STRATHCLYDE DISCUSSION PAPERS IN ECONOMICS CAN MIGRANTS SAVE GREECE FROM AGEING? A COMPUTABLE GENERAL EQUILIBRIUM APPROACH USING G-AMOS.

Transcription

1 STRATHCLYDE DISCUSSION PAPERS IN ECONOMICS CAN MIGRANTS SAVE GREECE FROM AGEING? A COMPUTABLE GENERAL EQUILIBRIUM APPROACH USING G-AMOS. BY NIKOS PAPPAS NO DEPARTMENT OF ECONOMICS UNIVERSITY OF STRATHCLYDE GLASGOW

2 Can Mgrants save Greece from Ageng? A Computable General Equlbrum Approach usng G-AMOS. Pappas, N. Fraser of Allander nsttute, Department of Economcs, Unversty of Strathclyde n Glasgow, UK Abstract: The populaton of Greece s projected to age n the course of the next three decades. Ths paper combnes demographc projectons wth a mult-perod economc Computable General Equlbrum (CGE) modellng framework to assess the macroeconomc mpact of these future demographc trends. The smulaton strategy adopted n Lsenkova et. al. (2008) s also employed here. The sze and age composton of the populaton n the future depends on current and future values of demographc parameters such as the fertlty, mortalty rates and the level of annual net mgraton. We use FIV-FIV software n order to project populaton changes for 30 years. Total populaton and workng age populaton changes are ntroduced to the G- AMOS modellng framework calbrated for the Greek economy for the year Postve net mgraton s able to cancel the negatve mpacts of an ageng populaton that would otherwse occur as a result of the shrnkng of the labour force. The polcy mplcaton s that a vable, long-lastng mgraton polcy should be mplemented, whle the mportance of polces that could ncrease fertlty should also be consdered. JEL classfcaton: J11, J21 Key words: CGE modellng, ageng populaton, mgraton, demography, Greece Acknowledgements: I would lke to thank Grant Allan, Karen Turner, Peter McGregor and Km Swales of the Fraser of Allander Insttute for useful comments. I would also lke to thank Katya Lsenkova for comments on demographc modellng. The usual dsclamer apples. 2

3 1. Introducton As n most western countres, populaton n Greece s projected to declne and age. Greece has tradtonally been a country of out-mgraton. However, n the last two decades the populaton outflow has reversed. Ths has generated a stock of foregn populaton, whch s estmated to be close to 10% of the natonals 1 (Konts et. al., 2006). Ths has been a real demographc shock. In contrast, the country has experenced a declnng fertlty rate (1.3 n 2004) and a contnuously ncreasng lfe expectancy. Fertlty rates n Greece have been lower than the European average untl the end of the 60 s and peaked durng the 70 s but never acheved the levels reached n the baby boom perod n Europe. The baby boom n Greece was late and small due to the poltcal nstablty of the post World War II perod. However, the fertlty rate has largely converged to the European average durng the last two decades. If the assumptons of ESYE (Natonal Statstcs Agency) on future trends of demographc varables (fertlty, mortalty, mgraton) materalse, then n the course of the next three decades (by 2034) s projected to ncrease slghtly by 0.54%. However at the same tme workng age populaton s expected to decrease by 6.13%. Under the same assumptons but wthout the presence of mgraton total populaton wll fall by 11.5% and workng age populaton by 19.76%. We are gong to try and analyse the mpacts, of these future demographc trends. At present very lttle attenton s pad to the macroeconomc mpacts that these wll have n the foreseeable future. Exstng lterature focuses on the mpacts of mmgraton upon the economy. Lanos (1996) found that wages of (llegal) mmgrants n Northern Greece were 40-60% lower than the ones of the local populaton. He also found that the vast majorty of mmgrants take jobs that the locals would never do. Ths study s rather descrptve, and thus does not assess the mpact of mmgraton on macroeconomc varables n a quantfable way. Sarrs and Zografaks (1999) use a Computable General Equlbrum model to assess the mpact of llegal mmgraton upon the Greek economy. In ther exercse, populaton stock s only adjusted by a one shot nflux of llegal mmgrants n the Greek economy. They fnd that almost a thrd of the 1 Contrary to many countres Greece defnes foregners by blood and not by brth. So, there are people that have been born n the country but have not got ctzenshp. In contrast, others have not been born and brought up n the country but are granted ctzenshp. So, when we refer to foregners we may nclude people that have been born n the country. 3

4 populaton s adversely affected from llegal mmgraton. However, the negatve mpact dsappears when they model labour market rgdtes. (Fxed nomnal wage). The paper at hand adds to the relevant lterature n lnkng demographc projectons wth consstent economc modellng. Thus t sheds lght on the effect of current and future fertlty, mortalty and mgraton trends wth the use of consstent economc and demographc modellng. Whle demographc changes have many economc mplcatons that are related to publc expendture, expected ncrease n demand for health care, fundng of pensons of a growng aged populaton etc., the focus of ths paper s to assess the mpact of future demographc trends upon the economy through the consequent changes n the labour market. We combne a demographc model, (FIV-FIV) and a Computable General Equlbrum (CGE) model of the Greek Economy (G-AMOS) calbrated for the year The projecton and smulaton horzon s three decades, a span that s long enough to allow for the populaton changes to be fed nto the rest of the economy and at the same tme not too long to make our populaton projectons completely unrelable. The remander of the paper s structured as follows. The next secton provdes the theoretcal framework upon whch our analyss s based. The two followng sectons provde detals on the demographc and the economc model. The economc mpact of the base projecton s dscussed n secton 5. Secton 6 focuses on senstvty analyss that we have performed by varyng values of demographc as well as economc parameters. Secton 7 concludes. 2. Theoretcal settng In ths secton we outlne the theoretcal foundaton for the smulaton results that are reported n the subsequent sectons. The focal pont s the labour market, whch s analysed usng an aggregate labour demand and supply framework (Lsenkova et.al., 2008). We present a long run approach, n order to dentfy the drecton n whch we expect the equlbrum wages and employment to move as the tme passes. Fgure 1 represents the nteracton of the labour supply and general equlbrum labour demand curves n the unfed Greek labour market, a unfed labour market, n whch workers can freely move between the dentfed sectors. When the workng age populaton s unchanged (n our case ths s equvalent to an unchanged labour force), the quantty 4

