abstract Structural change s nfluenced by many factors, ncludng technologcal changes and shfts n consumer preferences. These factors have an mportant mpact on economc performance through the reallocaton of resources from one economc actvty to another. Applyng SwssAGE, a general equlbrum model for Swtzerland, ths study focuses frst on the estmaton of changes n technology and consumer preferences. It shows then how economc hstory can be explaned n terms of these drvng factors. In partcular, t shows that decrease n captal/labour rato mtgates growth n real GDP and ncrease n export-orented producton heavly contrbutes to rapd growth n Swss trade across the perod 1990 to JEL classfcaton: D24; D58; C68; F14 Keywords: Technologcal change; Preference change; Dynamc CGE modellng; Swss structural change; Trade growth The author gratefully acknowledges Peter Dxon and Maureen Rmmer for ther comments and for the materal taken from ther research. Malng address : Ecoplan, Thunstrasse 22, CH Bern. E-mal: laurent@cretegny.ch. Ths paper has been prepared for presentaton at the 8th Annual Conference on global Economc Analyss, Lübeck, Germany, June 9-11, 2005.

2 1 Introducton Structural change s mportant to economc growth. Structural change s nfluenced by many factors, ncludng technologcal changes and shfts n consumer preferences. These factors have an mportant mpact on economc performance through the reallocaton of resources from one economc actvty to another. Ths study focuses frst on the estmaton of changes n technology and consumer preferences. It shows then how economc hstory can be explaned n terms of these drvng factors. In partcular, t shows that decrease n captal/labour rato mtgates growth n real GDP and ncrease n export-orented producton heavly contrbutes to rapd growth n Swss trade across the perod 1990 to The dffculty n the estmaton s that household preferences and ndustry technologes are naturally exogenous varables and are not observable. Ths dffculty s overcome by frst conductng an hstorcal smulaton. In ths smulaton, varables that can be observed across the perod 2001 to 1990 are exogenous and shock wth ther actual movements 1. On the other hand, naturally exogenous but unobservable varables are made endogenous. The estmaton of ther movements across the perod s realzed by applyng SwssAGE, a general equlbrum model for Swtzerland. Once the hstorcal smulaton has been completed, a decomposton smulaton can be conducted. In ths smulaton, estmates of unobservable exogenous varables are used to explan structural change n the economy across the perod n consderaton. The mplementaton of hstorcal and decomposton smulatons has been frst developed by Dxon and Rmmer (2002). Ther ntal motvaton for the hstorcal smulaton was to update nput-output tables from 1987 to As the hstorcal smulaton produces also estmates of changes n tastes and technology, they analyzed the Australan motor vehcle ndustry for the same perod. They found, for example, that preferences of users of cars strongly shft towards the mported product or that ndustry s technology moved largely towards the use of captal wth respect to. Usng these results, they were able then to decompose the performance of the motor vehcle ndustry over the perod 1987 to 1994 nto the parts attrbutable to movements n changes n tarffs relatve to changes n other varables (e.g. tastes and technology). They found out, for example, that the output for motor vehcles ncrease substantally despte the negatve mpact of the lower tarffs and the approxmatvely zero productvty growth n factor nputs durng ths perod. The decomposton smulaton showed that the reason of the large ncrease n output came from the strong shft n ndustry technologes favorng the use of motor vehcles. The use of hstorcal and decomposton smulaton has been appled recently to the US economy from 1992 to 1998 (Dxon and Rmmer, 2003; Dxon and Rmmer, 2004). The decomposton results show that technologcal changes reducng costs n export-orented ndustres or ncreasng nputs of commodtes that are heavly mported are the man determnant of the rapd growth n 1 We choose our base year to be 2001 n order to use the new sectoral classfcaton for Swtzerland. However all results are gven on the 1990 year bass. 2

3 nternatonal trade. The purpose of ths paper s to show how the estmaton of changes n technology and consumer preferences can be mplemented n a general equlbrum model for Swtzerland. SwssAGE s a dynamc, computable general equlbrum model of Swtzerland 2. It s based on a mnature verson (Dxon and Parmenter, 1996) of the Monash model developed by Dxon and Rmmer (2002). The structure of the model s presented n the followng secton. The hstorcal and decomposton closures are explaned n secton 3. Secton 4 presents the hstorcal smulatons across the perod 1990 to 2001 and a sketch model for helpng wth the nterpretaton of the results. In secton 5 we analyze both the growth n GDP and n trade for Swtzerland usng the decomposton smulaton across the perod 1990 to Concluson s gven n the last secton. 2 The Model SwssAGE s a dynamc, computable general equlbrum model of Swtzerland. Its theoretcal structure s based on a mnature verson (Dxon and Parmenter, 1996) of the Monash model developed by Dxon and Rmmer (2002). SwssAGE assumes a small open economy. It conssts of ndustres and nvestors, households, foregners and the natonal government. The model assumes constant return to scale n each actvty. We requre that margnal cost equals margnal revenue whch makes pure profts mpossble to earn for any actvty. Market clearng condtons mply that supply equals demand for commodtes. Fnally, the model s assumed to be recursve dynamc wth endogenous nvestment. Producers are assumed to mnmze ther cost subject to a producton technology represented by a two-stage producton functon. The nested structure of the nput sde s shown n fgure 1 3. The top level s a Leontef (LT) combnaton of ntermedate nputs and value-added. Followng Armngton (1969), ntermedate demand s represented as a composte of mported and domestc goods. Value-added s a combnaton of labour and captal usng a constant elastcty of substtuton (CES) functon. Producers are also assumed to maxmze ther revenue subject to transformaton fronters represented by a constant elastcty of transformaton (CET) functon 4. Fgure 2 reflects the structure of transformaton. The upper level s a mxture of all the commodtes produced by each ndustry. The lower level allows converson of each commodty nto goods destned for export or for local use and s governed by a CET transformaton fronter. It follows from the 2 SwssAGE uses the GEMPACK software developed at CoPS, Monash Unversty, Australa (Harrson and Pearson, 1996). 3 Most varables are defned n the appendx. The conventon s as follows. Indexes c, f and denote sets of commodtes, factors (L=Labour and K=Captal) and ndustres respectvely. Source of commodtes (D=Domestc and M=Imports) and destnaton of goods (D=Domestc and E=Exports) are respectvely denoted by ndexes s and d. Superscrpts n parentheses gve the type of user (1=Industres, 2=Investors, 3=Households and 4=Foregners) and add a short descrpton to the varable when t s necessary (prm=prmary, src=source, dst=destnaton, tot=total, mar=margn, pur=purchase, bas=basc, mp=imports, nd=industry). 4 In ths study however all ndustres are assumed to produce a sngle commodty. 3

