Structural Cange in Houeold Income Inequality in Korea: A New Decompoition Including Labor Supply Cange Culee Lee Scool of Economic Seoul National Univerity San 56-1, Sillim-dong, Kwanak-gu Seoul, Korea clee@prome.nu.ac.kr January 2003 Ti work wa upported by Korea Reearc Foundation Grant (KRF-2001-041-C00234). I tank two anonymou referee for teir elpful comment and uggetion, and Jiyun Kim for er excellent reearc aitance.
Abtract Ti tudy etimate te relative contribution of eac of te component of oueold income, including labor upply, to te oberved cange in oueold income inequality between 1988 and 1999. For te period 1988 to 1993, reduced inequality in te ourly wage of ead wa te dominant caue of te improvement in te oueold income ditribution. For te period 1997 to 1999, on te oter and, cange in oter income, along wit cange in te wage of ead, wa an important contributor to te widening of te income diparity. Canging our of ead wa alo a nontrivial factor, accounting for 15% of te rie in income inequality during te period. Te relative contribution of canging labor upply to te rie in income inequality after 1997 migt be muc greater, if te effect of canging employment of ead could be conidered. 2
1. Introduction After nearly a decade of eiter declining or table trend ince te mid 1980, te family income inequality in Korea arply increaed in te coure of te financial crii, and remained ig even after te economy recovered from te receion. According to te etimate provided by Korea National Statitical Office baed on te Houeold Income and Expenditure Survey, Gini ratio fell from 0.3115 in 1985 to 0.2817 in 1993, remained relatively table between 1993 and 1997, and ten rapidly roe from 0.2830 in 1997 to 0.3210 by 1999. Toug Gini ratio ligtly fell by te next year, it tayed a ig a 3174. Similar pattern are found from oter meaure of income inequality, uc a Atkinon index, income are by quintile, and percentile ratio (Lee and Hwang 1998, Cung and Coi 2001, Yoo 2001, Cung et. al. 2002). 1 A an important legacy of te financial crii, te dramatic rie in te extent of income inequality a drawn a great deal of attention from economit. 2 In particular, a number of tudie ave analyzed te ource of te cange in te tate of income ditribution. A recent tudy by Cung et al. (2002, capter 3), for example, decompoed te ource of te increae in family income inequality between 1997 and 2000. Tey uggeted tat te cange in te earning of oueolder explain more tan 100% of te oberved increae in family income inequality. Tey alo found tat te rie in earning inequality between 1997 and 2000 wa largely accounted for by te growt in inequality witin education or age group, rater tan between-group or compoitional 1 Due to te lack of completely reliable income data tat cover te entire oueold, tere i no definite conenu regarding te trend of oueold income inequality in Korea, epecially for te period prior to 1990. An (1997), for intance, reported tat te extent of income inequality roe during te 1980, wile many oter tudie found improvement in income ditribution, at leat from te mid-1980 (Kang and Hyun 2001). 2 See Lee and Ree (1999) for more general dicuion of te origin and impact of te Aian Currency Crii. 3
cange. Since earning are determined by our of work and ourly wage, an increae in te diparity in earning may reult from te following two ource: cange in te ditribution of ourly wage and of our worked. In pite of ti imple formula for earning inequality, te potential contribution of cange in our and employment to earning inequality a not been invetigated. Even in te United State, for wic finding te caue of te rie in income inequality during te 1980 wa one of te mot popular iue in te field of economic, only a few tudie ave examined te role of canging labor upply. Burtle (1990, 1993) and Moffitt (1990), for example, reported tat te growt in annual earning inequality for male worker during te 1970 and 1980 wa primarily due to growing inequality in ourly wage rate. On te oter and, Haveman (1996) uggeted tat te increae in earning inequality between 1973 and 1988 among working-age men wa largely produced by increaed variability in te amount tat potential breadwinner worked. 3 Te previou tudie on te U.S., mentioned above, ave largely focued on male worker. However, pattern of individual labor force participation and of our of work need to be undertood in te context of joint deciion made by family member. For example, fewer working our by a family ead migt be upplemented by increaed our worked by te poue or oter family member. Te exiting literature i nearly ilent about ow cange in employment and our of family ead and poue jointly contributed to te family income inequality. Burtle (1993) conider ti problem, but only incompletely. He rank individual into earning quintile according to family 3 Tee tudie ue variance of logaritm (VLN) of earning a te meaure of inequality. Tey decompoe te VLN of earning into (1) VLN of wage rate, (2) VLN of our worked, and (3) a covariance term between te two. Tey ten oberved cange in eac factor over time to 4