5 2 of labour suppled can only ncrease f the real wage s ncreased. The general equlbrum Labour demand s slopng downwards and ths s the case for most of the research that has utlsed the AMOS modellng framework (Gllmartn et. al. 2007, Lsenkova et. al. 2008, Turner, 2002). However, for combnatons of extreme product demand and factor substtuton elastctes, the general equlbrum labour demand curve can be upward slopng (McGregor et al, 1995). Note that the endogenously determned level of household ncome wll also affect labour demand n the general equlbrum. Fgure 1- Impact of an ageng and declnng populaton Wage NS 2034 NS 2004 W n, 2034 B W n, 2004 A ND 2004 ND 2034 N 2034 N 200 Employment In Fgure 1 we compare the long-run labour market equlbrum n 2034 wth that n 2004, under the assumpton that the base populaton projecton apples and that real per capta expendture of the Greek government remans constant. As mentoned before n the smulaton exercse there s no demand shock mposed assocated wth the ageng of the populaton. Instead, the assumpton s that government expendture per capta remans constant. One would expect that consumer and government demand patterns are age specfc and therefore shftng demographcs would alter the ndustral composton of consumer and government demand. However, for now we wll assume that government expendture per head remans constant over the smulaton perod. That s government expendture follows the changes of total populaton. The ntal equlbrum s represented by pont A, where the base-perod 2 We actually use a wage curve specfcaton for the labour market. (Blanchflower and Oswald, 1994; Layard et a ). 5

6 (2004) labour demand and supply curves ND 2004 and NS 2004 ntersect. Ths generates the ntal equlbrum employment and real wage levels w n,2004, N As mentoned before, over the perod up to 2034, the total populaton n Greece wll ncrease margnally but at the same tme age. Frst, the lower workng age populaton reduces labour supply at each wage level generatng an nward shft of the labour supply curve. The new labour supply curve NS 2034, s on the left of the orgnal labour supply curve NS On the other hand total populaton changes wll affect labour demand through changes n government expendture. The new demand curve s on the rght of the orgnal because durng ths perod government expendture (an element of fnal demand) ncreases n lne wth the slght ncrease of total populaton. The new equlbrum B, s at the ntersecton of the general equlbrum labour demand and supply curves ND 2034 and NS Both labour supply and labour demand shft have pushed the real wage upwards. However, whle ncreased labour demand tends to ncrease employment whle decreased labour supply tends to decrease t. We expect the supply effect to be of greater magntude n ths case. Fgure 2 - The Labour Market n 2034 comparng base projectons and hgher postve net mgraton every year Wage NS BASE B NS MIG W BASE W MIG C N BASE N MIG ND BASE ND MIG Employment In Fgure 2 we expand our nvestgaton to show the mpact on the labour market n 2034 of ncreased nward mgraton. Fgure 2 compares the labour market n 2034 under two dfferent scenaros. One smply takes the base populaton projecton. The second represents a stuaton n whch there s hgher postve net n-mgraton. Ths s 6

7 a polcy parameter that can be altered mmedately at the wll of the Greek government 3. The stuaton under the base projecton s gven by pont B, whch s the ntersecton of the labour demand and supply functons NS 2034, ND Ths was the fnal poston llustrated n Fgure 1. Now mpose net n-mgraton at a rate hgher than that assumed under base scenaro. Workng age populaton, and therefore the labour force, s gong to be hgher than under the base predcton for Ths s llustrated n Fgure 2 by an outward shft of the labour supply curve to NS MIG. Labour demand wll also shft outwards to ND MIG as a result of hgher populatonlnked government expendture. The new equlbrum s at pont C, the ntersecton of the new labour supply and demand curves. In ths case the shfts n labour demand and supply both work to ncrease equlbrum employment (compared to the equlbrum at B). Further, because we expect that the ncrease n labour supply wll domnate the ncrease n labour demand, pressure n the labour market should ease and the wage fall. The two fgures exhbt the mpact of populaton ageng and n-mgraton on the labour market. We expect these changes to have major mpacts through a tghtenng of the labour market, and consequences upon compettveness. These labour market changes wll have major mpacts on the economy overall. In partcular, wth an ageng populaton, we expect a negatve effect upon compettveness due to the ncreased wage, whch s a major cost of producton. Ths reduced compettveness has a negatve mpact on GDP. In mgraton s expected to counteract the tghtenng of the labour market (Sarrs and Zografaks, 1999). The Computable General Equlbrum framework we use allows us to explore such mpacts n greater detal by consstently modellng all the complex nteractons wthn the economy whle t gve us a suffcent degree of detal n terms of results. 3 An extensve dscusson of the current mgraton polcy n Greece can be found n Konts et. al. (2006) as well as n Katrougalos et.al. (2004) 7

