EFFECTS OF ASYMMETRIC OUTPUT SHOCKS AND OTHER VARIABLES ON POVERTY

Transcription

1 EFFECTS OF ASYMMETRIC OUTPUT SHOCKS AND OTHER VARIABLES ON POVERTY Dr Javed Iqbal Assstant Chef, Plannng Commsson Government of Pakstan Msbah Nosheen Assstant Professor Department of Economcs Hazara Unversty Mansehra Pakstan Tasawer Aslam Lecturer Educaton department Hazara Unversty Mansehra Shabana Parveen Lecturer Economcs department Hazara Unversty Mansehra Abstract Ths study analyses dfferent channels through whch stablzaton polces affect poverty n developng countres. It specally determnes asymmetrc effects of GDP per capta on poverty. The macroeconomc varables examned nclude publc expendture, nflaton, ncome levels and postve and negatve output growth, and the real exchange rate. In addton, several structural varables are also consdered. The data set s from 1990 to The estmaton method s a smple cross-secton, tme-seres OLS regresson model wth fxed effects. The results also show that there s no evdence of asymmetrc effect of the level of GDP per capta on poverty. The evdence proved n ths study shows that countres that have been successful n terms of economc growth are also very lkely to have been successful n reducng poverty. Key words: Poverty, asymmetrc growth, fxed effects, cross country. JEL classfcaton: I3, O1 COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 543

2 Asymmetrc Effects of Cycles and Crses on the poor 1. Introducton An ncreasng vew pont of emprcal research proposed that economc crses and cyclcal downturns may have an asymmetrc poverty effect. It mples that sharp output contractons or recessons may ncrease poverty rates extensvely; On the other hand, expansons have a tendency of contanng more lmted effect. To the extent that the austerty measures that are often at the core of macroeconomc adjustment programs lead to adverse short run output movements, and the asymmetrc effects may dmnsh consderably the benefts n longer run of adjustment for the poor. Therefore t s mportant to know about the causes of asymmetry. On the lnes of Agenor (2001), there are fve man causes of explanatons. The frst rests on parents' decsons concernng attendng of school by ther chldren school; the second s based on asymmetrc changes n expectatons and confdence factors; the thrd s a credt crunch at the frm level, wth ratonng nduced by ether adverse selecton problems or negatve shocks to net worth; the fourth emphaszes the mpact of borrowng constrants on household consumpton behavor; and the ffth s based on labor hoardng by frms facng hgh turnover costs. The remander of the paper s organzed as follows. Secton two explans fve alternatve channels through whch output shocks can have nonlnear effects on poverty. Secton III proposes fxed effects technque. Ths technque s then appled n Secton IV to a number of developed countres, usng panel data coverng the perod Fnally, Secton V summarzes the man results of the analyss, draws some polcy mplcatons. COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 544

3 2. Causes of Asymmetry 2.1 Schoolng Effects Durng an economc downturn a drop n real ncome may have an rreversble mpact on the human captal of the poor. Accordng to Lustg (2000), chldren n poor famles sometmes drop out of school and put to work n reacton to large adverse ncome shocks so compensatng the fall n the household's ncome. On the other hand, they are not sent back to school when ncomes ncrease agan. Thus, recessons generate some type of asymmetrc poverty effect n the sense that temporary negatve shocks have persstent effects. 2.2 Expectatons and Confdence Factors Durng recesson, consumers and frms may be more pessmstc than durng expansons when they are supposed to be more optmstc. If the overall economc poston s not lookng satsfactory by consumers and frms, a postve ncome shock n the form of credt constrants relaxaton may have a smaller mpact on prvate spendng decsons durng recessons than durng booms. Moreover, durng recessons f the perceved degree of uncertanty about future proftablty ncreases, even after a large postve polcy nduced shock, frms may be less wllng to nvest. If durng a recesson output and labor demand become less responsve to postve shocks, the ntal poverty ncrease nduced by hgher unemployment may be dffcult to reverse. 2.3 Adverse Selecton, Net Worth, and Credt Due to recesson, nterest rates may ncrease because slow economc actvty s supposed to ncrease the bankruptcy rsk. As a result, banks may ncrease the premum (Agenor and Azennan (1998)). An ncrease n the perceved rsk of default may also lead to a tghtenng of credt constrants f lenders are unwllng to lend to rsker borrowers, as mpled by adverse selecton models of the credt market (Jaffee and Stgltz (1990)). The tghtenng of credt COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 545