earning rater tan individual earning, and calculate counterfactual mean earning for eac earning quintile, auming no cange in te mean employment rate and our of work in eac earning group. He aee te contribution of te employment and our cange by comparing te actual and counterfactual cange in Gini-ratio for te period 1969-1979. Te tudy, after all, deal wit individual earning of men and women rater tan family earning. More recently, Lee (2000, 2001) contributed to te literature by developing a new decompoition metod baed on family income. Ti tudy found tat cange in labor market activity of family ead accounted for alf of te increaed gap between familie in te top and bottom income decile. Te purpoe of ti paper i to etimate te relative contribution of cange in eac element of oueold income, including our and employment of te ead and poue, to te cange in income inequality in Korea between 1988 and 1999. I ue a matced micro ample of te Urban Houeold Expenditure Survey and te Economically Active Population Survey. A noted above, te extent of family income inequality fell from te mid-1980 and reaced te lowet point in 1993. It remained table from 1993 to 1997, jumped during te financial crii, and ten peaked in 1999. Accordingly, decompoition are made for te tree ub-period, 1988-1993, 1993-1997, and 1993-1999. In particular, te decompoition analye given below will focu on te period 1988-1993 and 1997-1999 during wic te meaure of income inequality canged viibly. To my knowledge, te preent tudy i te firt attempt to decompoe te cange in te family income inequality in Korea into te contribution of labor upply and wage. Te organization of ti paper i a follow: I begin wit developing a metod of ae it relative contribution to te increae in te VLN of earning. 5
decompoing a cange in family income in te next ection. It will alo introduce te data ued in te analyi. Section 3 decribe te pattern of cange in te element of oueold income inequality. In ection 4, I preent te reult of te decompoition, explaining ow eac of te component of oueold income contributed to te cange in income inequality. I alo examine in te ame ection ow te reult cange were alternative meaure of inequality are ued. Te final ection ummarize te paper. 2. Metod and Data I begin wit a decompoition of total oueold income into everal component. Te definition of variable repreenting eac of te component i given in Table 1. Te average montly money income of oueold in a given income decile, denoted N, may be given a were H, H, (1) N H W P + H W P δ Q + W, and W tand for te mean montly our worked and te mean ourly wage rate for employed ead and poue, repectively; P and P tand for te employment rate for ead and, if married, poue; δ i te fraction of oueold in wic bot uband and wife are preent; and finally, Q tand for te mean income from oter ource. <Table 1 Here> A te meaure of income inequality, I ue te difference in te log of average oueold income between two income decile, ay, te top and te bottom income decile. Ti meaure, denoted by N, i te difference between te logaritm of te average income in te top decile and te logaritm of te average income in te bottom decile. Tat i, TOP N = ln[ N ] ln[ N BOTTOM ]. Uing an approximation, N can be 6
decompoed a (2) N Φ ( H + W + P ) + Φ ( H + W + P + δ ) + Φ ( Q ) were te aterik denote te difference in te log of a variable in te top and bottom decile of oueold. For example, TOP H = ln( H ) ln( H BOTTOM ). And Φ denote te weigt of eac of te tree income ource. For example, Φ [ ( H W P ) / N] indicate te earning of te ead a a proportion of te total oueold income. Q = It i poible to decompoe te cange in N over time, in ti cae, ay, between 1997 and 1999, by differentiating equation (2) totally, to obtain (3) N Q Φ H Φ Φ H Q + Φ + Φ W Q + Φ W Q + Φ P + Φ P + Φ ( H + Φ δ + Φ ( H + W + W + P ) + + P + δ ) + Te firt term on te rigt-and ide of equation (3), for example, repreent te rate of cange in te diparity in average our worked by oueold ead in te top and bottom decile, weigted by te relative are of oueold income derived from te earning of te ead. Te etimate of ti term will indicate te relative contribution of te cange in average working our of te ead to te rie in te meaure of income inequality between 1997 and 1999. Likewie, te econd and tird term ow te relative contribution of cange in ourly wage and in te employment rate of te ead, repectively. On te oter and, te fourt term repreent te effect of canging weigt, tat i, te relative importance of earning of te oueold ead a a ource of income. If earning of oueold ead are more unequally ditributed tan oter ource of income, an increae in te are of earning of te ead in te total oueold income would produce a rie in te magnitude of inequality. Te balance of ti paper i baed on a matced ample of Urban Houeold 7
