The Effects of Fiscal Consolidation on Macroeconomic Indicators in Cyprus

Transcription

1 Cyprus Economic Policy Review, Vol. 8, No. 1, pp (2014) The Effecs of Fiscal Consolidaion on Macroeconomic Indicaors in Cyprus Nicolea Pashouridou,a, Chrisos S. Savva b and Nicolas Syrichas c a Economics Research Cenre - Universiy of Cyprus, b Deparmen of Commerce, Finance and Shipping - Cyprus Universiy of Technology and c European Cenral Bank Absrac This paper implemens a recenly developed economeric model, he Facor- Augmened Vecor Auoregression (FAVAR), o esimae he dynamic effecs of fiscal consolidaion policies on key macroeconomic variables, such as aggregae oupu, price level, employmen, privae consumpion, invesmen and ineres rae. The resuls of our analysis show ha fiscal rerenchmen effors in he form of eiher a governmen expendiure reducion or a governmen revenue increase, lead o a fall in GDP driven by he negaive responses of invesmen, privae consumpion and employmen. As a resul of he conracionary effecs of consolidaion on economic aciviy, inflaion deceleraes. Fiscal ighening based on expendiure reducion resuls in a larger conracion in oupu han consolidaion hrough an equivalen revenue rise, especially in he medium erm. Thus, he policy mix beween expendiure cubacks and ax increases maers as i influences he cos of fiscal consolidaion in erms of los oupu. Wihin wo years, GDP falls cumulaively by 1.0 and 0.6 as a response o fiscal consolidaion amouning o 1.0 of GDP, associaed wih spending cus and equivalen revenue increases respecively. Keywords: Fiscal consolidaion, fiscal shocks, facor-augmened VAR, impulse response funcion. 1. Inroducion The global financial crisis and subsequen economic downurn has led o he accumulaion of imbalances and lef many European counries wih large public debs ha form a significan proporion of heir GDP. Greece, Ireland, Porugal, Spain and more recenly Cyprus had o resor o financial aid from European mechanisms and he Inernaional Moneary Fund condiional on he implemenaion of major fiscal consolidaion sraegies. These sraegies were based on he belief ha fiscal consolidaion would ulimaely se he scene for economic expansion and deb susainabiliy. However, recen experience has shown ha auseriy measures have resuled in severe conracion in economic aciviy and rise Corresponding auhor. Address: Economics Research Cenre, Universiy of Cyprus, P. O. Box 20537, 1678 Nicosia, Cyprus. n.pashouridou@ucy.ac.cy.

2 94 in unemploymen o hisorically record levels, wihou significan improvemens in naional deb o-gdp raio. In counries implemening fiscal consolidaion policies oupu growh underperformed he projecions prepared by inernaional organisaions. Blanchard and Leigh (2013) show ha fiscal mulipliers used in he compuaion of growh projecions have been oo low since he onse of he recen financial crisis. The value of fiscal mulipliers depends, among oher hings, on he sae of he economy (crisis vs. normal imes), he degree of openness of he economy, moneary policy and exchange rae regime, he naure of fiscal consolidaion (emporary vs. permanen measures), he composiion of fiscal consolidaion (revenue vs. expendiure measures) and he degree of fiscal consolidaion of rading parners (e.g. Boussard e al. 2012). Therefore, he availabiliy of esimaes of he effecs of fiscal policies on oupu is of paramoun imporance for forecasing and policymaking. This sudy aims o provide formal esimaes of he impac of fiscal policy shocks on various macroeconomic variables in Cyprus. The need for such esimaes for Cyprus has been pressing as he implemenaion of he Economic Adjusmen Programme and he inroducion of he Fiscal Compac necessiae he rigorous evaluaion of he effecs of fiscal measures on he economy. The impac of fiscal consolidaion is no resriced o programme counries bu has a wider applicabiliy as many counries around he globe have accumulaed large public debs. Wheher and under wha condiions reducions in governmen budge deficis are accompanied by long periods of recessions, lie in he hear of policy debae and reignie empirical and heoreical conroversies. On he one hand, sandard Keynesian models sugges ha fiscal adjusmen policies cause recessions. On he oher hand, neoclassical models emphasize he role of individuals expecaions and supply-side effecs, implying ha under cerain circumsances fiscal consolidaion plans can be expansionary or a leas non-recessionary Giavazzi and Pagano (1990) find ha fiscal adjusmens have very lile or no negaive effec on oupu. For example in Ireland and Denmark fiscal consolidaion plans were followed by high economic growh. They sress, however, ha hese plans can have non-keynesian impac only if hey are 112Fiscal adjusmen plans can have simulaing impac on privae consumpion if agens believe ha such adjusmens will increase heir lifeime income. If consolidaion is delayed, consumers will expec higher ax increases in he fuure, which, given he nonlinear increase in he disorionary effecs of axaion, will have a more disrupive effec on fuure oupu and income. By conras, imely consolidaion can improve expecaions of fuure income. Furhermore, in counries wih high public deb raios, resricive fiscal policies can reduce he risk of defaul and, in urn, boos he confidence in fuure policies; he subsequen ineres rae decline can spur economic aciviy.