8 3. The FIV-FIV software The software produces populaton projecton by usng the cohort component method, whch s the most wdely used. The cohort component method tracks each cohort and ts mortalty, fertlty, and mgraton over tme. Startng wth a base populaton, year-by-year deaths are subtracted, and brths and net mgraton are added to the populaton. Ths program allows us to project future populaton by sex and age structure, based on the current age-sex structure and addtonal assumptons about man demographc varables: mortalty, fertlty and mgraton. Projectons are made n fve-year cycles and can nvolve fve to nfnte years. Brths create a cohort of the age 0-4 at the end of the cycle. The rest of the cohorts face a probablty of mortalty and ther survvors are fve years older n the end of the cycle. Net mgraton s algebracally added to cohorts durng the cycle. The defned cohorts are presented n the followng table. Table 1 Cohorts dentfed n the FIV-FIV software Cohort Intal Fnal Cohort Intal Fnal ndex Age Age ndex Age Age 0 Born In the end of every cycle, the populaton of the cohorts 15 and 16 are added together to form the group 75+. Projectons are separately done for men and women. The frst step s to calculate the populaton alve n the begnnng of the cycle, whch comprses of cohorts one to sxteen. The second step s to calculate the populaton of cohort zero whch s the remanng and s born durng the cycle. Populaton alve s calculated by takng nto account the effect of mortalty for each cohort. That s: 8

9 (1) C F s,c = C I s,c Ss,c Where, C stands for the populaton of cohort c, sex s at a date 0 or fve ndcated by the superscrpt. S s the survval rate for each cohort. Mgraton s ncorporated the followng way. It s assumed that mgrants enter or leave the country followng a unform dstrbuton durng the fve year cycle. So, mgrants are exposed to mortalty for an average of one half of the cycle. Therefore for mgraton to be taken nto account the equaton above should be augmented the followng way. (2) C F s,c = C I S + M (1+ S )/2 for the cohorts one to sxteen. s,c s,c s,c s,c The number of brths wll depend on the number and age of women n chldbearng age, and ther age specfc fertlty rates. Women of the age group are the ones consdered of beng chldbearng age. These are seven cohorts, namely 15-19, 20-24, 25-29, 30-34, 35-39, 40-44, and The mean number of women wth a probablty of havng a chld for each of these cohorts wll be: (3) P = (P F F,C+ P I F,C)/2 where s the cohort ndex. The number of women s ether gven when we have superscrpt of zero, (snce t s gven by the prevous step) or calculated survval n the precedng when superscrpt s fve. The number of brths s calculated by usng age specfc fertlty rates, F, and for the whole cycle total number of brths wll be: 10 (4) B = 5 Σ 4 (P F ) F,C C Brths are calculated for males and females separately. Adjustng for the proporton of each gender and the correspondng survval rates males and females of the brth cohort are: (5a) C F 5 f,0 = g B S f,0 (5b) C = (1- g) B S m,0 m,0 Where g s the proporton of females born. Agan n order to augment the above equatons for mgraton we need to make assumptons about the tme and age dstrbuton of mgrants. The assumpton s that mgrants come n followng a unform dstrbuton, durng the fve year perod and that brths are equally dstrbuted between all ages 0 to 5. Ths s a suffcent approxmaton f we assume that the probablty of dyng declne n a lnear manner from brth untl the age of 5. Ths s not partcularly true especally for the newborns. In addton, very few mgrants are of the age of some 9

10 weeks or months. The software does not requre makng assumptons related to nfant mortalty of mgrants. Instead t adjusts arbtrarly the above approxmaton and assumes that the survval rate comes n wth a factor of one thrd nstead of a half. Equatons 5a and 5b take the form: (6a) C 5 f,0 = g B S s f,0 + M f,0 (2/3+1/3 S f,0 ) (6b) C 5 = (1- g) B S + M (2/3+1/3 S ) m,0 s m,0 m,0 m,0 We have now calculated the new cohort and the projecton to the next fve-year cycle s complete. In the next step each cohort ndex s ncreased by one and the fnal populaton of cohort s used as the ntal populaton of cohort +1. The data we have used for the projecton s publshed by ESYE. However there have been some gaps n data avalablty. For example, a problem we have faced s the fact that there s no data for the age and sex composton of mgrants. We have therefore assumed that mgrants have the age and sex composton of the foregn populaton of the country. The lack of data, that has been mentoned elsewhere (Konts et.al., 2006) as well as the software structure has caused small dscrepances between the ESYE Base projecton and the projecton we have generated under the same assumptons for the demographc varables. We are gong to refer to ths projecton as the Base scenaro although there are small dscrepances. The sze of these devatons s neglgble n relaton to the queston at hand. 10

