The relationship between shadow economy and unemployment rate. A ARDL causality analysis for the case of Romania

Transcription

1 The relaionship beween shadow economy and unemploymen rae. A ARDL causaliy analysis for he case of Romania Adriana AnaMaria Alexandru (Davidescu) Naional Scienific Research Insiue for Labour and Social Proecion, Buchares Deparmen of Saisics and Economerics, Academy of Economic Sudies, Buchares adrianaalexandru@yahoo.com Absrac The paper aims o invesigae he naure of he relaionship beween he shadow economy (SE) and unemploymen raes (boh regisered and ILO) for he case of Romania using Pesaran e al.() bounds ess approach for coinegraion. The sudy uses quarerly daa covering he period -. The size of Romanian shadow economy is esimaed using he currency demand approach based on VECM models, saing ha is size is decreasing over he analyzed period, from 6.5% a he end of o abou.5% of real GDP a he middle of. The ARDL coinegraion approach resuls show ha in shor-run boh ILO and regisered unemploymen rae has a negaive and saisically significan effec on he size of he shadow economy, while in he long-run he unemploymen raes has a posiive effec on shadow economy. The ARDL causaliy resuls revealed he exisence of a longrun unidirecional causaliy ha runs from unemploymen raes o shadow economy. Keywords : currency demand approach, ARDL coinegraion approach, unemploymen raes. Inroducion The paper aims o esimae he size of he Romanian shadow economy (SE) using he currency demand approach for quarerly daa covering he period - and o invesigae he relaionship beween he size of he shadow economy(se) and he regisered and ILO unemploymen raes for he case of Romanian daa using ARDL causaliy analysis. This dual srucure divides he paper ino wo main secions. The firs par briefly presens he esimaion of he size of Romanian shadow economy as % of official GDP using he currency demand approach. The second par presens an overview of he empirical research analyzing he relaionship beween shadow economy and unemploymen raes and invesigaes he

2 causaliy beween hem providing he daa and mehodology and finally he main economerical oupus. The main empirical resuls regarding he Romanian shadow economy esimaes are obained by boh naional and inernaional sudies using differen esimaion mehods and are presened in able. Table. The size of Romanian shadow economy(% of official GDP) Auhors Approach Period Size of SE (min-max) Albu(, 8,, ) Discrepancy beween acual %-.% and desired income Insiuul Național de Saisică Labour inpu mehod %-.5% Johnson(997, 998) Physical inpu mehod %-8.% Lacko(999) Physical inpu mehod %-.% Schneider e al.(,, DYMIMIC model and currency %-7.4% 4, 5, 6, 7, 9) demand approach Schneider e al.() MIMIC model %-6.7% As Schneider and Ense () saed, no approach is exemp from criicism, he empirical resuls being differen. So, if according o Naional Insiue of Saisics, he informal aciviy represens beween 4.5% and.5% of official GDP, Schneider e al.() esimaes he size of shadow economy in Romania o overcome he hreshold of 5% of official GDP.. Esimaing he size of Romanian shadow economy.. Meodology and daa The size of Romanian shadow economy was esimaed using one of he mos commonly used indirec mehods proposed by Cagand and Tanzi s ha assumes ha shadow (or hidden) ransacions are underaken in he form of cash paymens, so as o leave no observable races for he auhoriies. An increase in he size of he shadow economy will herefore increase he demand for currency. The size of he Romanian shadow economy based on he currency demand approach was esimaed using mulivariae coinegraion and vecor error correcion models (VECM) in order o invesigae wheher a long-run equilibrium relaionship exiss beween currency demand and is deerminans. Several empirical models were esimaed: M: C = α + α Y + α Tax + α Rr + α WS 4 () M: C = α + α Y + α Tax + α Rr + α WS 4 () M: C = β + β Y + β Tax + β Rr + β 4 Gov () M4: C = θ + θ Y d + θ Tax + θ Rr + θ 4 E 4 (4) M5: C = ϑ + ϑ Y d + ϑ Tax + ϑ Rr 5 (5)