3 95 considerably large and persisen. Alessina e al. (2012) use informaion on fiscal plans adoped by 16 OECD counries over a period of 30 years o analyse he impac of fiscal adjusmen on macroeconomic variables. They find ha he effecs of fiscal consolidaion depend on heir composiion (ax increases vs. expendiure reducions) and on wheher spending cus are permanen or ransiory. The cos in erms of los oupu due o consolidaion is much higher in he case of ax hikes compared o spending cus; adjusmens based on permanen measures are less cosly han sop and go ones. Erceg and Linde (2013) demonsrae ha in a currency union fiscal consolidaion is conracionary and more cosly compared o he case where moneary policy is independenly se. Oher sudies, using mainly Srucural Vecor Auoregression (SVAR) models, find ha posiive shocks on governmen spending simulae oupu a leas in he shor run (e.g. Blanchard and Peroi 2002; Cimadomo and Benassy-Quere 2012; Giordano e al. 2007). Moreover, here is evidence ha posiive shocks on axes suppress oupu (e.g. Blanchard and Peroi 2002; Cimadomo and Benassy-Quere 2012). Posiive effecs of governmen spending on consumpion is found, conrary o he predicions of neoclassical models where consumpion is expeced o fall due o negaive wealh effecs (e.g. Blanchard and Peroi 2002; Faas and Mihov 2001; Giordano e al. 2007). The resuls regarding he response of invesmen o posiive fiscal shocks vary beween negaive (e.g. Blanchard and Peroi 2002), insignifican (e.g. Faas and Mihov 2001) and posiive (e.g. Giordano e al. 2007). Neverheless, he majoriy of sudies on he effecs of fiscal policy shocks employ SVAR models where fiscal shocks are idenified as unexpeced exogenous changes in eiher governmen spending or axes. While his approach is compuaionally simple and can provide plausible resuls (under idenificaion assumpions), i is severely criicized. According o criics, VAR models, in order o conserve degrees of freedom ypically conain only a handful of variables ha canno convey all he relevan informaion. The fac ha he researcher observes less informaion han economic agens could lead o biased resuls and some conenious findings. A second problem arising from he use of SVAR analysis for fiscal policy is ha changes in fiscal policy ake ime o be implemened and herefore are anicipaed by economic agens. Since agens expec changes in fiscal policy in he very near fuure hey adjus heir decisionmaking in advance. Thus, wha he economerician idenifies as exogenous fiscal shock is in fac he resul of previous policy changes. One way o ackle he anicipaion problem is he so called narraive approach as in, for example, Romer and Romer (2010), Devries e al. (2011) and Alesina e al. (2012). These sudies idenify muli-period

4 96 consolidaion changes in fiscal policy (ax increases or governmen spending cus) by collecing records of official documens. Furhermore, as Alesina e al. (2012) saed shocks idenified hrough narraive mehod are modelled independenly and herefore are no affeced by he possibiliy ha some variables are omied in he esimaion. Neverheless, narraive approach sudies provide impulse response funcions for a small subse of variables ha scholars and policy makers are ineresed in; his is he hird limiaion of VAR models Besides he fiscal effecs on oupu, researchers and policy makers are also ineresed in he effecs on oher variables ha are ypically ignored in such small scale models. Since he undersanding of he ransmission mechanism of fiscal shocks is crucial for conducing fiscal policy, some oher echniques should be employed o overcome he aforemenioned shorcomings. Advances in economerics and saisics provide a naural soluion o he dimensionaliy problem of VAR models. In paricular, uilising facor analysis in a VAR framework overcomes he deficiency of informaion in sandard VAR models. Facors are a saisical insrumen o shrink he dimensionaliy of a large daase and a he same ime exploi all he available informaion abou is co-variaion. Thus augmening he VAR framework wih a small number of facors (Facor-Augmened VAR - FAVAR) he rich inerrelaion beween fiscal policy shocks and he real economy is adequaely capured. Furhermore, he inclusion of many business and consumer confidence indicaors in he daase addresses he anicipaion problem commonly found in fiscal policy lieraure. This sudy employs a FAVAR model o esimae he effecs of fiscal policy shocks on economic aciviy in Cyprus. More precisely, we invesigae he implicaions of fiscal consolidaion (expendiure cu, ax rise, defici reducion) on various macroeconomic aggregaes of he Cyprio economy, such as oupu, price level, employmen, privae consumpion and invesmen. Fiscal rerenchmen effors being eiher a governmen expendiure reducion or a governmen revenue increase, lead o a fall in GDP driven by he negaive responses of invesmen, privae consumpion and employmen; as a resul of he conracion in economic aciviy inflaion deceleraes. The res of he paper is organised as follows. Secion 2 oulines he empirical specificaions and describes he daa. Secion 3 presens and discusses he resuls of he empirical analysis. Secion 4 concludes. 213For insance, he effecs on labour, oupu componens, privae consumpion and invesmen.