4 x (1,src) c, x (1) c,d, x (1) c,m, Fgure 1: Structure of producton ( = LT ), x (1,prm) x (1,tot) x (1,src) c,,..., x (1,src) c, ( ) ( ) = CES x (1) c,s, x (1,src) c, = CES x (1) c,s, x (1,prm) x (1) c,d, x (1) c,m, ( = CES x (1) f, x (1) L, x (1) K, ) assumed nput-output separablty specfcaton that the composton of nputs s ndependent the composton of outputs. We assumed that nvestors behave n the same way as producers except to one dfference. They do not use drectly prmary factors as nputs to captal formaton. Households are assumed to maxmze a nested utlty functon subject to a budget constrant. As for producers and nvestors, the lower level allows for mperfect substtuton, represented by a CES functon, between domestc and mported commodtes. At the upper decson level, consumer preferences for composte commodtes are descrbed by a Klen-Rubn functon leadng to the lnear expendture system (LES) as t s shown n fgure 3. Foregners are assumed to purchase only commodtes produced domestcally. Ther demand for exports exhbts nfnte elastcty n prces expressed n foregn currency whch reflects that Swtzerland s a small open economy. Ths means that export prces are exogenously fxed by world prces. Publc consumpton s realzed through government demands for both mported and domestcally produced goods and servces. The commodty composton of government consumpton s assumed to be exogenously determned, Fgure 2: Structure of transformaton x (1,tot) ( ) q (1) c, = CET q (1) c,d,, q(1) c,e, q (1) c,d, q (1) c,e, ( = CET q (1) c,,..., q(1) c, ) ( ) q (1) c, = CET q (1) c,d,, q(1) c,e, q (1) c,d, q (1) c,e, 4

5 x (3,src) c ( u = LES Fgure 3: Structure of preferences x (3,src) c,..., x (3,src) c ; x(3,sub) c ( ) = CES x (3) c,s x (3) c,d x (3) c,m ),..., x (3,sub) c ( ) x (3,src) c = CES x (3) c,s x (3) c,d x (3) c,m whereas the aggregate publc consumpton s assumed to move wth aggregate prvate consumpton. Producers, nvestors and mporters are assumed to earn zero pure profts. The prce receved by producers s equal to the unt costs of producton and the prce receved by nvestors s equal to the unt costs of constructng captal. The basc prces of mports are defned as ther c..f. duty-pad prces n domestc currency. Regardng purchasers prces for each user, they are equal to the sums of basc prces and commodty taxes. Market clearng condtons mpose that supply equals demand. The supply for domestc commodtes s the sum over producers of commodtes and the demand s the sum over uses of commodtes. We assume nfnte elastcty of the supply of mports whch means that the supply of mports can be nterpreted as the percentage change n the total demand for mports. Regardng factor markets, the current supply of captal s equal to the demand n that perod for any ndustry whereas the supply of labour s equal to the aggregate demand for labour over ndustres. Captal stock avalable for use n each ndustry at the end of one perod s determned by the current captal stock deprecated at a gven rate plus and the current level of nvestment. Aggregate nvestment s endogenous to the model as well as the allocaton of nvestment across ndustres whch depends on rates of return. These are determned endogenously reflectng the nteracton of demand for captal wth exogenously gven captal supples. Fnally the model may allow for ndexaton of nomnal wage rates to the consumer prce ndex. The percentage change of the latter s defned as a weghted sum over the percentage change of the source-specfc good prces. 3 Hstorcal and Decomposton Closures Ths secton s entrely drawn from Dxon and Rmmer (2002) and Dxon and Rmmer (2003). It presents the underlyng theory of the hstorcal and decomposton closures. Detals of mplementng the hstorcal closure from the decomposton closure n SwssAGE s gven n the appendx. 5