11 4. G-AMOS: The Economc Model G-AMOS stands for Greek-AMOS, whch uses the AMOS modellng framework. (A Macro-Mcro Model Of Scotland) and has been developed by a team of researchers n the Unversty of Strathclyde. It s referred to as a Computable General Equlbrum modellng framework because t encompasses a range of behavoural assumptons, reflected n equatons whch can be actvated and confgured n many dfferent ways (Harrgan et al, 1991, p. 424). A wde range of model closures as well as parameter values s avalable to the user, as approprate for partcular applcatons. A good general descrpton of CGE modellng s gven n Greenaway et al (1992) and an extensve revew of regonal CGE models can be found n Partrdge and Rckman (1998). We have calbrated the G-AMOS modellng framework to data on the Greek economy gven n the form of a Socal Accountng Matrx (SAM) for the year The model has 3 transactor groups - households, frms and government - and 2 exogenous external transactors the rest of the EU (REU) and the rest of the world (ROW). The model has 25 actvtes/commodtes and these are lsted n Table A1.1 n Appendx 1. A condensed account of the model structure s gven n Table A2.1. In the verson of AMOS used here, producton takes place n perfectly compettve ndustres usng mult-level producton functons. Ths means that n every tme perod all commodty markets are n equlbrum, wth prce equal to the margnal cost of producton. Value-added s produced usng captal and labour va standard producton functon formulatons so that, n general, factor substtuton occurs n response to relatve factor-prce changes. Typcally, constant elastcty of substtuton (CES) technology s adopted, whch s the case n smulatons reported here. 4 In each ndustry ntermedate purchases are modelled as the demand for a composte commodty wth fxed (Leontef) coeffcents. These are substtutable for mported commodtes va an Armngton lnk. The composte nput then combnes wth valueadded (captal and labour) n the producton of each sector s gross output. Cost mnmsaton drves the ndustry cost functons (equaton 1 n Table A2.1) and the factor demand functons (equatons 7 and 8 n Table A2.1). 4 Leontef and Cobb-Douglas optons are avalable as specal cases. 11

12 Whlst the AMOS framework offers a wde choce of labour market closures, n the smulatons reported here the labour market s charactersed by a regonal barganng functon (also expressed as a wage curve, represented as equaton 5 n Table A2.1). Ths establshes a negatve relatonshp between the real wage and the unemployment rate (Mnford et al, 1994). Emprcal support for ths wage curve specfcaton s now wdespread (Blanchflower and Oswald 1994). The barganng functon s parametersed usng the regonal econometrc work reported n Layard et al. (1991): ln 0.113ln 0.40ln rwnt, = a ut+ rwnt, 1 where rw s the Greek real wage, u s the Greek unemployment rate, t s the tme subscrpt and a s a calbrated parameter. 5 To transform the real wage to the nomnal wage, we multply by the consumer prce ndex (equaton 2 n Table A2.1). Nevertheless, senstvty analyss has been performed n secton 6 where we examne extreme cases for the closure of the labour market. Perfect labour moblty s assumed between sectors, generatng a unfed labour market. Therefore, although wage rates vary between sectors n the base-year data set, n the smulatons wages n all sectors change by the same proportonate amount n response to exogenous shocks. The nomnal wage n each tme perod s then derved through the nteracton of the resultng wage curve and the general equlbrum labour demand curve (equaton 9 n Table A2.1). In the dervaton of the general equlbrum labour demand curve, t s mportant to note that all prces and ncomes are taken to be endogenous. The four man components of commodty fnal demand (represented by equaton 12 n Table A2.1) are consumpton, nvestment, government expendture and exports. Household consumpton s a lnear homogenous functon of real dsposable ncome and relatve prces (equatons 2, 11 and 13 n Table A2.1). Real government expendture per head s assumed to be constant (equaton 17, Table A2.1) and n these smulatons the populaton s determned exogenously usng the demographc model. 5 The calbraton s made so that the model, together wth the set of exogenous varables, wll recreate the base year data set. Ths calbrated parameter does not nfluence smulaton outputs, but the assumpton of ntal equlbrum s, of course, mportant. 12

13 Exports are determned by exogenous external demand va an Armngton lnk, makng exports relatve prce senstve (equaton 18, Table A2.1). The modellng of nvestment demand s a lttle more complex. In the mult-perod varant of the model, captal stock adjustment at the sectoral level, whch ultmately determnes aggregate nvestment demand, s dealt wth n the followng way. Wthn each tme perod, both the total captal stock and ts sectoral composton are fxed. The nteracton between ths fxed captal supply and captal demand at the sectoral level determnes each sector s captal rental rate (equaton 10, Table A2.1). The captal stock n each sector s then updated between perods va a smple captal stock adjustment procedure, accordng to whch nvestment equals deprecaton plus some fracton of the gap between the desred and actual level of the captal stock (equatons 6, 14 and 15 n Table A2.1). 6 Desred captal stocks are determned on costmnmsaton crtera, usng the user cost of captal as the relevant prce of captal (equatons 3 and 4 n Table A2.1). In the base perod the economy s assumed to be n long-run equlbrum, where desred and actual captal stocks are equal, wth nvestment smply equal to deprecaton. Investment as a source of product demand s then determned by runnng the demand for ncreased captal stock by sector through the captal matrx (equaton 16, Table A2.1). We nterpret the conceptual tme perods of the model as years: annual data are used for the calbraton and, where applcable, the estmaton of parameter values. As stated earler n ths secton, the structural characterstcs of the AMOS model are parametersed on a Socal Accountng Matrx (SAM) for Greece for In all sectors, the elastcty of substtuton between captal and labour n the producton of value added s 0.3 although senstvty analyss s conducted. Intermedate composte goods are assumed to be produced by a Leontef type producton procedure wth fxed coeffcents. Ths s requred because of the large number of zero entres. The default Armngton trade elastctes for mports and exports are 8.0. Ths s to reflect the fact that the country s consdered to be a prce taker n nternatonal markets. Sarrs and Zografaks (1999) also appled a hgh value of Armngton elastcty whle Ioakmoglou (1999) econometrcally estmated that 80% of varaton of prces n Greece s attrbutable to nternatonal prce changes. Nevertheless senstvty analyss 6 Ths process of captal accumulaton s compatble wth a smple theory of optmal frm behavour gven the assumpton of quadratc adjustment costs. The whole process s analogous to Tobn s q. 13