5 97 2. Empirical specificaion and daa 2.1. Empirical specificaion The FAVAR can be viewed as a VAR model which includes a vecor of observed variables and a vecor of unobserved variables, i.e. he facors. The facors summarise he informaion in a large se of economic series wih a smaller number of variables. By augmening he VAR wih a small number of facors he informaion se is enhanced considerably wihou grealy increasing he dimensionaliy of he model. Deails on he FAVAR model are given in he Appendix. To invesigae he effecs of fiscal policy on macroeconomic variables we use wo differen specificaions of he FAVAR model in erms of he fiscal variables included. The firs specificaion uses oal general governmen expendiure and revenue and he second employs governmen surplus/defici as a percenage of GDP. In paricular he vecor of observable economic variables is specified as follows: Y GGE, GGR, GDP, CPI ) (Specificaion A) ( Y DEF, GDP, CPI ) (Specificaion B) where ( GGE and GGR represen oal general governmen expendiure and revenue respecively, GDP denoes oal oupu, measured by real GDP and CPI denoes he overall price level, measured by he Consumer Price Index; DEF is general governmen defici/surplus as a percenage of GDP. All variables excep DEF are expressed in firs differences of logarihms and all series are sandardised prior o esimaion of he model. In order o obain disinc represenaions of he various aspecs of he economy, he facors are esimaed from wo differen blocks of daa. One se of facors is exraced from a daase of domesic series (mosly leading indicaors) and anoher from a group of foreign and inernaional variables. The number of domesic and foreign facors included in he FAVAR is chosen using informaion crieria (Bai and Ng 2002) vis-à-vis he availabiliy of degrees of freedom in esimaion. Table A1 (Appendix) shows he values of differen informaion crieria for he choice of he number of facors ogeher wih he percenage of variaion explained by each facor. Domesic economy is summarised in he FAVAR model by one facor as suggesed by he second informaion crierion; he facor explains abou 16 of he variaion in he daase. The daase of foreign and

6 98 inernaional series is represened by wo facors accouning for abou 45 of he variaion in he daa. 314 In he FAVAR model domesic facors are ordered afer he observable variables in Y followed by he foreign/inernaional facors. 415 Thus, fiscal variables are ordered firs followed by GDP, CPI, domesic and foreign facors. This srucure assumes ha oupu, prices and oher aspecs of domesic (e.g. ineres raes, consumpion, invesmen, employmen) and foreign economy respond conemporaneously o fiscal shocks, bu fiscal policy reacs wih a lag o oher domesic or foreign macroeconomic shocks. Prior o sudying he effecs of fiscal shocks, he lag lengh of he FAVAR was seleced and he sabiliy of he models was checked. For boh specificaions a lag order of wo was chosen based on informaion crieria. Furhermore, boh sysems saisfy he sabiliy condiion (see Table A2- A3) Daa The fiscal variables used in he analysis namely (i) oal governmen expendiure and (ii) oal governmen revenue were obained from he Quarerly General Governmen Accouns (Saisical Service) and were expressed in consan (2005) prices using GDP deflaor. Toal governmen revenue and expendiure were used o consruc a defici/surplus series as a percenage of GDP. The firs graph in Figure 1 shows he evoluion of real governmen expendiure and revenue (seasonally adjused) over he period 1995Q1-2013Q2. Boh series follow an upward rend up unil 2008 wih expendiure hisorically exceeding revenue excep in 2005Q1 and during Afer 2008 a slowdown is observed in expendiure growh followed by a reducion in Governmen revenue declined afer 2008 and remained well below expendiure unil he end of The second graph in Figure 1 plos general governmen defici/surplus as a percenage of GDP (seasonally adjused). Fiscal defici widened over as a resul of he slowdown in and he subsequen decline in governmen revenue ha was no accompanied by a cuback in spending. A similar paern is regisered afer 2008; defici, however, has been narrowing since he second half of The wo informaion crieria sugges six or five facors (explaining 65 and 62 of he daa variance respecively) bu he inclusion of such a large number of facors in he FAVAR model would limi he degrees of freedom and he accuracy of he esimaes. 415Conemporaneous dependence of domesic facors on foreign facors and Y is removed.

7 99 For he remaining variables in he analysis we used seasonally adjused quarerly daa over he period 1995Q1-2013Q2 covering domesic real economy (oupu, consumpion, invesmen, impors, expors, employmen, unemploymen, indices of indusrial producion and reail rade, ouris arrivals, ec.), foreign economies (GDP and indusrial producion), domesic and foreign prices, ineres raes, sock exchange indices and economic senimen indicaors, as well as domesic deposis, loans and reserves, and exchange raes. 516 In oal we used 200 variables. All non-saionary variables were rendered saionary via an appropriae ransformaion such as firs differences of levels or logarihms Resuls The esimaed FAVAR model in Specificaion A is used o simulae he responses of macroeconomic variables o he following: i. a negaive shock o oal governmen expendiure equal o one percen of (annual) GDP; ii. a posiive shock o oal governmen revenue equal o one percen of (annual) GDP The daa were obained from he Saisical Service of Cyprus, Cenral Bank of Cyprus, Eurosa, European Commission (DG-ECFIN) and Daasream. 617The complee lis of variables in he daase along wih heir ransformaions are available upon reques. 718The expendiure shock corresponds o a reducion in public spending of abou EUR 144 million (consan 2005 prices) and he revenue shock equals an increase in axes and conribuions of abou EUR 144 million (consan 2005 prices). According o he Cyprus Economic Adjusmen Programme he agreed consolidaion measures for 2013 and 2014 amouned o a leas EUR 351 million (abou EUR 298 million in consan 2005 prices) and EUR 270 million (abou EUR 229 million in consan 2005 prices) respecively (European Commission 2013; IMF 2013).