6 For any year the representaton of the model may be expressed n the followng compact form: F (X) = 0 (1) where F s a vector of m dfferentable functons of n varables X, n > m. The varables X nclude prces and quanttes applyng for a gven year and the m equatons n 1 mpose the usual condtons for appled general equlbrum models such as: demands equal supples, demands and supples reflect utlty and proft maxmzng behavor; prces equal unt costs; and end-of-year captal stocks equal deprecated openng captal stocks plus nvestment. In usng the model we always have avalable an ntal soluton, X 0, of equaton 1 derved manly from nput-output data for a partcular year. In smulatons, we compute the movements n m varables (the endogenous varables) away from ther values n the ntal soluton caused by movements n the remanng n m varables (the exogenous varables) away from ther values n the ntal soluton. In most smulatons the movements n the exogenous varables are from ther values n one year to ther values n the next year. Correspondngly, the results for the endogenous varables refer to movements from one year to the next. However, n the hstorcal and decomposton smulatons consdered n ths study, the movements n the exogenous varables refer to changes over several years rather than one year. Thus, n these smulatons, the movements n the endogenous varables refer to changes over the entre consdered perod. In order to be able to solve the model, we must close the model, that s we have to choose whch of the n m varables have to be ncluded n the exogenous set. In a decomposton closure we nclude n the exogenous set all naturally exogenous varables,.e., varables not normally explaned n a computable general equlbrum (CGE) model. These may be observable varables such as tax rates or unobservables such as technology and preference varables. On the other hand, hstorcal closures nclude n ther exogenous set observable and assgnable varables. Observables are those for whch movements can be readly observed from statstcal sources for the perod of nterest. Usually the observables nclude a wde array of macro and ndustry varables but not ntermedate nput flows of commodty to ndustry. Assgnable varables are naturally exogenous and are therefore exogenous n decomposton closures as well as hstorcal closures. The key feature of an assgnable varable n an hstorcal smulaton s that ts movement can be assgned a value (possbly not unque) wthout contradctng anythng that we have observed about the hstorcal perod or wsh to assume about that perod. We clarfy ths concept later n ths secton n the dscusson of equaton 2. Wth reference to the two closures we can partton the varables nto four parts. Let X HD, X H D, X HD H D, and X be the set of varables n the model, where H and H denote exogenous and endogenous n the hstorcal closure, and D and D denote exogenous and endogenous n the decomposton closure, then 6

7 Table 1: Categores of varables n the hstorcal and decomposton closures X H D X HD 1 Publc consumpton by commodty Commodty composton of publc consumpton Aggregate publc consumpton Rato of prvate to publc consumpton 2 Prvate consumpton by commodty Shfts n household preferences Average taste shft Average propensty to consume out of GDP 3 Imports by commodtes Shfts n mport vs domestc preferences 4 Aggregate nvestment Unform shft n nvestment/captal ratos 5 Exports by commodtes Shfts n export vs domestc transformaton 6-7 Employment and captal nputs Prmary-factor-savng techncal change by ndustry and captal/labour bas n techncal change 8 Output by commodty Commodty-usng techncal change n output X HD Populaton C..f. mport prces and f.o.b export prces n foregn currency Polcy varables, e.g. tax and tarff rates Demands for ntermedate nputs Demands for margn servces H D X - X H D denotes the set of varables that are exogenous n the hstorcal closure but endogenous n the decomposton closure, - X HD denotes the set of varables that are endogenous n the hstorcal closure but exogenous n the decomposton closure, - X HD denotes the set of varables that are exogenous n both hstorcal and decomposton closures, and - X H D denotes the set of varables that are endogenous n both hstorcal and decomposton closures. Table 1 gves examples of the parttonng of varables used n ths model. As ndcated, varables n X HD are populaton sze, foregn currency prces of mports and polcy varables such as tax rates and tarff rates. The values of these varables are readly observable (ncluded n H) and are not normally explaned n CGE models (ncluded n D). Examples of varables n X H D are demands for ntermedate nputs and demands for margns servces (e.g. road transport) to facltate commodty flows from producers to users. In the absence of end-of-perod nput-output tables, movements n these varables are not readly observable or assgnable (not ncluded n H) and are normally explaned n CGE models (not ncluded n D). Varables n X H D nclude, at the ndustry or commodty level, outputs, employment, captal, nvestment, exports, mports, prvate consumpton and numerous prce deflators. Also ncluded are numerous macro varables such as 7

8 the exchange rate and the average wage rate. CGE models normally am to explan the effects on these varables of polcy changes, changes n technology and other changes n the economc envronment. Hence these varables are naturally endogenous,.e. they belong to the D set, and because changes n ther values can be readly observed they belong to the H set. X HD contans the same number of varables as X H D wth each varable n X H D havng a correspondng varable n X HD. These correspondng varables are predomnantly unobservable technologcal and preference varables. Such varables are not normally explaned by CGE models and are therefore exogenous n the decomposton closure. However n the hstorcal closure they are endogenous wth the role of gvng the model enough flexblty to explan the observed movements n the varables n X H D. Table 1 shows examples of correspondng pars from X HD and X H D. As ndcated n the table, n our hstorcal smulaton we use shfts n household preferences to accommodate observatons on consumpton by commodty, twsts n mport-domestc preferences to accommodate observatons on mport volumes, etc. The prncples underlyng the four-way parttonng of the varables n the hstorcal and decomposton closures can be clarfed by an example. Let total ntermedate demand of commodty c, X c, be represented by the followng equaton: X c = B c, B c Z (2) where Z s the actvty level (overall level of output) n ndustry ; and B c, B c s the nput of commodty c per unt of actvty n ndustry wth B c, and B c beng technologcal varables whch can be used n smulatng the effects of changes n the nput of commodty c per unt of actvty n ndustry and the nput of commodty c per unt of actvty n all ndustres. In decomposton mode, B c, and B c are exogenous and Z and X c are endogenous. Suppose that movements n the actvty levels, Z, are not observed but that we have observed the movements over an hstorcal perod n total ntermedate demand, X c (possbly from nformaton on commodty outputs, mports and fnal usage). Suppose that we wsh to assume unform nput-cusng techncal change. Then n hstorcal mode we can use movements n B c to explan observed movements n X c and we can assgn a unform value (possbly zero) to the percentage movements n B c, for all ndustry. In ths example, Z s a member of X H D and the assgnable varable B c, s a member of X HD. X c s a member of X H D and B c s the correspondng member of X HD. Havng allocated the varables to the four categores, we can compute hstorcal and decomposton solutons, startng wth the hstorcal soluton of the form: X H = G H (X H ) (3) where X H and X H are the exogenous and endogenous varables n the hstorcal closure,.e. X H = X HD X H D and X H = X HD X H D, and G H s an m-vector of dfferentable functons. By observng and assgnng X H for two years, s and t, we can use equaton 3 to estmate percentage changes over the 8