Income and Expenditure Survey (UHS, ereafter) and Economically Active Population Survey (EAP) for te year 1988, 1993, 1997, and 1999. 4 UHS provide information on te ize of income from eac ource for urban oueold in wic two or more peron reide. In pite of a number of limitation, epecially te incomplete coverage of te population, UHS i te mot widely ued ource for meauring te extent of income inequality in Korea. 5 Te information on employment and our of work for oueold ead and poue come from EAP. 6 Te ample ued in ti tudy i a elected ample of UHS. Te oueold for wic te information on income i miing and toe not matced to EAP are excluded from te ample. Te number of oueold included in te ample to be ued i 33,671 for 1988 (66% of te entire UHS 4 Te Economically Active Population Survey (EAP) i conducted during te week containing te 15 t day of every mont. Te primary purpoe of EAP i to collect up-to-date information on te economic tatu of te population and on cange in te activity pattern of te labor force. Te ample to be urveyed i elected baed on multitage cluter ampling metod. Te ample for te 2000 EAP, for intance, i drawn from approximately 22,029 ordinary enumeration ditrict (ED ) wic are 10% of te 1995 Population and Houing Cenu. Te ED are reduced to Primary Sampling Unit (PSU), to Secondary Sampling Unit (SSU), and ten to Ultimate Sampling Unit, troug tree tage of ample election. In 2000, for example, 29,529 oueold out of 1,231 PSU were elected a ample of te 1995 Population and Houing Cenu. Te average number of oueold in te 2000 montly urvey wa about 28,807 oueold. Te purpoe of te Urban Houeold Income and Expenditure Survey (UHS) i to collect information on urban oueold income and expenditure for analyzing te variation in te level of living and te diparitie among te different ocio-economic group, and for obtaining weigt for te contruction of te conumer price index. Te ample of UHS i baed on te EAP ample. Te average number of oueold in te 2000 montly urvey, for example, wa about 5,216 oueold. See Korea National Statitical Office (2001a, 2001b) for more detailed feature of te two urvey. 5 By excluding ingle-peron oueold and rural oueold, UHS repreent 63.8% of te entire oueold. Furtermore, about 60% of te ample oueold are quetioned teir income. Terefore, le tan 40% of te entire oueold are covered by te income tatitic provided in UHS (Kang and Hyun 2001). 6 It report te our worked for te week preceding te urvey. Te montly our of work wa calculated by multiplying te weekly our by four, auming no cange in te our for te entire mont. 8
ample), 39,988 for 1993 (66%), 38,590 (63%) for 1997, and 33,928 for 1999 (59%). 7 For tudying te cange in income inequality, it would be ideal to ue te ample of all peron regardle of employment tatu. Since UHS doe not report income for oueold woe ead are not employed, owever, te analyi in ti paper a to rely on te ample of oueold wit employed ead. Accordingly, te decompoition metod preented in equation (3) i modified a follow, excluding te term repreenting te contribution of cange in employment of oueold ead. (4) N Q Φ H Φ Φ H Q + Φ + Φ W Q + Φ W Q + Φ P + Φ ( H + W ) + + Φ δ + Φ ( H + W + P + δ ) + A in oter tudie baed on UHS, it ould be noted tat ti tudy only deal wit te income inequality of te urban oueold wit two or more peron and an employed ead. 8 3. Cange in te Component of Houeold Income, 1988-1999 It i well documented tat variou meaure of oueold income inequality fell between 1988 and 1993, remained table from 1993 to 1997, and ten roe arply tereafter. Te income gap between te top and bottom income decile, te primary meaure of oueold income inequality employed in ti tudy, ow a imilar 7 Te ample for eac year wa contructed by pooling te data for montly urvey. Terefore, different montly urvey for te ame oueold, interviewed multiple time during te urvey year, are counted a different obervation. 8 A drawback of te inequality meaure ued in ti tudy i tat it doe not take into account te difference in te ize of te oueold between income decile. If te average ize of ric and poor oueold canged in a different manner, te oberved cange in te meaure of oueold income inequality may ditort te actual cange in te diparity in economic wellbeing. A preented at te bottom of Appendix Table, owever, te pattern of cange in te oueold ize i igly imilar acro different income decile. Terefore, te potential problem ariing from ignoring te equivalence cale ould not be eriou. 9
pattern. 9 A preented in Figure 1, te average income of low-income oueold grew fater tan tat of ig-income oueold between 1988 and 1993. During ti period, te average income of te bottom 10% oueold increaed by 140%, a compared to a 110% increae among te oueold in te top income decile. A a conequence, te difference in te log of te total oueold income between te top and bottom decile, denoted N above, dropped by 0.13. During te following ix year, on te oter and, te diparity in oueold income remained nearly uncanged: te average oueold income for te top and bottom decile increaed by, repectively 54% and 55%. After te financial crii, in contrat to te year prior to 1993, te average income for te bottom 10 t oueold fell arply (-20%), wile te income for oueold in te top decile grew at a repectable rate (10.5%). Accordingly, 1997 and 1999. 10 <Figure 1 Here> N increaed by 0.32 between Figure 2 indicate tat ift in te ourly wage rate of oueold ead ( W ) wa a igly important caue of te cange in te oueold income inequality between 1988 and 1993, and between 1997 and 1999. From 1988 to 1993, W grew by 176% for te ead of te bottom 10% oueold; te growt rate of W wa only 97% for te ead of te ricet 10% oueold. Between 1997 and 1999, on te oter and, W for te top 10% oueold increaed by 7%, wile W for te bottom income decile dropped 17%. Since te earning of oueold ead account for a dominant fraction of te total oueold income, tee uneven cange in W ould 9 Eac component of te average oueold income for eac of income decile i reported in Appendix Table. 10 If oueold income per capita i ued, te pattern of cange over time in te extent of income inequality and te reult of decompoition of canging income diparity do not cange muc, becaue te difference in te average oueold ize between te poor and te ric 10