14 usng the values of 2, 5 and 8 s performed later on. The speed of adjustment parameter for the adjustment of actual to desred captal stock s 0.5. Before dscussng the smulaton results t s mportant to clarfy a key characterstc of the G-AMOS model. G-AMOS s not a forecastng model. When t s parametersed on the base year data set, t s assumed that the economy s n long-run equlbrum. If there are no changes to the exogenous varables and the model s run n perod-by-perod mode, then the model wll smply report an unchangng economy. Results are presented as percentage changes from the base year value and thus demonstrate devatons from what would occur f nothng happened n the economy. 5. The Consequences for the Greek Economy of the Base scenaro for the projected populaton ESYE has examned three scenaros of future demographc trends. These are labelled low, medum, hgh and the related assumptons on the values of demographc parameters are shown n fgures 3,4 and 5. Fertlty s recorded to be 1.29 n the year Under the low scenaro t assumed to fall to 1.19 chldren per woman n chldbearng age by For the medum and hgh scenaros the correspondng fgures are 1.53 and 1.87 respectvely. Fgure 3 below shows the assumed trends. Fgure 3- Assumpton for the fertlty rate Low Medum Hgh

15 Lfe expectancy s assumed to rse n a lnear manner and to reach for men years, years, years n 2050 for the low medum and hgh scenaros respectvely. The correspondng numbers for women are 83.19, 85.32, Below we see the graphcal representaton of the assumptons beng made. Fgure 4- Lfe expectancy assumptons Men Low Men Medum Men Hgh Women Low Women Medum Women Hgh Fnally, the medum scenaro assumes that net-mgraton wll reman constant at the current level of 40,000 per annum. In the hgh scenaro the assumpton s Fgure 5- Assumptons for net mgraton Low Medum Hgh Usng the medum values for fertlty, lfe expectancy and net mgraton we have generated a populaton projecton for three decades after our base year whch suffcently approxmates the projected populaton by ESYE under the medum 15

16 assumptons. The table below shows changes for total and workng age populaton for fve year perods and the fgure shows the correspondng trajectores for our base scenaro. Table 2- Total and workng age populaton under the base scenaro % change Total % 1.4% 1.1% 0.5% -0.8% -2.8% -6.1% % change 0.0% 1.3% 2.0% 2.2% 1.9% 1.3% 0.5% Fgure 6- Percentage changes n total and workng age populaton under the base scenaro 3.00% 2.00% 1.00% 0.00% -1.00% -2.00% -3.00% -4.00% -5.00% -6.00% -7.00% Workng populaton changes Total populaton By 2034 the total populaton s expected to be 11.1 mllon people, whch s an ncrease of 0.54%. However, the group falls to mllon, a fall of 6.1% snce the year These exogenously generated populaton changes are fed nto the economc model as explaned above. The table below presents changes n the man macroeconomc varables. Table 3 - The mpact of the BASE demographc projectons on Greek aggregate economc ndcators GDP Real wage Unemployment Employment Export Prce Index CPI As we expect from the theoretcal dscusson, employment falls, n ths case by 3.23% n 2034, wth a correspondng declne n GDP of a lttle less at 2.82%. Note that because of the lnearsaton of the change n the workng age populaton n the 16

17 model, the results shown n Table 1 wll overestmate the reducton n the ntal years, where an ncrease n output and employment s expected, but underestmate the rate (but not the level) of declne n the latter perod of the smulaton where workng age populaton s fallng. One thng to notce s that the fall n employment s lower than the fall n workng age populaton. Workng age populaton falls by 6.13% whlst employment only declnes by 3.23%, mplyng an ncrease n the partcpaton rate and a fall n the unemployment rate to partally offset the negatve supply sde mpacts. Ths tghtenng of the Greek labour market s apparent as we see ncreases n nomnal and real Greek take home wage of 2.40% and 2.48% respectvely by the year GDP and employment follow a very smlar pattern as shown n Fgure 7. GDP falls both because of supply and demand effects. On the demand sde although we have an ncrease n government expendture, whch s appled n some sectors, both export demand and household demand fall because of the compettveness effect and reduced labour ncome respectvely. Remember we have assumed a hgh value of Armngton elastcty n order to reflect the fact that the country s largely a prcetaker. Changes n real and nomnal wage are shown n Fgure 8. Fgure 9 shows the CPI and export prce ndces. Although wth small devatons consumer prces ntally ncrease and then fall below equlbrum. By 2034 the prce of the composte exported good that has been calculated as the weghted average of the prce of exported commodtes rses by 0.35%. As a consequence, the demand for exported goods falls by 2.85%. From the supply sde, n each sector the captal stock wll adjust to changes n output demand but more slowly than the changes n employment so that the change n GDP wll slghtly lag the changes n total employment. There wll also be a tendency for producton to become more captal ntensve as the nomnal wage ncreases, so that there s some substtuton of captal for labour. 7 There s not full pass through of the ncreased wages to prces, that s prces ncrease by less than wages, prmarly because of the presence of mports from outwth Greece as elements of the consumpton basket and as ntermedate nputs n producton. 17

18 Fgure 7- Percentage Change of GDP and Employment GDP Employment Fgure 8- Percentage changes of nomnal and real after tax wage Wages - nomnal Wages - real Fgure 9 - Percentage changes for cp and the exported prce ndex Consumer Prce Index (CPI) Export Prce Index