9 nterval [s,t], x H, n the varables n X H. Thus we combne a large amount of dsaggregated nformaton on the economy (the movements n the varables n X H ) wth a CGE model to estmate movements n a wde varety of technologcal and preference varables, X HD, together wth movements n more standard endogenous varables, X H D. Next we move to the decomposton closure whch gves a soluton of the form: X D = G D (X D ) (4) where X D and X D are the exogenous and endogenous varables n the decomposton closure,.e. X D = X HD X HD and X D = X H D X H D, and G D s an m-vector of dfferentable functons. Followng the method poneered by Johansen (1960), we can express equaton 4 n log-dfferental or percentage change form as x D = B x D (5) where x D and x D are vectors of percentage changes n the varables n X D and X D, and B s an m by n m matrx n whch the j-th element B j = GD (XD ) Xj D Xj D X D (6) s the elastcty of the -th component of X D wth respect to the j-th component of X D. Wth the completon of the hstorcal smulaton, the percentage changes n all varables are known. In partcular the vector x D s known. Thus we can use equaton 5 to compute values for x D over the perod s to t. The advantage of workng wth equaton 5 rather than equaton 4 s that the former can be used to gve a decomposton of the percentage changes n the varables n X D over the perod s to t nto the parts attrbutable to movements n the varables n X D. Ths s a legtmate decomposton to the extent that the varables n X D are genunely exogenous, that s, can be thought of as varyng ndependently of each other. In settng up the decomposton closure, the exogenous varables are chosen wth exactly ths property n mnd. Thus, among varables n X D, we fnd polcy varables, technology varables, taste varables and nternatonal varables (e.g. foregn currency prces) all of whch can be consdered as ndependently determned and all of whch can be thought of as makng ther own contrbutons to movements n endogenous varables such as ncomes, consumpton, exports, mports, outputs, employment and nvestment. In ths study, we use the hstorcal closure, developed n the next secton, n estmatng changes n technology and tastes varables. Usng then the decomposton closure, we can compute, va equaton 5, the contrbuton of the movement n the j-th exogenous varable to the percentage movement n the -th endogenous varable Θ j = B j x D j (7) as the product of B j, the elastcty of the -th component of X D wth respect to the j-th component of X D, wth the percentage change n the j-th element of 9

10 x D. Because equaton 4 represents a non-lnear system of equatons, the effect on endogenous varable over the perod s to t of movements n exogenous varable j cannot be computed unambguously. The effects of movements n an exogenous varable depend on the values of other exogenous varables. In terms of equaton 7, the problem s to decde at whch values of the exogenous varables to evaluate B j. The most natural choce s to use md-pont values to evaluate the elastctes defned n equaton 6. In the decomposton analyss n ths study, we use a procedure due to Harrson, Horrdge, and Pearson (2000) whch, n effect, evaluates B j as the average of the values generated as we move the exogenous varables n the decomposton smulaton n small steps along a straght lne from ther values n year s to ther values n year t. 4 Hstorcal smulaton Hstorcal smulatons allow the estmaton of technology and preference varables (components of X HD ) once nformaton on observable varables have been ncorporated nto the model (components of X H D). As explaned n the prevous secton, ths means movng from the decomposton closure to the hstorcal closure or n other words endogenzaton of naturally exogenous varables and exogenzaton of naturally endogenous varables. As wll become apparent, the hstorcal closure s complcated and unusual. We are gong to develop the hstorcal closure n a seres of 8 steps. In ths process, the naturally endogenous varables for whch we have data are cumulatvely exogenzed. The steps are desgned n order to have a vald closure at the end of each step. We are thus able to perform an hstorcal smulaton at the end of each step. Comparson of results for successve smulatons show the effects of the addtonal data ntroduced at each step. Results from ths step-by-step approach appled for 1990 to 2001 are shown n table 3. At the end of all the steps, hstorcal smulatons gve estmates of technology and preference varables whch are reported n table 4. For the sake of clarty, n developng the hstorcal closure 5, we are gong to use a sketch model of SwssAGE. It wll be also useful for understandng the broad features of the results. No sectoral dsaggregaton s brought n the sketch model whereas the appled model s currently run wth 14 ndustres. The sketch model s presented n table 2. There are 8 equatons and 8 endogenous varables. X H D denotes the set of varables that are exogenous n the hstorcal closure but endogenous n the decomposton closure whereas X HD denotes the set of varables that are endogenous n the hstorcal closure but exogenous n the decomposton closure. At each step a naturally endogenous varable n the set X H D s swapped wth ts correspondng naturally exogenous varable n the set X HD. Thus the naturally endogenous varable becomes exogenous and ts correspondng naturally exogenous varable becomes endogenous. Equatons n table 2 are lsted n the order of the steps we are gong to go through. We label them accordng to ther correspondng step number. Equaton S-0 of the sketch model s the GDP dentty n constant-prce terms. The next equaton defnes the rato ψ C/G of prvate consumpton C to publc 5 The development of the hstorcal closure n SwssAGE s gven n the appendx. 10