ave exerted a trong effect on te trend of income inequality. <Figure 2 Here> Te weekly our of work for employed ead ( H ), a preented in Figure 3, ubtantially declined during te period under invetigation. For te period between 1988 and 1993, te magnitude of te decline in H wa relatively even acro different income decile, 6% and 4.5% for te bottom and top income decile, repectively. Terefore, te cange in H ould ave been a mild countervailing force of te decline in te income gap between te ric and te poor for ti period. For te period from 1997 to 1999, H fell for te lowet 10% oueold (-6%), wile it remained table for te top income decile. We may predict from ti pattern tat uneven cange in H wa at leat partly reponible for te rie in te oueold income inequality after 1997. <Figure 3 Here> Over te period under tudy, te ourly wage for employed poue ( W ) grew even more rapidly tan te pattern of te cange in W. Figure 4 ow tat for te period between 1988 and 1993 W wa imilar to tat of W ; a muc greater increae for te bottom income decile (180%) in comparion wit te rie for te top income decile (100%). For te period 1997-1999, owever, W increaed evenly acro different income decile, different from W of wic growt wa largely concentrated in igincome oueold. For ti period, terefore, te cange in ource of te rie in te oueold income inequality. <Figure 4 Here> W ould not be a major A te female labor-force participation rate increaed, te proportion of remained table over time. See Appendix Table. 11
employed poue ( P ) roe coniderably, epecially between 1988 and 1993. A indicated by Figure 5, te rie in P wa generally more pronounced among igincome oueold. For te top income decile, P increaed from 20% to 36% between 1988 and 1993. On te oter and, tere wa no gain for wive in te lowet 10% oueold. During te period from 1997 to 1999, P increaed from 8% to 10% for te lowet income decile, wile it fell by 3% for te ricet 10% oueold. Tee pattern indicate tat te labor-force participation of poue canged in te oppoite direction of te ift in oueold income inequality. Suc a countervailing effect wa particularly trong for te period between 1988 and 1993, in wic te decline in te income gap between te ric and te poor would ave been even greater ad tere been no cange in P. <Figure 5 Here> Te average weekly our of work of employed poue ( H ) ave diminied even more rapidly tan H ince 1988. Figure 6 ow tat te decline in H wa coniderably greater for poue in low-income oueold, wo initially worked longer, tan for toe in ig-income oueold. For te period between 1988 and 1993, in particular, H fell a muc a 13% for te bottom 10% oueold, wile of te oueold in te top income decile decreaed only by 3%. From 1997 to 1999, te fall in H wa larger for te ig-income oueold tan for te poor, but only by a relatively mall margin. Ti implie tat te cange in H H limited te extent of te improvement in oueold income inequality for te period 1988-1993, and offet te effect of oter factor leading to te widening of te income diparity for te year 1997 troug 1999. 12
<Figure 6 Here> Oter income (Q) i calculated a te difference between te average total oueold income and te average earning of ead and poue. Tu, it include earning of oter oueold member and non-labor income. Between 1988 and 1993, te rate of increae in Q wa muc iger for te oueold in te top income decile (142%) tan for te lowet income decile (41%). Tu, te cange in Q in toe year ould ave played te role of a countervailing factor of declining income inequality. For te period 1997-1999, too, te cange in Q ould ave ad a coniderable effect on oueold income inequality, becaue it roe 18% for te top decile, and fell 19% for toe in te bottom 10% oueold. Finally, te proportion of oueold woe ead are married (δ) ould ave ad little effect on te oueold income inequality, becaue it canged in a imilar manner for bot te ric and te poor. 4. Decompoition of te Cange in Houeold Income Inequality Te pattern of te cange over time in te component of oueold income, reviewed in te preceding ection, allow u to predict te direction of teir contribution to te ift in oueold income inequality. For intance, it wa uggeted tat ourly wage of ead ould ave been a major factor tat caued te decline in oueold income inequality between 1988 and 1993, and te dramatic turnaround of te trend after financial crii. Alo, it wa anticipated tat te cange in te our of work for ead ould ave contributed to te widening of te income gap between te ric and te poor for te period 1997 to 1999. In ti ection, I analyze in detail ow cange in eac of te component contributed to te ift in te oueold income inequality. I begin wit a baeline decompoition of te cange in te difference in te log of te 13