8 mn 100 FIGURE 1 Fiscal daa, seasonally adjused A. Toal expendiure and revenue of general governmen (consan 2005 prices) Toal general governmen expendiure Toal general governmen revenue B. Defici/Surplus as percenage of GDP Defici/Surplus ( of GDP)

9 101 The direcion of he wo shocks can be viewed as simulaing fiscal consolidaion effors via limiing spending (e.g. public secor wage bill, pensions, social ransfers, governmen subsidies, inermediae consumpion and public invesmen) and raising revenue (e.g. direc and indirec axes, social conribuions). Figure 2 and 4 plo he impulse response funcions (IRFs) of fiscal variables, GDP and CPI for he shocks described in (i) and (ii) respecively. The 68 boosrapped confidence inervals are also shown. 819 Similarly, Figure 3 and 5 show he IRFs of variables of ineres ha are included in he facors (domesic) and are no reaed as direcly observable. The responses of employmen, ineres rae, privae consumpion and invesmen are shown. 920 The ploed IRFs shown are in erms of percenage changes of he macro variables excep in he case of ineres rae where he response is in percenage poins A negaive shock o governmen expendiure leads o a small decline in GDP on impac (-0.2). However, he negaive reacion of GDP grows larger as horizon increases. Oupu conracs cumulaively by abou 0.6 and 1 in he firs and second year afer he shock respecively. Governmen revenue drops iniially by 3 as a response o he expendiure shock bu aferwards he effecs die ou. The large decline of governmen revenue on impac is parly driven by he los ax revenue as a resul of he expendiure shock ha affecs o a grea exen he compensaions of public secor employees. The price level does no exhibi any significan reacion o he expendiure shock alhough some downward price pressures arise abou a year laer. 819A boosrap echnique was applied o compue he confidence inervals as he models included generaed regressors (see Appendix for deails). The confidence inervals were obained as he 32nd and 68h percenile of he boosrapped disribuion consruced using 3000 replicaions. The 68 confidence inerval is ypically used in his lieraure (e.g. Blanchard and Peroi 2002; Cimadomo and Benassy-Quere 2012). 920The relevan ineres rae in his conex would be a long-run (e.g. 10-year) ineres rae on governmen bonds which is no available as a coninuous ime series from 1995 onwards in he case of Cyprus. Neverheless, we presen he response of a consumer ineres rae. 1021The IRFs obained from he FAVAR esimaed in sandardised firs differences of he (logs or levels) series were ransformed back o he original unis and hen cumulaed; IRFs show he change (percenage or percenage poins) in he level of he series as a response o a shock o he fiscal variable.

10 102 FIGURE 2 Impulse response funcions of fiscal variables, GDP and CPI, negaive governmen expendiure shock Governmen expendiure Governmen revenue GDP CPI

11 103 FIGURE 3 Impulse response funcions of oher macro variables, negaive governmen expendiure shock pp Employmen Ineres rae Privae consumpion Invesmen

12 104 FIGURE 4 Impulse response funcions of fiscal variables, GDP and CPI, posiive governmen revenue shock Governmen expendiure Governmen revenue GDP CPI

13 105 The responses of oher macro variables o he expendiure shock are also negaive as shown in Figure 3. The reacion of employmen and privae consumpion is insignifican on impac bu abou a year afer he shock employmen and consumpion fall by 0.6 and 0.7 respecively. The effec of he shock on invesmen is larger han on he oher macro variables and significan even on impac. Invesmen declines by 0.5 in he quarer of he shock and by 2.4 abou a year afer ha; subsequenly i coninues o conrac albei a slower raes. The effecs of governmen expendiure reducion on marke ineres rae are no found o be significan. A posiive shock on governmen revenue driven by, for example, a rise in axes and conribuions resuls in almos no change in oupu on impac; subsequenly GDP falls bu a a smaller rae compared o he oupu decline riggered by he expendiure shock. As a reacion o he revenue shock, GDP conracs by 0.4 in he firs year. The price level falls by 0.2 and 0.4 on impac and wihin he firs year respecively, as a response o he posiive revenue shock; aferwards he effecs of he shock on CPI fade away. The aggregae price level reacs more srongly o he revenue han o he expendiure shock. The sronges response o he revenue shock is esimaed for invesmen and employmen. Boh series reac insanly wih a decline; in he firs year employmen and invesmen are lower by 0.5 and 1.5 respecively. Afer he firs year he effec of he shock on employmen becomes smaller and evenually vanishes. The response of invesmen o he revenue shock appears more persisen. Privae consumpion also reacs negaively o he revenue shock; he shock enails lower disposable income, hrough higher direc axes and/or higher consumer prices via higher indirec ax raes, and consequenly lower real income. The consumer ineres rae increases on impac; aferwards he effec of he shock dies ou and becomes saisically insignifican. The FAVAR model in specificaion A does no allow us o examine he impac of simulaneous shocks o boh governmen expendiure and revenue. In specificaion B he fiscal variable of ineres is represened by governmen defici/surplus as a percenage of GDP. Thus, specificaion B allows us o ranslae he wo independen shocks o expendiure and revenue ino a single shock o defici. The esimaed FAVAR model in specificaion B is used o simulae he responses of macro variables o a shock o governmen defici generaed by a negaive shock o governmen