3 M6: ( C ) = γ Tax WS Rr Y M + γ + γ + γ + γ 4 pc 6 (6) where: C is he currency in circulaion ouside he banks (a he end of he period in millions C RON) normalized by he GDP deflaor; M is of currency ouside he banks and M (broad money);y is real gross domesic produc in millions RON, base year Ypc (=); is real GDP per capia; Y d is real disposable income (GDP a curren Y prices- axes) normalized by GDP deflaor; d is real disposable income (GDP a curren prices-axes+social benefis) normalized by GDP deflaor; Tax is oal ax revenues normalized by GDP; Rr is -monh Real Romanian iner-bank offered rae (ROBOR); Rr is -monh Real Romanian iner-bank offered rae (ROBOR);WS is he raio of gross wages and salaries in naional income; Gov is he raio of final governmen consumpion expendiure in GDP; E is he raio of privae consumpion expendiure in GDP; ε i, i =,.., 6 are he error erms. Regarding he sign of he variables in he models, we expec a posiive impac on currency demand for income, axes, wages, governmen consumpion, privae consumpion and a negaive effec from he par of ineres raes. The ime span covered by he series is from :Q o :Q, he number of observaion is 4. Apar from he real ineres raes and he real currency ouside banks, he daa were seasonally adjused by means of ramo seas mehod. All series were expressed in logarihmic form. The main sources used o collec he daa are: Eurosa, Naional Bank of Romania and Naional Insiue of Saisics. A descripion of he variables and heir sources is summarized in he able of Appendix... Empirical resuls... Esimaing he currency demand approach As a preliminary sep, we carried ou ess of non-saionariy for each series. Furhermore, Augmened Dickey-Fuller (ADF) and Phillips-Perron (PP)) ess were employed in order o idenify each variable s inegraion level. Boh ess were consisen wih he hypohesis ha he variables are level-unsaionary bu firs difference saionary. Since i has been deermined ha he variables under examinaion are inegraed of order hen he Panula principle and he Johansen and Juselius coinegraion es were be performed. The ess sugges a leas one coinegraing vecor exiss for each model considered, meaning ha here is a long-run equilibrium relaionship beween he variables of he model. The expeced posiive impac of axes on currency demand can be inerpreed, following Tanzi: if he level of axaion increase, economic agens will be encourage engaging ax-evading aciviies, using currency, due o he inracabiliy of cash, and han he currency rises (Brambilla Macias and Cazzavillan, 9). If he ineres rae increases, he economic agens ge ride o heir currency holdings. The negaive expeced coefficien on he inflaion variable reflecs he fac ha rising raes on inflaion erode he value of money and encourage smaller volumes of currency holdings.

4 Given he non-saionariy of our series, radiional esimaion mehods are ruled ou and we mus hen esimae a VAR model in which we shall include a mechanism of error correcion model (ECM). The resuls of he VECM models are presened in able. Table. Empirical resuls of VECM models Analyzing he empirical resuls for all he six models, he bes model was considered he model M in which he coefficiens for oupu, ax burden and wages have a posiive long-run effec, while ineres rae ake he pressure off on currency demand. This implies ha an increase in he ax rae is likely o moivae individuals o paricipae in he underground economy in an aemp o increase heir income. The coefficien on ineres raes is negaive as expeced reflecing he increasing opporuniy cos in holding currency balances when ineres raes are rising. All coefficiens are srongly saisically significan. The significance of he error correcion erm (ECT) shows causaliy in a leas one direcion. The lagged error erm ( EC ) in our resuls is negaive and highly significan.