total oueold income between te top and bottom income decile, baed on equation (4). <Table 2 Here> 4.1. Income Gap between te Top and Bottom Income Decile A. 1988-1993 For te period between 1988 and 1993, in wic te difference in te log of income dropped by 0.13, it turn out tat reduced inequality in te ourly wage of ead explain more tan 100% of te decline in te meaure of oueold income inequality. In fact, ad tere been no cange in oter factor, te cange in W would ave produced a decline in te meaure of income inequality twice a large a te actual decline in magnitude. Te cange in te wage of employed poue ( W ) played a upporting role, accounting for anoter 15% of te decreae in te income gap. Te trong equalizing effect of te cange in wage wa offet to a coniderable extent by te cange in oter component of oueold income. In particular, te ift in oter income (Q) and te are of te earning of poue ( Φ ) eliminate, repectively, 78% and 58% of te overall decline in te meaure of inequality. Te ditribution of Q became more unequal, leading to a rie in income inequality. 11 An increae in te are of poue earning affected te meaure of inequality in te ame direction, becaue it wa more unequally ditributed tan te total oueold income. 12 In addition, te cange in te proportion of employed poue ( P ), wic reulted from te dramatic 11 Te difference in te log of oter income between te top and te bottom income decile increaed by 0.538 between 1988 and 1993. 12 Altoug te earning of employed poue are more equally ditributed tan te total oueold income, te proportion of oueold wit employed poue i muc lower among te poor tan among ig-income oueold. 14
rie in te labor-force participation of poue in iger-income oueold, offet anoter 25%. Meanwile, a decline in te are of oter income ( Φ ) reduced te meaure of inequality, becaue it ditribution wa more unequal tan tat of te total income. q B. 1997-1999 For te period between 1997 and 1999, too, te cange in te wage of ead ( W ) wa te major reaon for te rie in income inequality, altoug it relative contribution (54%) wa muc maller tan in te previou five year. An increae in te inequality in oter income (Q) made a coniderably large contribution (28%) to te widening of te income gap. A rie in te are of oter income ( Φ ) explain anoter 10%. It i alo notable tat canging our of ead ( H ) wa a nontrivial factor, accounting for 15% of te increae in te meaure of inequality. On te oter and, a decline in te are of ead earning ( Φ ) wa a countervailing factor; te increae in te meaure of inequality would ave been 8% greater witout it. q 4.2. Rigt Tail v. Left Tail of te Income Ditribution A widening (reduction) of te income diparity between te top and bottom income decile could reult from eiter a fater (lower) growt of income of te ricet 10 t oueold or deterioration (improvement) of te poition of te lowet 10 t oueold, in comparion wit te oueold in te middle. Likewie, te relative contribution of eac of te component of oueold income could come from eiter left or rigt tail of te income ditribution. Similar decompoition, alo baed on 15
equation (4), are conducted eparately for te difference in te log of income between te top 10 t and te average, and between te average and te bottom 10 t oueold. Te reult are reported in Table 3 and 4. <Table 3 and 4 Here> A. 1988-1993 For year 1988 to 1993, te overall decline in income inequality come equally from bot tail of te income ditribution [ee raw (1) in Table 3 and 4]. Te effect of canging wage of ead ( W ) wa alo evenly allocated in bot ide [ee raw (3) in Table 3 and 4]. However, te contribution of te cange in oter income (Q), te mot important countervailing factor of te decline in te inequality, i largely concentrated in te lower alf of te ditribution [ee raw (10) in Table 3 and 4]. Te cange in te weigt of oter income ( Φ ) reduced te diperion of income only for te rigt-and q ide of ditribution. At te oppoite ide, it actually increaed te inequality. Te reult ow tat te overall contribution of H and Φ were mall, becaue tee variable in eac ide canged in te oppoite direction [ee (2) and (11) in Table 3 and 4]. B. 1997-1999 For tee year, about 60% of te increae in te income difference between te top and bottom 10 t oueold were produced by deterioration of te poition of te pooret 10 t relative to te average oueold. Ti confirm te reult of earlier tudie reporting tat te arp rie of income inequality among urban working familie wa mainly due to te collape of low-income oueold (Cung et al. 2002). For te upper alf of te income ditribution, te cange in oter income (Q), wa te major 16