11 consumpton G. Equaton S-2 relates the sum of prvate consumpton and publc consumpton to GDP va the average propensty to consume ψ C/GDP. The followng equaton relates mports M to GDP, the exchange rate φ and to an mport/domestc preference varable Υ src. Equaton S-4 defnes the rato ψ I/K of nvestment I to captal K. The next equaton relates exports X to the exchange rate and to an export/domestc preference varable Υ dst. Equaton S-7 relates the captal/labour rato to the exogenous rate of return ROR, a technology shft varable A, the exchange rate and to a captal/labour twst varable Υ prm. Ths equaton s derved assumng that the value of margnal product of captal equals the rental on captal, R, or equvalently, MPK = A R P K P K P P f 1/A = R (8) K PK = A ROR P where P K and P are the prce ndexes for captal goods and for domestc goods. The rate of return s vewed as the rato of the rental prce of captal to the asset prce. Recognzng that the margnal product of captal, MPK, s a monotoncally decreasng functon of the captal/labour rato, technology has a negatve mpact on ths rato. The rato of the cost of unts of captal to the domestc prce ndex can be vewed as a measure of real devaluaton of the Swss franc, snce the former ncludes mport prces but not export prces whereas the latter ncludes export prces but not mport prces. In equaton 9, the real devaluaton s nterpreted as a decreasng functon of the exchange rate whch means that t has a postve mpact on the captal/labour rato. The last equaton of the sketch model s the producton functon relatng real GDP to prmary factors, captal K and labour L, and to the technology shft term. Labour s assumed to be fxed. Step 0 - Naturally exogenous varables A prelmnary step to the 8 steps s to shock all the varables that are exogenous n both the hstorcal and decomposton closures (the set of varables X HD lsted n table 1) for whch we have nformaton on ther actual movements across the perod 1990 to These varables are populaton and tarff rates. In ths step we ntroduce also nformaton on the general prce level. In terms of the sketch model ths means that we have to exogenze the prce ndex P and to endogenze the exchange rate φ. The prce ndex s thus now consdered as the numerare. The frst column n table 3 reports the mpact of these shocks on the man macro varables of the model. Shock to populaton has no mpact on the economy as we assume for the present study a untary ncome elastcty for consumers n the hstorcal smulatons. The decrease n tarff rates from 1990 to 2001 has lttle mpact on the economy as the major decrease comes from agrculture whch has an output share n total producton less than 2%. The ncrease n the absolute prce level has no mpact on real varables but only on (9) 11

12 Table 2: Sketch model Step Equaton X H D X HD 0 GDP = C + I + G + X M P φ 1 C/G = ψ C/G G ψ C/G 2 C + G = ψ C/GDP GDP C ψ C/GDP 3 M = m GDP, P M /φ, Υ src M Υ src wth m GDP > 0, m φ > 0 and m Υ > 0 4 I/K = ψ I/K I ψ I/K 5 X = x P X /φ, Υ dst X Υ dst wth x φ < 0 and x Υ > 0 φ A 7 K/ L = k ROR, 1/A, φ, Υ prm K Υ prm wth k A < 0, k φ > 0 and k Υ < 0 8 GDP = 1/A f(k, L) GDP A ID wth f A < 0 and f K > 0 the dfferent prces of the economy. In partcular, the ncrease n prce ndex decreases approxmately by the same amount the exchange rate snce t s defned n foregn currency per unt of Swss franc. Step 1 - Publc consumpton In the frst step, we ntroduce nformaton on the structure and overall quantty of publc consumpton. Ths means that we observe the commodty composton of publc consumpton. However, as there s only one domestc good n the sketch model, only the movement n aggregate publc expendture can be used as example. To accommodate ths nformaton, government spendng G are exogenzed and the rato of real prvate to real publc consumpton ψ C/G s endogenzed. Results n the second column of table 3 show the effects of the shocks from step 0 plus the shock to publc consumpton appled n ths step. We can see that the 16.14% ncrease n aggregate real government spendng from 1990 to 2001 generates a decrease n prvate consumpton of 2.34% (from 0.02% to -2.32%) and an extra 0.86% ncrease n aggregate captal (from 0.21% to 1.07%). The former s explaned by the constant average propensty to consume out of GDP and the sze of prvate consumpton relatve to publc consumpton. The reason for the latter s hghly captal-ntensve government spendng relatve to prvate consumpton expendture. Ths frst step n hstorcal smulatons allows us to estmate a frst varable that s naturally exogenous but unobservable. As menton earler, the rato of real prvate to real publc consumpton ψ C/G s endogenous and ts resultng value n ths step s equal to %. Ths large value s explaned manly by 12

13 Table 3: Macro results from hstorcal smulatons for 1990 to 2001 (percentage changes) Varables Step 0 Step 1 Step 2 Step 3 Step 4 Step 5 Step 6 Step 7 Step 8 Real GDP Real nvestment Real prvate consumpton Real publc consumpton Real exports Real mports Aggregate employment Aggregate captal Average prm-fac-usng tech change Real devaluaton Exchange rate Prce deflator for GDP Prce deflator for nvestment Prce deflator for consumpton Prce deflator for publc consumpton Real wage for consumers APC out of GDP C/G rato I/K rato