5 The error correcion erm for model M - as -. (.) indicaes a high rae of convergence o equilibrium, which implies ha deviaion from he long-erm equilibrium is correced by % over each quarer. The coefficien of he error correcion erm is significan a % level and greaer han uniy implying a high speed of adjusmen owards equilibrium.... Obaining he size of Romanian shadow economy Afer esimaing he vecor error correcion model (VECM) and obaining he coefficiens for he long-run relaionship of model M,we compue C ~ using all he coefficiens in equaion (7).Then, we se he ax variable 4 equal o zero and re-esimae he equaion, keeping all he oher coefficiens unchanged o obain Ĉ : C ˆ = Y Tax.5 R WS (7) The difference beween hese wo variables- Ĉ and C ~ -gives he amoun of exra currency (EC) or illegal money in he economy. Following Tanzi (98), we assume equal velociy in boh he formal and informal and esimae i as follows: Y = v (8) M EC Equaion (8) yields he velociy of money in he Romanian economy. Y is he gross domesic produc, Mcorresponds o oal currency and deposis in circulaion and exra currency (EC) for exra currency or illegal currency. The difference beween M and EC can be inerpreed as he amoun of legal money used in economy. Once we esimae he velociy from equaion (8), he dimension of shadow economy using he currency demand approach can be obained muliplying EC by he velociy of money: EC v = Y shadow (9) The main idea behind he model is ha a rise in he underground economy will cause an increase in demand for money. Following he Ahumada e al. (7), we proceed o correc our esimaes using heir suggesed mehod 5 : Y ˆ shadow Cshadow β Yshadow β = ( ) = ( ) () Y C Yˆ official official official where: Y is he GDP, C is he currency, while β is he income elasiciy. The correcion Yˆ shadow basically deflaes he wrong raio ( ) ha we obained using inappropriaely he Yˆ official assumpion β =. Equaion () correcs he esimaion when β =. The long-run relaionship beween our variables was derived normalizing C. 4 The only reason ax raes would affec he use of cash is ha cash is used as a means of evading axes in he underground economy (Spiro, 994). 5 Ahumada e al.(7) show ha i is wrong o assume he same velociy of money when he hypohesis β = is rejeced by he economeric esimaion of he currency demand model. This is our case, since our model gives us a coefficien β =. 8.

6 The empirical resuls of currency demand approach based on VECM models emphasizes ha here is a general downward rend in he size of he shadow economy as % of official GDP for he period - wih an highligh on wo low periods, Q and 8Q4. Thus, he size of he shadow economy 6 as % of official GDP measures approximaely 6.6% in Q and follows a downward rend afer regisering he value of % by 8. For he pas few quarers, here is a slighly upward rend in he size of Romanian shadow economy. The resuls are consisen wih sudies by Schneider (7) and Albu (7,, ) which show a mainly downward rend of informal economy in Romania % of official GDP Q Q Q Q 4Q 5Q 6Q 7Q 8Q 9Q Q Fig.. The size of Romanian shadow economy as % of official GDP. The relaionship beween unemploymen rae and shadow economy in Romania. A ARDL causaliy analysis The aim of his second par of he paper is o invesigae he naure of he relaionship beween unemploymen raes 7 and he size of he Romanian shadow economy and o idenify he direcion of causaliy beween hem using ARDL coinegraion and causaliy approach. According o Giles and Tedds (), wo opposing forces deermine he relaionship beween unemploymen and he informal economy. On he one hand, an increase in he 6 I is imporan o noe ha because of is undeecable naure and characer, i is nearly impossible o measure precisely he size of economic aciviies aking place in he informal economy of any counry in he world, wheher developed or less developed. Given his, any heoreical or empirical inference derived from hese resuls should always be regarded as an approximaion. In he face of hese difficulies, he resuls drawn from hese esimaes should be inerpreed wih due reserve, given he limiaions of he mehods. 7 Regarding he Romanian unemploymen daa, here are wo measures available for unemployed persons: he firs is he regisered unemploymen rae, who is calculaed by Naional Agency for Employmen (NAE) and based on saemens of people who pass by employmen agencies and said ha hey are unemployed and he ILO unemploymen rae, who is published quarerly by he Naional Insiue of Saisics and is based on labour force survey (LFS).