caue of riing inequality. Growing diparity in Q account for te larget portion (47%) of te increaed income gap. Te increae in te weigt of oter income ( Φ ) explain anoter 8%. On te oter and, canging wage of ead are reponible for 36% of te increae in te inequality. For te poorer ide of te income ditribution, canging wage of ead account for nearly 70% of te oberved increae in income inequality. Te te relative decline in te our of work among ead of low-income oueold wa te econd mot important factor, explaining 24%. Finally, rie in te diparity in oter income (Q) contributed to a coniderable extent (16%) to te rie in te income inequality. Q 4.3. Income Gap between te 2 nd and 9 t Decile Te difference in te log of income between te top and bottom 10 t oueold, ued above a te meaure of oueold income inequality, may not deliver te full picture of canging inequality. By focuing on te gap between te ricet and te pooret, in particular, it fail to capture any cange in te middle of te income ditribution. In order to upplement ti weakne, at leat partially, a imilar decompoition metod i applied to te difference in income between te econd and nint income decile. Te reult i reported in Table 5. <Table 5 Here> A. 1988-1993 Te decline in te difference in te log of income between te econd and nint decile (-0.0628) wa about alf in magnitude of te reduced income gap between te top and lowet 10 t oueold. Te reult of te decompoition are generally imilar 17
to toe for te difference between te firt and tent decile, reported in Table 2. Te cange in te wage of ead wa te dominant factor of reduced income diparity. Growing relative importance of poue earning wa te bigget countervailing factor of te trend toward a more equal ditribution of income. Tere are, owever, ome notable difference, too. Te abolute ize of te effect of cange in oter income (Q) wa relatively mall in ti cae (-13%), compared to te reult for te difference between te top and bottom decile (-78%). In addition, te equalizing effect of increaing weigt of oter income ( Φ ) wa maller (11% v. 31%). Ti reult, if Q combined wit te reult of te decompoition eparately performed for eac ide of te income ditribution (reported in Table 3 and 4), implie tat te relatively trong effect on inequality of te cange in Q i largely explained by te relative decline of Q among te lowet 10 t oueold. B. 1997-1999 Te rie in te difference in te log of income between te econd and nint decile for ti period (0.0869) wa only 27% of te increaed difference in income between te top and lowet 10 t oueold. Since te increae in inequality wa muc more pronounced among te oueold on te lower ide of te income ditribution, ti reult implie tat te relative decline of te lowet 10 t oueold wa te key factor of te growing income inequality during ti period. Compared to te cae of te difference between te top and bottom decile, canging wage of ead ( W ) explain muc larger fraction, more tan 100%, of te rie in inequality. Te relative contribution of canging our of ead ( H ) wa greater (23%). A oppoe to te reult for te difference between te top and bottom decile, te effect of cange in oter income (Q), 18
in term of it ditribution and relative are, wa negative. 13 5. Concluion Te dramatic rie in variou meaure of income in Korea after te financial crii i drawing increaing attention from economit. Altoug a few tudie ave examined te ource of canging income diparity, te role of labor upply a not been invetigated. Ti tudy a developed a new metod of decompoing te cange in oueold income into everal component, including te ift in employment and our worked of ead and poue of oueold, and etimated te relative contribution of eac of te component to te oberved cange in oueold income inequality between 1988 and 1999. Matced ample of Urban Houeold Income and Expenditure Survey and Economically Active Population Survey for te year 1988, 1993, 1997, and 1999 were ued in te analyi. Te difference in te log of income between te top and bottom income decile, te primary meaure of oueold income inequality ued in ti tudy, dropped by 0.13 between 1988 and 1993, remained table from 1993 to 1997, and ten increaed by 0.32 from 1993 to 1999. For te period 1988 to 1993, reduced inequality in te ourly wage of ead wa te ingle mot important caue of te improvement in te oueold income ditribution. In fact, ad tere been no cange in oter factor, te cange in te wage of ead would ave produced a decline in te meaure of inequality twice a large a te actual decline in magnitude. Te trong equalizing effect of te cange in 13 Similar decompoition analye were conducted, uing quintile, intead of decile, a te unit of dividing oueold according to income. Te reult for decompoing te cange in te diparity in income between te top and bottom quintile, a can be expected, are placed alfway between te reult for te top and bottom decile and for econd and nint decile. For intance, cange in wage and our worked of ead explain, repectively, 67% and 16% of te rie in te diparity in income between te top and bottom 20% oueold. 19