14 106 expendiure and a posiive shock o revenue equal o one percen of annual GDP each Figure 6 and 7 show he IRFs of various macro variables o he posiive shock on defici, along wih he 68 boosrapped confidence inervals. Fiscal consolidaion measures ha curb governmen spending and raise governmen revenue via increases in axes and conribuions, lower GDP on impac by as lile as 0.1. The response of GDP o such a shock is quie persisen. Oupu conracs hroughou he horizon bu he effecs of he shock become smaller afer he second year. Cumulaively GDP falls by 1.2 and 3.0 in he firs and second year respecively as a response o he defici shock. As a resul of he fiscal consolidaion shock which riggers adverse effecs on oupu, he price level reacs negaively on impac; i declines by 0.2. Aferwards, CPI exhibis a hump-shaped response which is no found o be saisically significan. The reacion of he price level fades away faser han ha of oupu. The cumulaive drop in CPI is esimaed a abou 0.4 during he firs wo years afer he shock. Employmen reacs negaively o he defici shock as i declines by 0.1 on impac and by 0.8 (cumulaively) in he firs year. The iniial response of privae consumpion o he shock is raher couner-inuiive. I increases on impac and declines quie fas afer he second quarer. However, he response of privae consumpion is consisen wih he dynamic reacion of he consumer ineres rae o he shock. Ineres rae drops on impac by abou 2 percenage poins, possibly as a resul of low demand for borrowing iniiaed by governmen hrough is expendiure cuback. Subsequenly ineres rae rises slowly before he effecs of he shock die ou. The effec of he defici shock on invesmen is insignifican boh on impac and during he firs year. Aferwards invesmen falls rapidly unil he end of he horizon when i sabilises a abou 18 below is pre-shock level. The effecs of he defici shock on invesmen are much more persisen han hose on employmen or consumpion. 1122The shock o defici as a percenage of annual GDP equals wo percenage poins. The shock simulaed amouns o fiscal consolidaion measures of abou EUR 288 million (in consan 2005 prices).

15 FIGURE 5 Impulse response funcions of oher macro variables, posiive governmen revenue shock Employmen pp Ineres rae Privae consumpion Invesmen

16 of GDP FIGURE 6 Impulse response funcions of defici, GDP and CPI, posiive defici shock Defici/Surplus GDP CPI

17 FIGURE 7 Impulse response funcions of macro variables, posiive defici shock pp Employmen Ineres rae Privae consumpion Invesmen Table 1 summarises he effecs of all fiscal consolidaion shocks considered on he differen macro series a various poins of he horizon. All ypes of shocks resul, a leas a year afer he shock, in a decline in GDP. The shock o defici can be viewed as simulaing simulaneously a reducion in expendiure and an increase in revenue, and herefore is impac on GDP is sronger han he effecs of he wo disinc shocks. Moreover, he reacion of GDP o he defici

18 110 shock is more negaive han he sum of he effecs of he wo separae shocks in he firs and second year. Employmen and invesmen, which are he main drivers of aciviy decline in he case of he defici shock, exhibi a more negaive response o he combined shock han he sum of he responses o he wo disinc shocks only in he medium erm. Looking separaely a he expendiure and revenue shocks we find ha he former leads o a larger oupu conracion driven on impac by he more negaive reacion of invesmen and subsequenly inensified by he more adverse reacion of employmen and consumpion. The faser reacion of prices o he revenue shock seems o resrain he conracion in aggregae oupu hrough a smaller drop in privae consumpion and subsequenly employmen and invesmen compared o he case of a negaive expendiure shock. The resuls of he analysis indicae ha fiscal rerenchmen based on expendiure cus affecs economic aciviy more adversely han consolidaion based on revenue increases, via higher axes and conribuions, due o he sronger negaive reacion exhibied by invesmen, privae consumpion and employmen. Apar from he faser reacion of he price level in he case of a revenue shock ha could lead o a smaller oupu conracion, we discuss wo possibiliies which could explain he larger GDP decline in he case of a negaive expendiure shock. The negaive governmen spending shock which leads o cus in public secor salaries, pensions and social ransfers is esimaed o be followed by a larger drop in privae consumpion compared o an equivalen posiive revenue shock as he spending shock migh affec a larger proporion of lower income consumers (e.g. pensioners, social assisance recipiens) wih higher marginal propensiy o consume. The revenue shock associaed wih increases in direc and indirec axes ends o influence more srongly higher as opposed o lower income households, hus consumpion is affeced o a smaller degree. I should be noed ha here we look a he reacion of aggregae invesmen. Hisorically, public invesmen consiues abou 12 of oal invesmen; moreover abou 7.5 of governmen expendiure goes o invesmen spending. A negaive shock on governmen expendiure, herefore, riggers a direc drop in invesmen via reduced public invesmen, and consequenly lower employmen. A posiive governmen revenue shock is followed by a smaller reducion in invesmen, compared o he case of a negaive expendiure shock, as he effecs of he revenue shock on invesmen are raher indirec. The reducion in aggregae invesmen is possibly driven by he decline in privae invesmen due o