19 The fact that prces do not move upwards mmedately s somethng to notce. In a smple back of the envelope model of a two good prce taker economy that s presented n Appendx 3, t can be shown that n the short run prces wll move up or down dependng on the relatve labour ntensty of the exportng and non exportng sectors. It should be noted that when we run the model forward for more perods than our conventonal projecton horzon, and the negatve supply shock becomes bgger, prces do move upwards. Fgure 10 shows the changes by 2034 n sectoral output and employment generated by the base demographc projecton for Greece. At ths stage we must underlne once agan, that we do not take nto account any compostonal changes n government and household demand that may sprng from the changes n age composton of the populaton. For example, there s much dscusson of the mplcatons for health care and educaton provson resultng from longer lfe expectancy (Economc Polcy Commttee and European Commsson, 2006). Such compostonal demand changes are lkely to be mportant but they are not the subject of ths paper. These dsaggregated results therefore prmarly reflect more general demand-sde factors. These are the extent to whch the sector supples export, nvestment, and household consumpton or government demand. Fgure 10- Impact on sectoral output and employment. (Percentage changes by the year 2034) Food Products Beverages Tobacco Agrculture Huntng Forestry Fshng Mnng and Quarrng Textles Leather Footwear Wood Paper Prntng Publshng Machnery Equpment Coke Refned Petroleum Product Other Manufacturng Transport Equpment etc Electrcty Gas and Water Supply Constructon Wholesale Retal Transport and Storage Hotels and Restaurants Employment Post Telecommuncatons Fnance Insurance Real Estate Actvtes Rentng Machnery Equpment Output Research and Development Computer and Related Actvtes Other Busness Actvtes Publc Admn Educaton Health and Socal Work Other Servces By 2034 most sectors wll exhbt lower employment and output than equlbrum. There s a wde varaton n the mpacts, rangng from an output ncrease of 0.3% for Publc Admnstraton output to a decrease of 5.9% for Transport and 19

20 Storage. In general those sectors sellng most of ther output to government demand (Publc Admnstraton, Health and Socal Work, Educaton) are affected least because government expendture, whch n these smulatons s lnked to total populaton, remans relatvely constant over the whole smulaton perod, and t actually ncreases by only 0.54% by In addton these sectors are sheltered n the sense that they are not subject to nternatonal competton. The extent of the negatve effect upon other sectors (whch are mostly ht much harder) s determned by two factors. Frst, labour ntensve sectors are worst affected because of the ncreased cost of labour. Second, the sectors that are more exposed to nternatonal trade feel the negatve compettveness effect more strongly. For example, sectors such as Agrculture Huntng Forestry Fshng clearly suffer these negatve compettveness effects. Transport and Storage s the sector that realses the bggest decrease. Ths s somethng that should be attrbuted to the hgh labour ntensty and export propensty of the sector. However, Coke and refned Petroleum Products s labour exportng a lot (31% of output) but s not so labour ntensve (7% of nput goes to wages) and thus the negatve mpact appears to be only -2.6%. As argued above, n all sectors employment falls by more than output because as the prce of labour rses frms substtute labour for captal and captal stock takes tme to adjust to optmal levels. 6. Senstvty analyss It s well known n CGE lterature that closures of markets are partcularly mportant n the generaton of results. The AMOS modellng framework allows a range of labour market closures. In order to test how senstve the results are to varaton n labour market assumptons, we conducted smulatons usng the base populaton projecton under two lmtng cases. The frst mposes wthn-perod fxed labour supply. Ths means that labour supply s a gven proporton of the labour force, where the labour force s adjusted perod-by-perod through demographc changes. There s therefore assumed to be no unemployment or partcpaton rate adjustment as the labour market tghtens. In any ndvdual tme perod, ths stuaton s represented by a vertcal labour supply curve. The second, alternatve, assumpton s that there s so much slack n the local labour market that an ncrease n labour demand s met by a change n employment but no change n the real wage. In ths case, n each tme perod the labour supply curve would be horzontal and employment adjusts to 20

21 changes n labour demand through changes n the unemployment and partcpaton rates. In the fxed labour supply case we expect the employment reducton and the wage ncrease to be larger than under our base case. In the fxed real wage scenaro, the opposte results hold: we expect the employment to fall and the wage ncrease to be less than n the conventonal wage curve confguraton. When we attempt to run the model wth a fxed real wage, we get a relatvely stable GDP trend whle under the fxed labour supply closure GDP follows the lnear reducton that we exogenously mpose to the labour force. In a smlar smulaton by Lsenkova et. al (2008) the model fals to solve. The problem s that n ths case the model produces a negatve unemployment rate. The key practcal pont s smply ths: the populaton constrants mpled by demographc projectons, combned wth fxed government expendture per head, put upward pressure on wages and contnuously reduce unemployment. On the other hand, where labour supply s completely nelastc, the whole adjustment to the labour force contracton must come through hgher wages. As we have argued already, we expect employment and output to fall by more under ths labour market closure, as partcpaton and unemployment rates are not allowed to adjust. Ths wll be combned wth an even grater pressure upon wages. The fgure below shows GDP trends for the dfferent labour market closures. In Table 3 I present the percentage changes n the man aggregate ndcators for 2034 under the fxed labour supply, fxed real wage and wage curve labour market closures. Fgure 11- GDP trends under dfferent labour market closures %change FLS RWR Barganng year 21