4 constrants may ncrease the mpact of the ntal recesson on borrowng and spendng. The resultng fall n labor demand and thus the effect on poverty may also be many folds. If expansons are not characterzed by an equvalent reducton n the perceved rsk of default, adverse selecton problems may mpart an asymmetrc bas to output shocks. 2.4 Borrowng Constrants and Household Consumpton Operatng at the household level, credt constrants may also represent a source of asymmetry n the response of poverty to output shocks. If durng recessons such constrants become bndng as a result of adverse selecton, or net worth effects, because household wealth may be also be adversely affected by sharp drops n asset prces, t may hnder the households ablty to smooth consumpton. The evdence suggests that n developng countres consumpton smoothng and rsk sharng reman extremely mperfect. The poorest households are least secured aganst shocks because of ther nablty to accumulate assets and because of asymmetrc nformaton problems hgh transacton costs whch may completely prevent access to prvate market nsurance or credt mechansms to smooth ncome fluctuatons for them. Resultantly, through borrowng and lendng consumpton smoothng s smply not feasble for poor households. 2.5 The Labor Hoardng Hypothess Because of an asymmetrc effect on employment and productvty, recessons may aggravate poverty (Agenor (1999, 2001)). In a downturn, unsklled poor workers are generally the frst to lose ther jobs as frms hoard ther sklled labor force because of hgh turnover costs durng recessons. Therefore frms are n general not on ther sklled labor demand curve and sklled workers wages are hgher than ther margnal product. Durng expansons, the prorty for frms s to COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 546

5 recover the productvty losses and foregone proft opportuntes. If the greater degree of complementarty between sklled labor and physcal captal s hgh, they therefore, ncrease fxed nvestment and reduce ther demand of unsklled labor. Ths, n turn, may produce a strong degree of persstence n unsklled unemployment and poverty n the early stages of an expanson. 3. Panel Data Estmaton Technque Here we dscuss the relevant panel data econometrc technques appled n ths study, namely fxed effects models. The panel specfcatons (equaton 3.1) do not allow the ntercept and coeffcent for crosssecton to dffer, therefore, these estmated coeffcents were fxed. y t X X... X 1 1, t 2 2t k kt t 3.1 Panel data regresson dffers from the regular tme-seres and cross-secton regresson n that t has a double subscrpt on ts varables, denotng the cross-secton, for example ndvduals, households, frms, countres etc. and t denotng tme. Although not very flexble, these specfcatons already recognzed the benefts, such as ncreased degrees of freedom and potentally lowerng the standard errors of estmated coeffcents of jont estmaton of coeffcents. The technque s to control for ndvdual heterogenety. It has the beneft of beng more nformatve, t has more varablty, less collnearty among the varables, more degrees of freedom and more effcency and are better able to study the dynamcs of adjustment and are also suted to study the duraton of economc states lke unemployment and poverty. The modern panel data specfcaton: / y X u t t t 3.2 COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 547