19 Expendiure Revenue Defici ( of GDP) Expendiure Revenue Defici ( of GDP) Expendiure Revenue Defici ( of GDP) Expendiure Revenue Defici ( of GDP) Expendiure Revenue Defici ( of GDP) 111 TABLE 1 Effecs of fiscal consolidaion shocks on macroeconomic variables Macro variable GDP CPI Employmen Privae consumpion Invesmen Type of shock On impac Four quarers afer he shock Eigh quarers afer he shock

20 112 higher axes (e.g. profi axes) and by subdued consumer demand because of higher indirec axes (e.g. VAT) and lower disposable income. As a resul of he more negaive responses of privae consumpion and invesmen o he governmen expendiure shock, employmen, especially in he medium erm conracs by more compared o he case of a posiive revenue shock. 4. Conclusions The objecive of his sudy was he invesigaion of he dynamic effecs of fiscal rerenchmen policies on key macroeconomic variables using a Vecor Auoregression model augmened by facors ha summarise a large number of domesic and foreign series. This framework of analysis known as Facor-Augmened Vecor Auoregression (FAVAR) has been widely used in he recen lieraure o examine he effecs of fiscal policy. The FAVAR model allows us o race he effecs of fiscal shocks no only on he variables which are direcly included in he VAR model i.e. fiscal variables, GDP and prices, bu also on a number of oher macroeconomic series of ineres, such as employmen, privae consumpion, invesmen and ineres rae. The resuls of our analysis show ha fiscal consolidaion effors being eiher a governmen expendiure reducion or a governmen revenue increase, lead o a fall in GDP driven by he negaive responses of invesmen, privae consumpion and employmen. As a resul of he conracionary effecs on economic aciviy, inflaion deceleraes. Similar reacions of macroeconomic variables o fiscal policy shocks were found by oher sudies in he lieraure. Blanchard and Peroi (2002) find ha a posiive shock on governmen spending is followed by an increase in oupu and consumpion, while a posiive ax shock (ax increase) leads o a decline in oupu and privae invesmen. Faas and Mihov (2001) find ha a posiive shock o governmen spending is followed by an increase in consumpion and employmen. Oher sudies (e.g. Cimadomo and Benassy-Quere 2012; Giordano e al. 2007) also repor a posiive oupu reacion o expansionary governmen spending shocks. Furhermore, i is found ha a ax cu can simulae he economy (e.g. Cimadomo and Benassy-Quere, 2012 for Germany; Mounford and Uhlig 2009). Our analysis for Cyprus reveals ha fiscal ighening based on expendiure reducion resuls in a larger conracion in oupu han consolidaion based on revenue increase especially in he medium erm. Wihin wo years, GDP falls cumulaively by 1.0 and 0.6 as a response o fiscal consolidaion (amouning o 1.0 of GDP) associaed wih spending cus and equivalen

21 113 revenue increases respecively. Thus, he policy mix beween expendiure cubacks and ax increases maers as i influences he cos of fiscal consolidaion in erms of los oupu. This finding is he opposie of ha repored in Alessina e al. (2012) where hey find ha ax hikes are much more cosly in erms of oupu loss han spending cus; heir esimaed effecs however are he average across a number of OECD counries and fiscal consolidaion plans. Erceg and Linde (2013) demonsrae ha in a currency union an expendiure-based consolidaion depresses oupu by more han a ax-based consolidaion during he firs hree years; hey conclude ha he composiion of fiscal consolidaion should ake ino accoun he moneary policy regime of he counry as he laer affecs he cos of he consolidaion. Regarding he size of he esimaed effecs compared o sudies for oher counries we can loosely infer he following. The shor-erm response of oupu (four quarers) o he governmen expendiure shock is close o ha found in oher sudies for he Unied Kingdom and Germany bu lower han he responses esimaed for France, Ialy, Spain, Porugal and he Euro Area (see Boussard e al. 2012, Table 1). However, he medium-erm reacion of oupu o he expendiure shock esimaed for Cyprus is closer o ha repored for he Euro Area (see Boussard e al. 2012, Table 1). The shor- and medium-erm reacion of oupu o a governmen revenue shock (ax increase) for Cyprus found here, resembles ha esimaed for Germany, France and he Euro Area (see Boussard e al. 2012, Table 2). Overall, our resuls show ha fiscal ighening policies of magniude similar o ha in he Cyprus Economic Adjusmen Programme, ha aim o reduce governmen defici can lead, ceeris paribus, o a decline in GDP in he firs and second year afer he enforcemen of he measures by 1.0 and 3.0 respecively. The framework of analysis developed here, however, is more general and can be used o analyse he effecs of oher fiscal policies on macroeconomic indicaors. The model can be exended by including more deailed sub-caegories of governmen expendiure (e.g. compensaion of employees, inermediae consumpion, public invesmen, social ransfers, ineres paymens, ec.) and revenue (e.g. income ax, indirec ax, conribuions, ec.) in order o simulae he impacs of more specific governmen policies on he macroeconomy.