wage wa offet to a coniderable extent by ift in oter income and te are of earning of poue. For te period 1997 to 1999, on te oter and, cange in oter income, along wit cange in te wage of ead, wa a coniderably important contributor to te widening of te income diparity. Canging our of ead wa a nontrivial factor, accounting for 15% of te rie in income inequality. During te year 1988 to 1993, te overall decline in te meaure of income inequality come equally from bot tail of te income ditribution. Diminied wage diparity among oueold ead wa reponible for te rinking of te pread on bot ide of te income ditribution. In te year 1997 to 1999, on te oter and, te arp rie of income inequality wa largely due to te collape of low-income oueold. For te upper alf of te income ditribution, te rie in te diparity in oter income wa te major caue of te widening of income gap between te ricet 10 t oueold and toe in te middle. For te lower ide of te income ditribution, te cange in wage wa te mot important factor tat produced te increae in te diparity between te pooret 10 t and average oueold. If te difference in te log of income between oueold in te econd and nint decile i ued a te meaure of income inequality, cange in te wage of ead till tand out a te dominant factor of te ift in income inequality. Since te earning of ead account for te lion are of te total oueold income and only te oueold wit employed ead are included in te ample, it i not too urpriing to oberve tat cange in te wage of ead emerged a te mot important contributor to te ift in income inequality. Meanwile, it i not a fully predictable reult tat cange in oter income, in term of te are in te total oueold income and te tate of ditribution, ad a large effect on te rie in income 20
inequality between 1997 and 1999. It i alo notable to find tat te cange in te our worked account for a repectable fraction of te rie in te meaure of income inequality between 1997 and 1999. In interpreting te reult, te effect of canging labor upply call for a cloer attention. It i well documented tat te unemployment rate oared during te economic receion following te financial crii. 14 Job loe, rater tan lowered wage, ould ave delivered a more devatating impact on te economic wellbeing of lower-income oueold. Terefore, if te effect of canging employment of ead i examined, uing a ample covering te non-employed, te relative contribution of canging labor upply to te rie in income inequality after 1997 ould come out muc grater tan uggeted by te reult of ti tudy. Te reult of a imilar analyi conducted for te United State from 1969 to 1989 i uggetive wit regard to ti point. Te relative decline of labor upply among ead of lower-income oueold accounted for 45% of te rie in te difference in te log of income between te top and bottom income decile. Of ti 45%, canging employment explained 26%, and ift in our worked 19% (Lee 2000). A verification of ti conjecture will ave to wait for te releae of improved data tat will allow a more complete decompoition analyi. Reference An, K. (1997), Trend in and Determinant of Income Ditribution in Korea, Journal 14 Te number of te unemployed increaed from 452,000 in October 1997 to 1,378,000 by Marc 1998. Te average unemployment rate in 1998 wa 6.8%, 4.2% point greater tan in te previou year. 21
of Economic Development 22 (2). Burtle, G. (1990), Earning Inequality over te Buine and Demograpic Cycle, in Burtle, G.., ed., A Future of Louy Job? Waington DC: Brooking Intitution, 77-117. Burtle, G. (1993), Te Contribution of Employment and Hour Cange to Family Income Inequality, American Economic Review Paper and Proceeding 83, 131-135. Cung J., and K. Coi (2001), A Decompoition of Houeold Income Inequality of Worker (in Korean), Korean Economic Review 49. Cung, J., D. Hwang, B. Lee, and K. Coi (2002), Te State and Policy Iue of Income Inequality and Poverty (in Korean), Korea Labor Intitute Haveman, R. (1996), Earning Inequality: Te Influence of Canging Opportunitie and Coice. Waington DC: Te AEI Pre. Kang, S., and J. Hyun (2001), Suggetion for Improving Statitic for Undertanding Income Ditribution (in Korean), Paper preented at te KDI Conference on Income Ditribution. Korea National Statitical Office (2001a), Te Economically Active Population Survey Annual Report. Korea National Statitical Office (2001b), Annual Report on te Houeold Income and Expenditure Survey. Lee, C. (2000), "Te Relation of te Growt in Income Inequality to te Organization of Work and te Structure of Conumption." R. Fogel., Te Fort Great Awakening and te Future of Egalitarianim, Cicago: Univerity of Cicago Pre, 272-283. Lee, C. (2001), "Cange in Employment and Hour, and Family Income Inequality in te United State, 1969-1989." International Economic Journal 15, 27-49. Lee, J., and S. Hwang (1998), Decompoition of Wage Income Inequality and Analyi 22
of it Caue (in Korean), KDI Policy Reearc 20, (3/4) Lee, J., and C. Ree (1999), "Social Impact of te Aian Crii: Policy Callenge and Leon." IER Working Paper No. 2, Seoul National Univerity. Moffitt, R. (1990), Te Ditribution of Earning and te Welfare State, in Burtle, G.. ed., A Future of Louy Job? Waington DC: Brooking Intitution, 201-230. Yoo, K. (2001), Analyi of te Trend and Caue of te Income Cange after te Financial Crii, Paper preented at te KDI Conference on Income Ditribution. 23