6 Where yt s the dependent varable, α s a scalar, t y t / X t and are k-vectors of non-constant explanatory varables and parameters for = 1, N cross-sectons and t = 1, T tme perods. Most panel data applcaton utlzes a one-way error component model for the dsturbances. If we assume cross-secton heterogenety n the error term of equaton 3.2 then the error can be specfed as: u t v 3.3 t Where denotes the unobservable ndvdual specfc effect and s tme-nvarant, accountng for the specal effect that s not ncluded n the regresson. The remander dsturbance vt vares wth cross-sectons and tme. The error of a panel then becomes the sum of a well-behaved error component ( v t ) and an ndvdual specfc effect ( ). These ndvdual specfc effects can be cross-secton specfc effects ( ) or tme-seres specfc effects ( ) and can be modeled utlzng fxed or random effects models. / yt X t ut 3.4 The fxed effects model allows the ntercept term to dffer across cross-sectons, whle keepng the slope coeffcents constant. The fxed effects model assumes that the ndvdual specfc effects are fxed parameters to be estmated and the remander dsturbance s ndependently and dentcally dstrbuted. A fxed effects model s approprate f the focus s on a specfc set of cross-sectons for example N frms, and the nference s restrcted to the behavour of these sets of cross sectons. Least square dummy varables estmaton models estmate the ndvdual specfc effects by makng use of dummy varables and OLS regresson. Ths fxed effects model estmates K parameters for the X-regressors, one parameter for the ntercept and N-1 parameters for the ndvdual fxed effects. The model s very appealng due to ts smplcty, but the loss of degrees COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 548

7 of freedom when estmatng a large number (N-K) of coeffcents places huge restrctons on the regresson. The problem of estmatng too many parameters can be solved by the Wthn Q estmaton. The Wthn estmaton dffer from the Least square dummy varable models n that the ndvdual effects are no longer estmated. The fxed effects are calculated by demeanng the data, subtractng the wthn mean from each varable, and estmatng OLS usng the transformed data. The Wthn model for cross secton ndvdual effects becomes a smple regresson: X X v v y y 3.5 t t Where y, and v denotes the average of each varable for each cross-secton over the tme X perod. The ndvdual cross-secton effects can be solved under the assumpton that 0, then the frst order condtons to recapture the ndvdual effects are: aˆ y ˆ X 3.6 ~ y ˆ ˆ X 3.7 N 1 The total ndvdual effect s the sum of the common ntercept and the constructed ndvdual component. The Wthn method avods the problem of too many parameter estmates but an mportant dsadvantage arses when demeanng the data n that X-regressors that are themselves dummy varables whch cannot be used because you wll wpe-out the ndvdual effects. One wll thus not be able to use dummy varables, whch capture for example sex, race or relgon, n your specfcaton when the wthn model s used. COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 549

8 4. An Applcaton to Developng Countres Ths secton presents, n a cross-secton emprcal framework, some prelmnary results on the relatonshp between macroeconomc factors and poverty n selected South Asan and Latn Amercans developng countres. The macroeconomc varables examned nclude publc expendture, nflaton, ncome levels and output growth, and the real exchange rate. In addton, several structural varables are also consdered. The data set s from 1990 to The dependent varable s the logarthm of the poverty rate (POV), measured by the poverty gap for the populaton as a whole. The explanatory set of varables used n the regressons s defned as follows. INF s the nflaton rate n consumer prces; SUB_GDP s the log of the rato of total subsdes and other current transfers over GDP, whch ams at capturng level effects of changes n publc spendng; ILL s the youth llteracy rate n percent of the populaton aged 15-24, whch ams to capture the level of educaton of the labor force; HOS s the log of hosptal beds per 1,000 persons, whch measures overall health condtons; GDP s GDP per capta at PPP exchange rates, whch captures the level of economc development; GDPGR s the annual growth rate of GDP per capta, measured at PPP exchange rates, whch can be vewed as ether a proxy for the rate of return on nvestment, or as a measure of cyclcal movements n output; EXC s the annual rate of change of the real exchange rate (defned such that an ncrease s a deprecaton); COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 550