22 114 Appendix A.1. Facor-Augmened Vecor Auoregression (FAVAR) Model Le X denoe a N 1 marix ha conains a large number of economic ime series; Y is a M 1vecor of endogenous variables ha are a subse of X. The usual approach is o employ a VAR or srucural VAR using daa for Y alone o esimae various macroeconomic relaionships. Neverheless, in many applicaions, addiional economic informaion (no fully capured by Y ) may be relevan o modelling he dynamics of he series in Y. Therefore, le us suppose ha F, a K 1 vecor of unobserved facors, can summarize mos of he X, i.e. K is much smaller han N. The unobserved informaion conained in facors can be viewed as reflecing conceps ha canno be easily represened by specific series bu are capured by a wide range of economic variables (see e.g. Bernanke e al. 2005; Sock and Wason 2005). The join dynamics of ( F ', Y ') and he saic represenaion of a dynamic facor model X, F, Y are given by he following equaions: F F 1 ( L) Y Y 1 (1) X f y F Y e (2) where ( L) is a ( K M) ( K M) marix lag polynomial of finie order d, whose parameers could be subjec o a priori resricions as in a srucural VAR f seup. The error erm has mean zero and variance-covariance marix Q. is y a marix of facor loadings wih dimensions N K, is N M ; e is a mean zero vecor of errors exhibiing some degree of cross-correlaion. 1 Facors F are unobserved and herefore mus be esimaed joinly wih or prior o he esimaion of equaion (1). The dynamic evoluion of each economic series in X is governed by he K facors and he M elemens of he variables of ineres Y, which are common o all elemens of X, plus an idiosyncraic componen. The saic represenaion of he dynamic facor model described by equaion (1) and (2) where F can also include lags of he facors allows he esimaion of he space spanned by he facors by applicaion of he principal componens mehod 1 In he principal componen mehod employed for he esimaion of facors he crosscorrelaion beween error erms in e ends o zero as he number of series in X (i.e. N ) becomes large (Sock and Wason 2002).

23 115 (see Sock and Wason 1998, 2002, 2005). Then he FAVAR model (1) can be K M esimaed using a smaller number of variables ( ) han he dimension of X. Moreover, he FAVAR model ness he simple VAR model. Equaions (1) and (2) can be esimaed in wo ways: (i) a wo-sep procedure based on principal componens and (ii) a single-sep Bayesian maximum likelihood (Bernanke e al. 2005). Here we use he wo-sep procedure which has he advanage of being compuaionally simple and easy o implemen. Furhermore, in conras o he Bayesian maximum likelihood mehod, i requires fewer disribuional assumpions and allows for some cross-correlaion in he idiosyncraic error erms e. In he firs sep of he wo-sep procedure he space spanned by he facors of he ~ daa marix X 0, denoed by F 0 is esimaed by he firs K principal componens of X 0, where X 0 is he par of he marix X ha does no include Y as he laer is reaed as being direcly observable. 2 Sock and Wason (2002) showed ha he principal componens mehod yields consisen esimaors when boh cross secion and ime series dimensions are sufficienly large. ~ In he second sep F 0 can replace F in he FAVAR model and subsequenly equaion (1) can be esimaed as a sandard VAR. Furhermore, like a sandard VAR, he FAVAR model requires assumpions/resricions for he idenificaion of policy shocks. Here idenificaion is achieved by employing a recursive ordering of he variables (i.e. Cholesky idenificaion scheme) where fiscal variables are ordered firs, followed by oher macroeconomic variables of ineres (oupu and prices) and facors. 3 A similar variable ordering, where macro series such as oupu and prices respond conemporaneously o a fiscal shock, is followed in oher works sudying he effecs of fiscal policy (e.g. Giordano e al. 2007; Peroi 2004). The esimaion of he FAVAR model in he second sep involves he esimaed ~ facors and herefore generaed regressors, hus confidence inervals for he F 0 impulse response funcions obained from he FAVAR are compued using a boosrap echnique (Kilian 1998). 2 In he principal componen esimaion a normalisaion is imposed on he esimaed facors and heir loadings in order o reach a unique soluion as alernaive ses of facors and loadings ha span he same space as ~ can be found. F 0 3 Oher idenificaion schemes which are available in he VAR lieraure (e.g. Leeper e al. 1996; Bernanke and Mihov 1998) can also be employed in he FAVAR seup.