22 Table 4- Man aggregate ndcators n 2034 under dfferent labour market closures Gross Domestc Product Employment Unemployment Consumer Prce Index Real after tax consumpton wage Barganng Fxed Real Fxed Labour (LNJ estmates) Wage Supply Indeed our results are consstent wth what economc theory suggests. Employment wth fxed labour supply s expected to fall by 6.13%, exactly n lne wth the reducton n the labour force (and workng age populaton), whlst under the barganng closure the employment reducton s only -3.23%. In the case of fxed real wage we have a mnor change of 0.14%, whch s generated, by the demand changes that I have exogenously mposed. The correspondng fgures for GDP are -5.41%, % and -0.01% respectvely. The wage rate shows a reducton of an ncrease of 2.48% under the Barganng closure and a 4.11% ncrease under fxed labour supply. By defnton there s no change under the fxed real wage closure. Unemployment falls to 7.34% (from 9.12%) under Barganng whle under fxed real wage t lands at 5.30% as the labour market s purely quantty clearng. By defnton, under fxed labour supply, the unemployment rate remans unaltered. We have also tested how senstve these economc results are to changes n the value of captal labour elastcty as well as to changes n the value of Armngton elastcty. The values we have mposed for the elastcty of substtuton between captal and labour are 0.8 and 1, whch corresponds to the Cobb-Douglas case. For the Armngton elastcty we have mposed 2.0, whch s the value that s most commonly used, and 5.0. Recall that the value mposed n the analyss of the mpacts of the base scenaro s 8.0. The table below presents the changes n the man aggregate ndcators under each scenaro. 22

23 Table 5- Man Aggregate Indcators by 2034 under dfferent values of elastcty of substtuton between local and mported goods and elastcty of labour and captal substtuton (BASE) (CD) 0.3 CES 1 CES 0.8 CES 0.3 CES 0.3 CES Arm.=8 Arm.=8 Arm.=8 Arm.=5 Arm.=2 GDP Real wage Unemployment Employment Export Prce Index CPI The expectaton s, that wth the mposton of a negatve labour supply shock employment wll fall more rapdly n the cases where substtutablty between captal and labour s easer as frms lay off workers n the lght of ncreased wages. Indeed employment falls by more under the Cobb-Douglas assumpton by less when we have elastcty of substtuton equallng 0.8 and even by less wth the value of 0.3. Reversely, the wage rate ncreases by more the lower the substtutablty between captal and labour. When layng off workers s not easy more pressure falls upon wages. In partcular by the year 2034 employment falls by 4.58%, 4.38% and 3.23% for the values of unty (Cobb-Douglas), 0.8 and 0.3 respectvely. The correspondng fgures for GDP are 3.93%, 3.79%, 2.82% respectvely. When lookng at real wage, not surprsngly the rankng s reversed and the correspondng mpacts are 1.26%, 1.43%, 2.48%. The fgure below, shows GDP trajectores for dfferent degrees of substtutablty. Fgure 12 - GDP trends under dfferent values of elastcty of substtuton between captal and labour %change year Cobb Douglas 0.8 CES BASE

24 It should be noted that although the results are obvously dependng on the values of elastctes, the magntude of the varaton can be consdered of mnor mportance n relaton to the varaton that s caused by varyng demographc parameters as we shall see n the next secton. We have also mposed dfferent values of elastctes for the substtuton of local and foregn produced goods. In ths case, the hgher the value of elastcty, the easer t s for frms to substtute local wth foregn nputs. Thus, the more dampenng the mpact wll be for the local economy. As the labour market tghtens, producng goods locally becomes more expensve and local goods are substtuted by mports. The easer ths swtch the worse the news for the local economy s employment and GDP. Alternatvely, the wage rate ncreases by more the lower the substtutablty between local and mported goods. When mport substtuton s dffcult frms nsst on producng locally. Fnally, we expect that the hgher the value of elastcty the less the prce of exported goods (nverse of compettveness) wll devate from zero. In fact mposng elastcty to equal nfnty would be equvalent to assumng that the law of one prce holds. The values of elastctes we have mposed are 2,5 and 8. Indeed, employment falls by 2.77% when the value of elastcty s 2, by 3.11% when elastcty s 5 and 3.23% when t s 8. For GDP the correspondng fgures 2.30%, 2.69% and 2.82% respectvely. As expected the rankng of mpacts s agan reversed when one looks at the behavour of real wages and prce of exported goods. Hence the correspondng ncreases of real wage are 2.94%, 2.60% and 2.48% respectvely. The fgures for compettveness are 1.12%, 0.53% and 0.35% wth the mpact of elastcty beng even clearer. Agan, as before the magntude of devatons n the results s small and one can clam that the values of these elastctes are not the factors that generate the results. The fgure below shows the trajectores of GDP under dfferent assumptons. 24

25 Fgure 13 - GDP trends under dfferent values of elastcty of substtuton between local and mported goods %change Armngton 2 Armngton BASE (8 Armngton) year 7. Varant demographc scenaros Havng checked the senstvty of results to dfferent labour market closures and key elastcty values we turn to assessng the senstvty of results to varyng demographc parameters. These parameters are the fertlty rate, the male and female lfe expectancy, and the net-mgraton rate. The tables below summarse the combnaton of assumptons that s used n each scenaro and gve the projected changes of total and workng age populaton by 2034 n each case. Table 6 Combnaton of assumptons for each demographc scenaro Scenaro Fertlty Lfe expectancy Mgraton 1 Base projecton Base Base Base 2 Hgh fertlty Hgh Base Base 3 Low fertlty Low Base Base 4 Hgh lfe expectancy Base Hgh Base 5 Low lfe expectancy Base Low Base 6 Hgh mgraton Base Base Hgh 7 Low mgraton Base Base Low 25