9 URB s the log of the relatve share of the urban populaton n proporton to total populaton; TOT s the log of the terms of trade; IMEX s the rato of the sum of mports and exports of goods and servces n percent of GDP, and ams to capture exposure to external shocks. Theoretcally t s suggested that nflaton should have a postve effect on poverty. The poor are more exposed to nflaton than hgher-ncome groups as a result of a varous reasons. For example, ther ncome s often defned n nomnal terms, and they often do not beneft from ndexaton mechansms. In perods of hgh nflaton, therefore, the purchasng power of ther resources may fall. Moreover, they have few real assets, such as land, and usually no ndexed fnancal assets-wth whch to protect themselves from the effect of prce ncreases. The fscal varables, subsdes and current transfers as a proporton of GDP have a pror an ambguous effect. The effect of an across-the-board cut n transfers and subsdes, for nstance, may be negatve; but to the extent that t s accompaned by better targetng, there may be no sgnfcant effect on the poverty rate. An ncrease n the llteracy rate s expected to be postvely correlated wth poverty, whereas, an mprovement n health ndcators should be nversely related to poverty. Both the level of GDP per capta and ts rate of growth are expected to be negatvely correlated wth the poverty rate. The effect of real exchange rate deprecaton s n general ambguous; t s lkely to lead to a reducton n poverty f t benefts small farmers n the tradable sector, as s the case n many low-ncome developng countres. A hgher rate of urbanzaton also has n general an ambguous effect on aggregate poverty; to the extent that rural-to-urban mgraton translates nto greater access to publc servces n urban areas, t may lower poverty; COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 551

10 but to the extent that nflows of workers n the urban sector lead to an excess supply of labor n the nformal sector and lower wages therefore, t may ncrease poverty. An mprovement n the terms of trade may reduce poverty f t represents an ncrease n the relatve prce of agrcultural commodtes (thereby beneftng small farmers n rural areas) or a fall n the prce of mported consumpton goods (beneftng mostly households n urban areas). An ncrease n macroeconomc volatlty (assocated wth output shocks, nflaton, or fluctuatons n the real exchange rate) s expected to ncrease poverty, possbly through ts adverse effect on growth, as dscussed earler. Fnally, the degree of openness has a pror an ambguous effect on poverty. Recent research on the mpact of economc crses on the poor has emphaszed that temporary negatve ncome shocks may have permanent effects on poverty. It has also been recognzed that over the short run, busness cycles and economc crses may have an asymmetrc effect on poverty whch means that poverty may ncrease more durng perods of economc contractons than t decreases n an equvalent expanson. To account for the possblty of asymmetrc output shocks n both the level of real GDP per capta and ts growth rate, observatons on the level and growth rate of GDP per capta are splt nto two sub-samples. In both cases, two dummy varables, dummy postve and dummy negatve, are created; for the level of GDP per capta, the frst s equal to 1 tmes GDPPC when GDPPC at perod t s hgher than GDPPC at tme t - 1 (and 0 otherwse), and the second s equal to 1 tmes GDPPC when GDPPC at perod t s lower than GDPPC at tme t - 1. For the growth rate, the frst s equal to 1 tmes REALGR when REALGR s postve and 0 otherwse and the second equal to 1 tmes the absolute value of REALGR when REALGR s negatve and 0 otherwse. COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 552

11 Table: 1 Asymmetrc Effects on Poverty Rate (OLS wth fxed effects) Dependent varable: Log of poverty gape (1) (2) (3) (4) (5) Inflaton (-0.86) (-0.919) (-1.232) (-2.25) (-1.819) Illteracy (-1.046) (-0.926) (-2.757) Hosptal beds (-0.73) (-0.269) (-0.929) (-0.499) NEGDR (0.733) (-0.765) (-0.824) (-0.694) (-0.826) POSGR (-0.643) (-0.852) ( ) (-1.072) (-0.649) NEGPC (2.238) (2.220) (2.063) (2.246) (2.171) POSPC (1.223) (1.205) (1.135) (1.275) (1.144) Exchange rate (3.053) (3.114) (2.465) (3.402) (3.063) Openness (0.502) (1.718) Urbanzaton (1.397) (1.341) (2.812) Terms of trade (-2.338) (-2.243) (-1.28) (-2.289) (-2.347) COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 553