24 116 A.2. Esimaion of facors, selecion of lag order and sabiliy condiions No. of facors ICP1 TABLE A1 Esimaion of he number of facors Domesic daa ICP2 of variance explained by facor Foreign / inernaional daa ICP1 ICP2 of variance explained by facor Esimaed no. of facors No. of variables (cross secion) No. of periods (ime series) No. of lags in FAVAR Final predicion error TABLE A2 Crieria for lag order selecion AIC Specificaion A BIC Final predicion error Specificaion B * * * * 8.38* * Noe: The symbol * denoes he number of lags seleced by each saisic. AIC BIC

25 117 TABLE A3 Sabiliy condiions, eigenvalue modulus Specificaion A Specificaion B Noe: The modulus of all eigenvalues is less han uniy and herefore he sysems saisfy he sabiliy condiion. References Alesina, A., Favero, C., and Giavazzi, F., (2012), The oupu effec of fiscal consolidaions, NBER Working Paper Bai, J., and Ng, S., (2002), Deermining he number of facors in approximae facor models, Economerica 70: Barro, R., (1974) Are governmen bonds ne wealh?, Journal of Poliical Economy 82: Bernanke, B., and Mihov, I., (1998), Measuring moneary policy, Quarerly Journal of Economics 113(3), Bernanke, B., Boivin, J., and Eliasz, P. S., (2005), Measuring he effecs of moneary policy: A facor-augmened vecor auoregressive (FAVAR) approach, Quarerly Journal of Economics 120: Blanchard, O., and Peroi, R., (2002), An empirical characerizaion of he dynamic effecs of changes in governmen spending and axes on oupu, Quarerly Journal of Economics 117: Blanchard, O., and Leigh, D., (2013), Growh forecas errors and fiscal mulipliers, IMF Working Paper WP/13/1. Boussard, J., de Casro, F., and Salo, M., (2012), Fiscal mulipliers and public deb dynamics in consolidaions, European Economy - Economic Papers 460, European Commission, Direcorae-General for Economic and Financial Affairs.

26 118 Cimadomo, J., and Benassy-Quere, A., (2012), Changing paerns of fiscal policy mulipliers in Germany, he UK and he US, Journal of Macroeconomics 34: Devries, P., Guajardo, J., Leigh, D., and Pescaori, A., (2011), An Acion-based analysis of fiscal consolidaion in OECD counries, IMF Working Paper No. 11/128. Erceg, C. J., and Linde, J., (2013), Fiscal consolidaion in a currency union: Spending cus vs. ax hikes, Journal of Economic Dynamics and Conrol 37: European Commission, (2013), The economic adjusmen programme for Cyprus, Second review Auumn 2013, European Economy - Occasional Papers 169, December 2013, Direcorae General Economic and Financial Affairs. Faas, A., and Mihov, I., (2001), The effecs of fiscal policy on consumpion and employmen: Theory and evidence, No Cenre of Economic and Policy Research. Galí, J., Gerler, M., and López-Salido, J. D., (2007), Mark-ups, gaps, and he welfare coss of business flucuaions, Review of Economics and Saisics 89: Giavazzi, F., and Pagano, M., (1990), Can severe fiscal conracions be expansionary? Tales of wo small European counries, NBER Macroeconomics Annual 5: Giordano, R., Momigliano, S., Neri, S., and Peroi, R., (2007), The effecs of fiscal policy in Ialy: Evidence from a VAR model, European Journal of Poliical Economy 23: Inernaional Moneary Fund, (2013), Cyprus: Second review under he exended arrangemen under he exended Fund faciliy and reques for modificaion of performance crieria, IMF Counry Repor No. 13/374, December Kilian, L., (1998), Small-sample confidence inervals for impulse response funcions, Review of Economics and Saisics 80: Leeper, E., Sims, C., and Zha, T., (1996), Wha does moneary policy do?, Brookings Papers on Economic Aciviy 2: Mounford, A., and Uhlig, H., (2009), Wha are he effecs of fiscal policy shocks?, Journal of Applied Economerics 24: Peroi, R., (2004), Esimaing he effecs of fiscal policy in OECD counries, IGIER Working Paper no Romer, C. D., and Romer, D. H., (2010), The macroeconomic effecs of ax changes: Esimaes based on a new measure of fiscal shocks, American Economic Review 100: Sock, J. H., and Wason, M. W., (1998), Diffusion indexes, NBER Working Paper No

27 119 Sock, J. H., and Wason, M. W., (2002), Macroeconomic forecasing using diffusion indexes, Journal of Business and Economic Saisics 20: Sock, J. H., and Wason, M. W., (2003), Has he business cycle changed? In: Proceedings of he Symposium on Moneary Policy and Uncerainy pp Sock, J. H., and Wason M. W., (2005), Implicaions of dynamic facor models for VAR analysis, NBER Working Papers No