26 Table 7 Percentage changes of total and workng age populaton by 2034 under dfferent scenaros % -6.13% Total % 0.5% BASE Low Hgh Low Hgh Low Hgh Scenaro Expectancy Expectancy fertlty Fertlty Mgraton Mgraton % -5.81% -7.2% -4.8% % 0.68% % 1.8% -1.7% 3.6% -5.5% 6.6% Comparng the columns of table 6 we see that the varaton n net-mgraton s the demographc varable that causes the greater varaton n the sze of total and workng age populaton and thus n sze of shocks ntroduced nto the economc model. The amount of net mgraton assumed n the low mgraton scenaro s 20,000 per year whle n the hgh mgraton scenaro t s 60,000 per year. The values of fertlty and lfe expectancy are the ones that are assumed by ESYE n the correspondng scenaros and presented n secton 5. In partcular, assumng hgh and low lfe expectancy results n GDP beng 2.57% and 2.99% below base respectvely by The correspondng fgures when we vary fertlty are -1.94% and 3.55%. However, the dscrepancy of results between the scenaros of hgh and low netmgraton s much bgger. Assumng low (20,000) net-mgraton results n GDP beng 6.86% below base, but assumng hgh mgraton results n GDP beng above base by 0.88%. The paramount mportance of net-mgraton has led us to generate further scenaros by varyng even more the amount of net-mgraton whle holdng the other two varables constant. We generate a scenaro wth zero net mgraton as well as scenaros that nvolve 80,000 and 100,000 annual net-mgraton. The results summarsed n the table and the fgures below are llumnatng and the postve mpact of n-mgraton s very apparent. Note that n the scenaro where we assume zero net mgraton (natural change only) GDP falls by 11.33% n

27 Table 8 Man aggregate economc ndcators for dfferent net-mgraton scenaros BASE GDP Real wage Unemployment Employment Export Prce Index CPI Zero Gross Domestc Product Real wage Unemployment Employment Export Prce Index CPI Low Mgraton (20,000) GDP Real wage Unemployment Employment Export Prce Index CPI Hgh Mgraton (60,000) GDP Real wage Unemployment Employment Export Prce Index CPI Eghty GDP Real wage Unemployment Employment Export Prce Index CPI Hundred GDP Real wage Unemployment Employment Export Prce Index CPI

28 Fgure 14 - GDP % change under dfferent scenaros for the amount of annual mgraton %change Hgh Mgraton BASE Low year Mgraton eghty hundred Fgure 15 Real wage % change under dfferent scenaros for the amount of annual mgraton 6.00 %change Hgh Mgraton BASE Low Mgraton eghty hundred year The mportance of net-mgraton stems from the fact that mgrants due to ther age structure have an mmedate mpact upon the sze of the workng age populaton as a great deal of them s wthn the age-group. The thng to notce s that the hgher the level of mgraton the hgher the GDP and lower the real wage. The result to hghlght s that only wth scenaros that nvolve 60,000 postve net-mgraton per annum or more wll the GDP reman above equlbrum levels. In ths case the real wage remans practcally stable as the level of net-mgraton s completely off settng the tghtenng of the labour market. The results of the economc model are consstent wth the theoretcal framework, whch we put n place n Secton 2 and they are manly drven by the labour market and the subsequent mpact on compettveness. Fgure 15 shows the real wage change 28

29 assocated wth the dfferent mgraton scenaros. In ths case, the 60,000 n-mgraton case produces a neglgble change n the real wage. We know that n ths smulaton, the workng age populaton s rsng, but more slowly than the total populaton. In terms of the analyss n Secton 2, the outward shft of the labour demand curve, as government expendture rses n lne wth total populaton, s greater than the outward shft n the labour supply curve as the workng age populaton ncreases. There s therefore an ncrease n employment but also a small ncrease n the wage. However, for the examples where n-mgraton s less than 60,000, the wage ncrease s much greater and s not accompaned by any expanson n government expendture. 8. Conclusons Ths paper attempts to explore the consequences of projected demographc developments n Greece. We have nteracted a demographc model (FIV-FIV) and an economc model G-AMOS whch s an economc model for the Greek economy. We conclude that f current demographc trends contnue, that s under what we call the base scenaro there wll be a negatve mpact upon economc actvty, whch wll stem from a tghtenng of the labour market. GDP falls by 2.82% whle the real wage ncreases by 2.48% n the next three decades. Employment also falls by 3.23%. The demographc varable wth the greatest mportance appears to be the level of annual net-mgraton. Mgrants are manly of younger ages and are often accompaned wth chldren, and thus ther presence tends to mtgate the tendency for populaton ageng. The mportance of net-mgraton s apparent n the results of the zero net-mgraton scenaro. Ths scenaro answers the followng queston. What would happen f Greek borders close completely and the Greek populaton changes depend only on the underlyng brth rate and lfe expectancy? Ths s the worst scenaro of all the ones we have examned and t reveals to a great extend the postve mpact that mgrants have had and wll contnue to have n the years to come. Accordng to ths scenaro, GDP and employment fall by 11.33% and 12.99% respectvely whle the real wage ncreases by 8.33%. Remember that n the course of ths paper we have not examned, the mpacts of a change n the composton of publc and prvate demand that may occur due to a change n the demographcs, nether we have explored the mpact of the lkely 29