12 Adj. R Number of obs Std error of Regresson Note: Values n parentheses are t statstcs Table 1 summarzes some of the basc emprcal results. They ndcate that nflaton, hosptal beds and openness, do not have a statstcally sgnfcant mpact on poverty. The results also show frst that GDP per capta both postve and negatve, exchange rate and terms of trade are sgnfcant. The fact that the level of nflaton has ncorrect sgn and not sgnfcant may have far more to do wth the way t has been defned and used n ths analyss. Snce t s pckng up only the level of nflaton n that year and the poverty levels are based on observatons wth dfferent tme ntervals and hence reflectng changes over that tme perod as well as economc developments n that partcular year, t would be dffcult to solate the mpact of nflaton n that year unless t was dramatcally hgh. The llteracy rate, and the volatlty of exchange rate were not only nsgnfcant, they also had the wrong sgn. The rate of deprecaton of the real exchange rate also has also a negatve effect on poverty, whch s smlar wth the vew that mprovements n the relatve prce of tradables beneft farmers producng exportables n the agrcultural sector. The degree of openness has a postve but nsgnfcant effect on poverty. Ths postve mpact of openness on poverty can be explaned by the fact that greater access to mported captal goods leads frms to substtute away from unsklled labor n the producton process, whch results n an COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 554

13 ncrease n unemployment whch ultmately wll tend to ncrease poverty. Ths s n accordance wth the result gven by Agenor (2002). Urbanzaton has a postve sgn and s sgnfcant n some regressons so we can say that nflows of workers n the urban sector leads to an excess supply of labor n the nformal sector and lower wages there and as a result t ncreases poverty. An mprovement n the terms of trade can decrease poverty f t represents an ncrease n the relatve prce of agrcultural commodtes (thereby beneftng small farmers n rural areas) or a fall n the prce of mported consumpton goods (beneftng mostly households n urban areas). Ths s the case n our analyss. The coeffcent has a negatve sgn and statstcally sgnfcant almost n all cases. The results also show that there s no evdence of asymmetrc effect of the level of GDP per capta on poverty: both varables have coeffcents that are ndstngushable. Postve growth rates of output have no statstcally sgnfcant effect on poverty, whereas hgher negatve growth rate of per capta affect sgnfcantly and postvely the poor. Fnally, the fxed effects are all statstcally sgnfcant, suggestng that country-specfc factors are mportant n determnng the behavor of poverty rates. Conclusons The paper analyses the varous channels through whch stablzaton polces affect poverty n developng countres. It determnes asymmetrc effects of GDP per capta on poverty. The macroeconomc varables examned nclude publc expendture, nflaton, ncome levels and postve and negatve output growth, and the real exchange rate. In addton, several structural COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 555

14 varables are also consdered. The data set s from 1990 to The estmaton method s a smple cross-secton, tme-seres OLS regresson model wth fxed effects. The results also show that there s no evdence of asymmetrc effect of the level of GDP per capta on poverty; both varables have coeffcents that are mpossble to dfferentate. Postve growth rates of output have no statstcally sgnfcant effect on poverty, whereas hgher negatve growth rate of per capta affect sgnfcantly and postvely the poor. The evdence proved n ths study show that countres that have been successful n terms of economc growth are also very lkely to have been successful n reducng poverty so government s target on poverty-reducton must therefore create an envronment that s benefcal to growth. COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 556