7 unemploymen rae may involve a decrease in he informal economy because i is posiively relaed o he growh rae of GDP and evenually negaively correlaed wih unemploymen (Okun's law). On he oher hand, increase in unemploymen leads o an increase in people working in he informal economy because hey have more ime for such aciviies. Dell Anno and Solomon (7) saed ha here is a posiive relaionship in he shorrun beween unemploymen rae and U.S. shadow economy for he period Using SVAR analysis, hey invesigae he response of he shadow economy o an aggregae supply shock (impac of he shadow economy o a emporary shock in unemploymen). The empirical resuls show ha in he shor-run, a posiive aggregae supply shock causes he shadow economy o rise by abou 8% above he baseline... Mehodology and daa The daa used in he research covers he period :Q- Q. The variables used are as follows: he size of he Romanian shadow economy expressed as % of official GDP (SE) obained by currency demand approach; ILO unemploymen rae (ILO_UR) and regisered unemploymen rae (R_UR). The unemploymen raes were seasonally by means of ramo seas mehod. The main source of he daa for unemploymen raes is he Naional Insiue of Saisics (Tempo daabase) and he Naional Bank of Romania. In order o invesigae he relaionship beween shadow economy and unemploymen rae, we esimae he models: SE = α + β R _ UR () SE = α + β ILO _ UR () where: SE is he size of Romanian shadow economy as % of official GDP obained hrough ARDL models; R _ UR is he regisered unemploymen rae; ILO _ UR is he ILO unemploymen rae; α,α are consans; ε,ε is he disurbance erms. According o Pesaran e al.() approach, in he firs sep we have verified he coinegraion using he bounds es approach. In he second phase, once he coninegraion confirmed, we esimae he long-run coefficiens of he level equaions ()-() and he shor-run dynamic coefficiens using he ARDL error correcion models. To ascerain he goodness of fi of he ARDL models, diagnosic and sabiliy ess are conduced. The diagnosic es examines he serial correlaion, funcional form, normaliy, and heeroscedasiciy associaed wih he model.the parameer sabiliy was esed using he CUSUM and CUSUMSQ ess. Finally, we esed he direcion of causaliy wihin he condiional Granger causaliy ess using he ARDL mechanism as a long-run conex. The F-saisics for he shor-run causaions and he saisics of ECTs(error correcion erm) for he long-run causaions mus be saisically significan o achieve Granger causaliy beween he shadow economy and he unemploymen raes.

8 .. Empirical resuls Invesigaing he possibiliy of coinegraion beween he shadow economy and he unemploymen raes using he bounds ess wihin he ARDL modeling approach, he empirical resuls show ha in shor-run boh ILO unemploymen rae and regisered unemploymen rae has negaive and saisically significan effec on shadow economy; however in he long-run boh unemploymen raes have a posiive effec on shadow economy. The long-run esimaed R_UR and ILO-UR coefficiens sae ha % increase in regisered unemploymen rae yields an average.4% increase in he size of he shadow economy, while a % increase in ILO unemploymen rae yields an average.7% increase in he size of he shadow economy. Finally, we esed he direcion of causaliy wihin he condiional Granger causaliy ess using he ARDL mechanism. Table.Resuls of Granger Causaliy for SE and R_UR and SE and ILO_UR The empirical resuls revealed he exisence of a long-run uni-direcional causaliy ha runs from he regisered unemploymen rae o he shadow economy. In able, he significance of F-saisic value confirms he shor-run causaliy from regisered unemploymen rae o shadow economy. The negaively and significance of error correcion erm provide evidence of long-run causaliy. The error correcion erm is significan wih he expeced sign ha confirm long-run unidirecional causaliy from unemploymen raes o shadow economy as we concluded hrough bound es. Therefore, in he long-run we have proved ha here is a unidirecional causaliy ha runs from unemploymen rae (regisered or ILO) o shadow economy. 4. Conclusions In his paper, we used he currency demand approach in order o obain a measure of he Romanian shadow economy expressed as % of official GDP and we invesigae he relaionship beween unemploymen rae and he size of he shadow economy using ARDL causaliy analysis. The Romanian shadow economy as % of official GDP is esimaed using vecor error correcion models, and is size is decreasing over he analyzed period, from 6.5% a he end of o abou.5% of real GDP a he middle of.