14 the large ncrease n government spendng and to a less extend by the small decrease n prvate consumpton. Step 2 - Prvate consumpton The second step allows us to mpose the change over the perod 1990 to 2001 n the commodty structure and overall quantty of prvate consumpton. Therefore consumpton by commodty and aggregate consumpton s exogenzed, whch enable us to endogenze consumer preferences and to free the lnk between the sum of prvate and publc consumpton and GNP. In terms of the sketch model, ths means exogenzaton of aggregate prvate consumpton C and endogenzaton of the average propensty to consume ψ C/GDP as there s only a sngle locally produced and consumed good. The nformaton on movements n consumpton by commodty has a strong postve effect on aggregate prvate consumpton (from -2.77% to 15.74%) as commodty expendture ncrease for most of them. The ncrease n global consumpton has mnor effects on the components of the GDP dentty except from exports on whch t has a major negatve mpact. The declne n exports relatve to mports mples a deteroraton of the balance of trade and a decrease n the exchange rate. As the prce ndex for captal goods ncreases relatve to the prce ndex for domestc goods, aggregate captal decreases whch reduces aggregate nvestment. Hstorcal smulatons generate estmates of changes n household preferences for a gven commodty. They are resduals of the dfference between the effectve percentage change n consumpton per household of a gven commodty and the percentage change mpled by the model gven movements n household consumpton, prces and ncome for 1990 to In other words they gve a measure of changes n household preferences not explaned by theory embedded n the model. An estmaton of 1 for a gven commodty means a 1 percent growth rate of consumpton per household for ths good hgher than would be expected on the bass of changes n total expendture per household and changes n prces. Results of the estmaton of changes n household preferences between 1990 and 2001 are reported n column 1 n table 4. It gves the contrbuton of a change n household preferences for a gven commodty to growth n output of ths commodty. These results quantfy the strong shft n household preferences n favour of health and nsurance across the perod 1990 to 2001 whch appear wth a large postve entry. On the other hand, households reduce ther nterest between 1990 and 2001 n prmary goods, hotels and restaurants as well as n manufacturng products. Overall, there s a shft n households preferences between 1990 and 2001 from prmary and secondary sectors towards servces sectors. Fnally, n ths step we know that aggregate prvate consumpton ncreases across the perod 1990 to Ths has naturally an mpact on the rato of real prvate to real publc consumpton estmated n the prevous step. As prvate consumpton declnes relatve to publc consumpton between 1990 and 2001, the hstorcal smulaton generates a negatve value for the percentage change 14

15 Table 4: Results on commodtes from hstorcal smulatons for 1990 to 2001 Contrbuton (%) to Swss output growth of changes n H hold pref. Interm. demand M/D rato X/D rato Prmaryfactor techncal progress Captal/ labour bas (1) (2) (3) (4) (5) (6) Prmary goods Manufacturng products Energy Constructon Wholesale and retal trade Hotels and restaurants Transport and communcaton Fnancal ntermedaton Insurance Busness servces Publc admnstraton Educaton Health Other servces n the rato of real prvate to real publc consumpton. Step 3 - Imports In ths step we observe the movement n aggregate mports by commodty. The mport observatons are therefore accommodated by endogenzaton of twst varables n the mport/domestc preferences of ndustres, captal creators and households. Ths means n terms of the sketch model exogenzaton of mports M and endogenzaton of the mport/domestc preference varable Υ src. Applcaton of the shocks for mports by commodty results n a large ncrease of aggregate mports. Ths ncrease n mports has lttle effect on GDP, captal and nvestment. As the sum of prvate and publc consumpton s a fxed share of GDP, the man effect of the decrease n mports s a large ncrease n exports (from % to 8.91%). Ths ncrease n exports s accompaned by a decrease n the exchange rate. Estmates of mport/domestc preference varables are computed as the dfference between the actual percentage change n the mport/domestc rato and the percentage change mpled under the theory. In addton to standard theory, the latter allows for demand pressures. It captures the dea that when output of a gven commodty n the domestc economy s growng rapdly, there s a tendency for demand shfts to occur towards mports. Ths s explaned by shortages and lengthenng queues and s unrelated to movements n relatve prces. Smlarly, when output of a gven commodty s growng slowly there s a tendency for shfts to occur towards the domestc product. Estmates of mport/domestc twsts take account of changes that can not be explaned by changes n relatve prces and demand pressures. 15

16 Results n column 3 of table 4 show the percentage contrbutons of the mport/domestc preference shfts to sales n Swtzerland of domestc commodtes. Most commodtes have postve entres for the perod 1990 to Ths means that twsts aganst mports occur for these commodtes. The largest bas n favour of the domestc varety occurs for manufacturng products. Prmary goods experence a large twst n preferences also towards domestc goods, as t s the case for hotels and restaurants, transport and communcaton. Bases n favour of mports across the perod 1990 to 2001 are present only n nsurance and other servces. Step 4 - Investment The fourth step s to ncorporate n the model nformaton on nvestment between 1990 and The lack of data does not allow us to ntroduce movements on nvestment n each ndustry. Ths means that we can te down only aggregate nvestment I. Ths s done by allowng the rato of nvestment to captal ψ I/K to be determned endogenously. The ntroducton of the small ncrease n aggregate nvestment for 1990 to 2001 has no mpact on GDP and captal snce labour, technology and rates of return are assumed to be fxed. In addton, the ncrease s so small that, under consumpton lnked to GDP and mports gven n the prevous step, t does not affect exports and the exchange rate. In ths step, hstorcal smulatons produce an estmate of the rato of nvestment to captal between 1990 and Its value of 0.03% reflects the decrease n aggregate captal stock relatve to nvestment. Step 5 - Exports Here we use nformaton on exports. Ths allows us to exogenze exports by commodty and to endogenze the twst varables n the export/domestc technology of ndustres. In terms of the sketch model ths means exogenzaton of exports X and endogenzaton of the export/domestc technology varable Υ dst. However, n usng export observatons to estmate the shft varable exports versus locally used goods, we found ndetermnacy between ths shft term and the exchange rate. The problem s that an ncrease n exports can be explaned ether by an ncrease n the export/domestc twst shft varable or by an decrease n the exchange rate. However an decrease n the exchange rate mples an decrease n captal and so a decrease n GDP whch creates an nfeasblty n the GDP dentty. The soluton to ths problem s to ntroduce nformaton on the exchange rate. In order to be able to exogenze the exchange rate, we need to gve the supply of export freedom to make the observed movement n exports compatble wth the demand functon for exports. Ths requrement s satsfed through the endogenzaton of a technology varable. In terms of the sketch model we need thus to exogenze the exchange rate φ and endogenze the prmary-factor-usng techncal progress varable A. Comparson of steps 4 and 5 n table 3 shows that movements of exports by commodty are accommodated by both a substantal demand shft reflected 16