15 References Agénor, Perre-Rchard, (1998) Stablzaton Polces, Poverty and the Labor Market: Analytcal Issues and Econometrc Evdence, unpublshed, the World Bank. Agénor, Perre-Rchard, (2001) Asymmetrc Effects of Monetary Polcy Shocks, unpublshed, the World Bank. Agénor, Perre-Rchard, and Joshua Azenman, (1998) Contagon and Volatlty wth Imperfect Credt Markets, IMF Staff Papers, 45, Agénor, Perre-Rchard, and Joshua Azenman, (2000) Savngs and the Terms of Trade under Borrowng Constrants, Workng Paper No. 774, Natonal Bureau of Economc Research. Forthcomng, Journal of Internatonal Economcs. Bernanke, Ben S., and Mark Gertler, (1989) Agency Costs, Net Worth and Busness Fluctuatons,'' Amercan Economc Revew, 79, Bernanke, Ben S., Mark Gertler, and Smon Glchrst, (2000) The Fnancal Accelerator n a Quanttatve Busness Cycle Framework,'' n Handbook of Macroeconomcs, ed. by John B. Taylor and Mark Woodford, North Holland. De Ferrant, Davd, Gullermo E. Perry, and Lus Servén, (2000) Securng our Future n a Global Economy, the World Bank. De Janvry, Alan, and Elsabeth Sadoulet, (2000) Growth, Poverty, and Inequalty n Latn Amerca: A Causal Analyss, Revew of Income and Wealth, 46, Demrguc-Kunt, Asl, Enrca Detragache, and Poonam Gupta, (2000) Insde the Crss: An Emprcal Analyss of Bankng Systems n Dstress, Workng Paper No. 00/156, Internatonal Monetary Fund. Domac, Ilker, (1999) The Dstrbutonal Consequences of Monetary Polcy: Evdence from Malaysa, PRE Workng Paper No. 2170, the World Bank. Dxt, Avnash, and Robert S. Pndyck, (1994) Investment under Uncertanty, Prnceton Unversty Press (Prnceton, New Jersey. Hodrck, Robert J., and Edward C. Prescott, (1997) Postwar U.S. Busness Cycles: An Emprcal Investgaton,'' Journal of Money, Credt, and Bankng, Horton, Susan, and Dpak Mazumdar, (2001) Vulnerable Groups and the Labor Market: The Aftermath of the Asan Fnancal Crss, n East Asan Labor Markets and the Economc Crss, ed. By Gordon Betcherman and Rzwanul Islam, the World Bank. COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 557

16 Jackman, Rchard, and John Sutton, (1982) Imperfect Captal Markets and the Monetarst Black Box: Lqudty Constrants, Inflaton and the Asymmetrc Effects of Interest Rate Polcy,'' Economc Journal, 92, Jaeger, A., and M. Parknson, (1990) Testng for Hysteress n Unemployment: An Unobserved Components Approach, Emprcal Economcs, 15, Jaffee, Dwght, and Joseph Stgltz, (1990) Credt Ratonng,'' n Handbook of Monetary Economcs, Vol. II, ed. by Benjamn M. Fredman and Frank H. Hahn, North Holland. Levnsohn, James, Steven Berry, and Jed Fredman, (2001) Impacts of the Indonesan Economc Crss: Prce Changes and the Poor, unpublshed, Unversty of Mchgan. Lokshn, Mchael, and Martn Ravallon, (2000) Short-Lved Shocks wth Long- Lved Impacts? Household Income Dynamcs n a Transton Economy, PRE Workng Paper No. 2459, the World Bank. Lustg, Nora, (2000) Crses and the Poor: Socally Responsble Macroeconomcs, Workng paper No. POV-108, Inter-Amercan Development Bank. Macklem, T., (2000) Asymmetry n the Monetary Transmsson Mechansm: What can we Learn from VARs?,'' unpublshed, Bank of Canada. Morduch, Jonathan, (1995) Income Smoothng and Consumpton Smoothng, Journal of Economc Perspectves, Wodon, Quentn, Rodrgo Castro-Fernandez, Khoon Lee, Gladys Lopez- Acevedo, Cornne Saens, Carlos Sobrado, and Jean-Phlppe Tre, (2001) Poverty n Latn Amerca: Trends ( ) and Determnants, unpublshed, the World Bank. World Bank, (2000) Global Economc Prospects, the World Bank. World Bank World Development Indcators (Varous Issues) CD-Rom, Washngton, D.C.: World Bank. COPY RIGHT 2012 Insttute of Interdscplnary Busness Research 558