9 The dynamic relaionship beween shadow economy as % of official GDP obained hrough VECM models and unemploymen rae for he case of Romania has been invesigaed using quarerly ime series daa from -. To invesigae long-run causal linkages and shor-run dynamics, ARDL causaliy ess are applied. Coinegraion es resuls shows ha in shor-run boh ILO unemploymen rae and regisered unemploymen rae has negaive and saisically significan effec on shadow economy; however in he long-run boh unemploymen raes have a posiive effec on shadow economy. The ARDL causaliy resuls revealed he exisence of a long-run uni-direcional causaliy ha runs from he boh unemploymen raes o he shadow economy. One possible explanaion of he negaive impac in he shor run can be he he inabiliy of he labor marke o provide more jobs wheher hey are official or 'hidden'/ unregisered (clandesine in some exen of course) in he even of rising unemploymen, underlining a limiaion of opporuniies o work in he informal economy. Bu on he long run, an increase in unemploymen leads o an expansion of he aciviies in he shadow economy, underlining a posiive impac on he size of he shadow economy. References Albu, L.L., Nicolae, M.().Use of households survey daa o esimae he size of he informal economy in Romania, MPRA Paper 486, Universiy Library of Munich, Germany. Albu, L.L.(8). A Model To Esimae Spaial Disribuion Of Informal Economy, Journal for Economic Forecasing, Insiue for Economic Forecasing, vol. 5(4), - 4. Albu, L.L., Iorgulescu, R. and Sanica, C.(). Esimaing Hidden Economy and Hidden Migraion: The Case of Romania, Journal for Economic Forecasing, Insiue for Economic Forecasing, vol. (), Albu, L.L., Ghizdeanu, I. and Sanica, C.(). Spaial Disribuion of he Informal Economy. A Theoreical and Empirical Invesigaion, SCIENZE REGIONALI, FrancoAngeli Ediore, vol. (), 6-8. Ahumada, H., Alvaredo, F. and Canavese, A.(7). The moneary mehod and he size of he shadow economy: a criical assessmen. Rev, Income Wealh 5 (), 6 7. Cagan, P. (958). The Demand for Currency Relaive o he Toal Money Supply. Journal of Poliical Economy, 66(), 8. Davidescu A.A. and Dobre, I.(). Revisiing he relaionship beween U.S. shadow economy and he level of unemploymen rae using bounds es approach for coinegraion and causaliy, Economic Compuaion and Economic Cyberneics Sudies and Research, /, vol.46, pg.9-4, A, ISSN X. Dell Anno, R. and Solomon H.(7). Shadow economy and unemploymen rae in USA. There is a srucural relaionship? An empirical analysis, Annual Meeing of he European Public Choice Sociey, Finland. Giles, D.E.A. and Tedds, L.M. (). Taxes and he Canadian Underground Economy, Canadian Tax paper n.6. Canadian Tax Foundaion. Canada: Torono.

10 Johnson, S., Kaufmann, D. and Shleifer, A.(997). The unofficial economy in ransiion, Brookings Papers on Economic Aciviy, Fall, Washingon D.C. Johnson, S., Kaufmann, D. and Zoido-Lobaón, P.(998). Regulaory discreion and he unofficial economy. The American Economic Review, 88/, Narayan, P.K.(5). The saving and invesmen nexus for China: evidence from coinegraion ess, Applied Economics,7:7, Pesaran, M. H., Shin, Y. and Smih, R. J.(). Bounds Tesing Approaches o he Analysis of Level Relaionships, Journal of Applied Economerics, 6: Schneider, F. and Ense, D. (). Shadow Economies: Size, Causes and Consequences, Journal of Economic Lieraure 8, 7-. Schneider, F. and Buehn, A.(7). Shadow economies and corrupion all over he world: revised esimaes for counries, Economics - The Open-Access, Open-Assessmen E-Journal, Kiel Insiue for he World Economy, vol. (9 (Versio), -5. Schneider, F., Buehn, A. and Monenegro, C.(). Shadow Economies All Over The World: New Esimaes For 6 Counries From 999 To 7, Working Papers wp, Universiy of Chile, Deparmen of Economic. *** Quarerly Naional Accouns daabase, Eurosa. *** Quarerly Labor Force Survey daabase, Eurosa. *** Quarerly Governmen Finance Saisics daabase, Eurosa. *** Quarerly Ineres Raes daabase, Eurosa. *** Quarerly Moneary and Financial Saisics daabase, Eurosa. *** Employmen and Unemploymen daabase, Eurosa. *** Tempo daabase, Naional Insiue of Saisics, *** Monhly Bulleins of Naional Bank of Romania, -,

11