17 n huge twsts n favour of exports (see below) and a supply shft reflected n prmary-factor-usng techncal change 6 of -9.71%. The ncrease of 9.71% n technologcal progress and the ncrease of 10.31% n captal, nduced by a apprecaton of the Swss franc and by the mprovement n technology, contrbute both of them to the ncrease of 13.18% n GDP. The large ncrease n captal has also an mpact on real wage for consumers (from 1.53% to 13.28%) and on the rato of nvestment to captal (from 0.03% to -9.58%). The former rses reflectng the ncrease n the captal/labour rato and the consequent ncrease n the margnal product of labour. Regardng the latter, t s negatve ndcatng the rapd growth n aggregate captal relatve to aggregate nvestment across the perod 1990 to As n the case of mports, the export/domestc technology varable acts lke a resdual. Estmaton of ts value s the dfference between the actual percentage change n the export/domestc rato and the percentage change mpled under the standard theory. A postve value means thus a twst towards exports not explaned by changes n relatve prces. The percentage effects on output growth of the export/domestc technology shfts are gven n column 4 of table 4. Postve entres mean twsts n favour of exports from from 1990 to Wth one excepton, results show a twst towards exports for all commodtes. These upward shfts n the export demand curves are consequence of growth n the rest-of-world economy. The excepton s prmary goods whch shows a small twst toward domestc products. A possble explanaton s the large reducton of over 60% n export subsdes under the reform of the agrcultural polcy. The contrbuton of the twst toward exports to output growth s hghest for manufacturng products. Large twsts towards exports occur for hotels and restaurants, as well as for nsurance. Fnancal ntermedaton, transport and communcaton, wholesale and retal trade, and energy are examples of twsts towards exports across ths perod but of a smaller ntensty. Step 6 - Employment In ths step we ntroduce nformaton on employment. In terms of the sketch model ths rases no macro closure ssues snce employment s already exogenous. However n the dsaggregated model, observatons on employment by ndustry allow estmaton of shfts at the ndustry level n prmary-factor techncal change. On comparng the results n table 3 for step 5 wth those for step 6, we see that the mposton of the labour shocks ncreases aggregate employment substantally. From all the prevous steps we know that the ncrease n employment can have lttle effect on GDP snce all ts components are approxmately fxed. As GDP does not change, the prmary-factor techncal change ncreases from -9.71% to -6.73%. If t were to decrease, t would contradct equaton S-7 6 A negatve value of one for the change n the technology varable decreases prmary-factor nputs per unt of GDP by one per cent. Wth prmary-factor nputs held constant, GDP ncreases by one per cent. A negatve change for the technology varable has thus a postve mpact on GDP and on captal. 17

18 as the captal decrease requred to keep GDP constant n equaton S-8 mples an ncrease n the technology varable. Although the prmary-factor techncal change ncreases, the mpact on captal from steps 5 to 6 s postve due to the suffcently large ncrease n aggregate employment whch makes the change n the rato of captal to labour negatve. Fnally, the ncrease n employment has a negatve mpact on the margnal product of labour and thus on real wage for consumers whch decreases from 13.28% to 8.87%. Column 5 n table 4 reports a measure of prmary-factor techncal progress. For each commodty, t s defned as the ncrease n output equal to the share of prmary factors n a gven ndustry s costs multpled by the value of the techncal change. The latter accounts for the dfference between the actual percentage change n prmary-factor nput to a gven ndustry and the percentage change n ths ndustry s output 7. Wth two exceptons, results show a postve entry ndcatng techncal progress. The largest ncrease s n fnancal ntermedaton (18.61%) followed by other servces (12.22%). Prmary goods, wholesale and retal trade, and health experence techncal progress around 8.5% from 1990 to To a less extend, technologcal progress s also present n manufacturng products, energy, constructon, transport and communcaton, as well as n educaton. Step 7 - Captal The ntroducton of movements on captal n each ndustry s the focus n ths step. Captal varaton n each sector s accommodated by allowng a captal/labour technology twst varable to move. In terms of the sketch model we requre exogenzaton of K and endogenzaton of Υ prm. A postve value wll favour labour whereas a negatve value wll favour captal. Captal/labour twsts n technology are cost neutral n the sense that they do not affect a gven ndustry s total nputs per unt of actvty. In the sketch model, exogenzaton of captal amounts to turn off equaton S-7 by the endogenzaton of ths twst varable. Results for step 7 n table 3 show that the applcaton of the captal shocks decrease aggregate captal by 10.34% compared to step 6. As GDP and labour are fxed, the decrease n captal wth respect to the prevous step mples an ncrease n techncal progress and an ncrease n real wage for consumers. Ths s reflected n the results by the decrease n the technology varable from -6.73% n step 6 to -9% n step 7 and by the ncrease from 8.87% to 13.39% n real wage. Captal-labour bases n technology act n a smlar way to mport-domestc bases n preferences or to export-domestc bases n transformaton technology. They account for the dfference between the actual percentage change n the captal/labour rato n a gven ndustry and the percentage change explaned by the movements n ths ndustry s unt costs of usng labour and captal. A negatve value means that the ndustry s technology changes so that at any 7 The model assumes a Leontef producton technology to combne each commodty nput and prmary-factor nput. Ths means that prmary-factor composte s not senstve to factor prces. 18

19 gven rato of the wage rate to the rental rate on captal, the ndustry chooses a captal/labour rato hgher n 2001 than n From 1990 to 2001, the last column of table 4 reports huge twsts n technology n favour of labour for educaton and busness servces. A plausble cause s the low level of nterest rate n Swtzerland durng ths perod. Large twsts favourng labour occur also n hotels and restaurants, constructon and other servces. On the other hand, manufacturng products, transport and communcaton, and nsurance present large twsts n technology n favour of captal across the perod 1990 to Step 8 - Producton The last step allows the model to take account of nformaton on output changes by commodty 8. Shocks to supply of domestc good are absorbed by a unform (across ndustres) endogenously determned movement n a commodty-usng techncal change varable for current producton and captal creaton. Ths approach allows the model to ncrease the use of commodtes that experence rapd growth n supply and conversely reduce the use of commodtes that experence slow growth n supply. However, n order to avod ndustres expandng (reducng) ther use of goods per unt of output wthout reducng (ncreasng) ther use of other nputs per unt of output, we allow for a cost-neutralzng endogenous movement n an all-nput-savng technology varable. Ths means for nstance that a 20 per cent nput-usng techncal change n a gven commodty requred to absorb the observaton on output mples a 2 per cent all-nputsavng techncal change n a gven ndustry f 10 per cent of ths ndustry s costs were accounted for by nputs of the commodty. Results of ths fnal step n the last column of table 3 show almost no movement n aggregate varables. From all the prevous steps, GDP, labour and captal are fxed leavng the addtonal nformaton from ths step to be embedded n the commodty-usng techncal change varable. Column 2 of table 4 shows the contrbutons to growth n sales of domestc products of nput-usng techncal change n producton and captal creaton. Estmate of nput-usng techncal change accounts for the dfference between the actual percentage change n the sales of a gven commodty to the domestc ndustres and the percentage change explaned by movements n producton and captal-creatng actvtes of ndustres usng ths commodty. Results show that the use per unt of output of prmary goods, constructon, hotels and restaurant, health and other servces declne substantally from 1990 to However commodty-usng techncal change substantally stmulates output growth of fnancal ntermedaton and nsurance across ths perod. To a less extent, t s also the case for energy as well as for transport and communcaton. 8 Although ntermedate demand s not explctly represented n the sketch model, ths fnal step would mean exogenzaton of GDP and endogenzaton of a techncal change varable n ntermedate demand, A ID. 19

20 5 Decomposton smulaton As explaned n secton 3, decomposton smulatons allows us to decompose hstory nto the parts attrbutable to changes n varables such as those dentfed n table 5. Most of these varables (e.g. technology and taste varables) are naturally exogenous varables and changes n these varables across the perod 1990 to 2001 are estmated n hstorcal smulatons. Applyng shocks n the decomposton smulaton dentcal to theses estmated changes n naturally exogenous varables generate values for naturally endogenous varables dentcal to those specfed n the hstorcal smulaton. However, snce technology and taste varables are exogenous n the decomposton smulaton, questons about the effects of changes n these varables can be answered. The decomposton analyss s presented frst n terms of the macroeconomc results. In the frst subsecton they are explaned accordng to each relevant group of varables lsted n column n table 5. The second subsecton gves an analyss across theses groups of varables to fnd out factors underlyng the macroeconomc results. The last subsecton presents sectoral results of the decomposton analyss. 5.1 Macroeconomc results Macroeconomc results of the decomposton smulaton are shown n table 5. Column 1 through 9 gve the contrbuton of the group of varables descrbed n each column headng to the total percentage change (the last column) n each of the endogenous varable dentfed n the frst column. Therefore the sum of these contrbutons gves the total percentage change n the last column for each endogenous varable. We start by lookng at each man group of exogenous varables 1 to 9 ndvdually. We therefore omt rrelevant groups of varables such that changes n tarffs (column 2) and changes n publc expendture (column 7). Changes n export/domestc technology The frst column n table 5 gves the effects of shfts n export versus domestc technology. Estmates of these shfts are provded n hstorcal smulatons and show, wth one excepton, a twst towards exports for each commodty (column 4 n table 4). Ths mples naturally an ncrease n aggregate exports whch leads to an apprecaton of the Swss franc. As producton s now drected more towards exports, mports rse, nduced by evaluaton of the exchange rate, to satsfy domestc demand. The largest twst n favour of exports are n manufacturng products whch accounts for 70% of aggregate exports. Because they are labour ntensve relatve to producton of commodtes enjoyng favourable shfts towards domestc consumpton, there s a small reducton n aggregate captal and as well as n aggregate nvestment. Another effect of the technology shft towards exports s the small reducton n real prvate and publc consumpton. The explanaton les n the negatve change n nomnal GDP due to the larger decrease n ts prce deflator than 20