Table 1 Definition of Variable Ued Variable N H H W W P P δ Q Φ Φ Φ Q Definition of Variable Average montly money income of oueold in a given income decile Mean montly our worked by employed ead of oueold Mean montly our worked by employed poue Mean ourly wage rate of employed ead of oueold Mean ourly wage rate of employed poue Employment rate for ead of oueold Employment rate for poue Te fraction of oueold in wic bot uband and wife are preent Te mean montly income from oter ource Te difference in te log of an income variable between two te top and te bottom income decile; e.g., TOP H = ln( H ) ln( H BOTTOM ) Te weigt of te income earned by te ead of oueold Te weigt of te income earned by poue Te weigt of te income from oter ource 24
Table 2 A Decompoition of te Cange in Houeold Income Inequality, 1988-1993 and 1997-1999: Difference in Income between te Top and Bottom Decile (1) (2) 1988-1993 1997-1999 Variable Etimate Contribution Etimate Contribution N -0.1299 100.00 0.3239 100.00 Φ 0.0096-7.38 0.0481 14.86 H (3) Φ -0.2523 194.27 0.1742 53.76 W (4) Φ ( H + W ) -0.0111 8.53-0.0273-8.44 (5) Φ 0.0094-7.20 0.0044 1.34 H (6) Φ -0.0200 15.40 0.0048 1.48 W (7) Φ 0.0337-25.92-0.0221-6.83 (8) P Φ δ -0.0004 0.30-0.0054-1.66 (9) Φ ( H + W + P + δ ) 0.0673-51.86 0.0126 3.89 (10) Φ q Q 0.1018-78.37 0.0920 28.41 (11) Φ qq -0.0399 30.73 0.0331 10.21 (12) ε -0.0279 21.48 0.0096 2.97 Labor Supply: (2)+(5)+(7) Wage: (3)+(6) Compoition: (4)+(9)+(11) 0.0526-0.2723 0.0164-40.49 209.67 28.16 0.0303 0.1790 0.0183 9.37 55.25 5.66 25
Table 3 A Decompoition of te Cange in Houeold Income Inequality, 1988-1993 and 1997-1999: Difference between te Income of te Top Decile and te Average Income (1) (2) 1988-1993 1997-1999 Variable Etimate Contribution Etimate Contribution N -0.0653 100.00 0.1277 100.00 Φ 0.0178-27.18 0.0045 3.49 H (3) Φ -0.1236 189.19 0.0457 35.75 W (4) Φ ( H + W ) -0.3622 55.44 0.0020 1.53 (5) Φ 0.0049-7.45-0.0023-1.82 H (6) Φ -0.0226 34.60 0.0076 5.95 W (7) Φ 0.0285-43.59 0.0147 11.49 (8) P Φ δ -0.0008 1.19-0.0012-0.98 (9) Φ ( H + W + P + δ ) 0.0339-51.84-0.0120-9.39 (10) Φ q Q 0.0121-18.47 0.0595 46.57 (11) Φ qq 0.0215-32.96 0.0101 7.92 (12) ε -0.0007 1.08-0.0007-0.52 Labor Supply: (2)+(5)+(7) Wage: (3)+(6) Compoition: (4)+(9)+(11) 0.0511-0.1462 0.0192-78.23 223.78-29.36 0.0168 0.0533 0.0001 13.16 41.71 0.07 26
Table 4 A Decompoition of te Cange in Houeold Income Inequality, 1988-1993 and 1997-1999: Difference between te Average Income and te Income of te Bottom Decile (1) (2) 1988-1993 1997-1999 Variable Etimate Contribution Etimate Contribution N -0.0646 100.00 0.1962 100.00 Φ -0.0102 15.81 0.0473 24.09 H (3) Φ -0.1257 194.77 0.1357 69.17 W (4) Φ ( H + W ) 0.0202-31.36-0.0006-0.33 (5) Φ 0.0049-7.63 0.0051 2.62 H (6) Φ -0.0033 5.07-0.0006-0.31 W (7) Φ 0.0112-17.34-0.0279-14.23 (8) P Φ δ 0.0001-0.21-0.0038-1.93 (9) Φ ( H + W + P + δ ) 0.0214-33.12 0.0047 2.41 (10) Φ q Q 0.0743-115.09 0.0309 15.73 (11) Φ qq -0.0488 75.61-0.0024-1.24 (12) ε -0.0087 13.51 0.0079 4.03 Labor Supply: (2)+(5)+(7) Wage: (3)+(6) Compoition: (4)+(9)+(11) 0.0059-0.1290-0.0072-9.17 199.83 11.13 0.0245 0.1351 0.0017 12.47 68.86 0.85 27
Table 5 A Decompoition of te Cange in Houeold Income Inequality, 1988-1993 and 1997-1999: Difference in Income between te Second and Nint Income Decile (1) (2) 1988-1993 1997-1999 Variable Etimate Contribution Etimate Contribution N -0.0628 100.0 0.0869 100.00 Φ 0.0197-31.38 0.0196 22.51 H (3) Φ -0.1266 201.61 0.1156 133.07 W (4) Φ ( H + W ) -0.0129 20.60 0.0164 18.89 (5) Φ 0.0038-6.07 0.0065 7.43 H (6) Φ -0.0016 2.59-0.0009-9.98 W (7) Φ 0.0293-46.68-0.0249-28.65 (8) P Φ δ -0.0022 3.57 0.0024 2.74 (9) Φ ( H + W + P + δ ) 0.0317-50.52-0.0195-22.44 (10) Φ q Q 0.0083-13.22-0.0222-25.56 (11) Φ qq -0.0069 10.93-0.0123-14.21 (12) ε -0.0054 8.57 0.0063 7.20 Labor Supply: (2)+(5)+(7) Wage: (3)+(6) Compoition: (4)+(9)+(11) 0.0528-0.1282 0.0173-84.13 204.21-19.00 0.0011 0.1148-0.0154 1.30 132.07-17.76 28
Appendix Table Component of te Average Houeold Income Year All Bottom 9t 8t 7t 6t 5t 4t 3d 2d Top 1. Total family income 1988 643711 199828 321280 387981 447269 509686 579697 665194 784131 973831 1567943 1993 1455383 481922 758407 913178 1052070 1193537 1347330 1538502 1788892 2158890 3320824 1997 2260730 713350 1131095 1394526 1627200 1861674 2115996 2425443 2825758 3410129 5102131 1999 2199180 570300 1000375 1266483 1491799 1724971 1993665 2310477 2704053 3289775 5639358 2. Montly earning for employed family ead 1988 484955 164578 271925 322303 362055 397690 449675 501079 580138 717769 1082163 1993 1062869 427626 651766 757210 842560 918285 1022264 1144049 1280166 1501248 2037584 1997 1531335 579650 930690 1090571 1224468 1341665 1495511 1677249 1871416 2214731 2887402 1999 1528737 452603 806966 1019197 1178426 1304179 1464000 1666408 1968804 2302505 3123948 3. Weekly our worked for employed family ead 1988 57.73 56.87 59.72 60.58 59.57 58.97 58.63 59.05 57.39 55.53 52.56 1993 53.77 53.65 55.04 54.52 54.57 54.80 54.82 53.89 53.74 52.48 50.18 1997 52.27 52.00 53.10 53.59 53.31 53.25 53.54 52.44 51.62 50.80 49.05 1999 52.31 48.84 52.65 53.44 54.07 53.44 53.56 52.81 52.61 51.71 49.44 4. Montly earning for employed poue 1988 228582 100374 122552 136233 145833 164709 181565 199072 228185 277626 553740 1993 554567 235260 270565 312512 347486 378005 415070 471905 533129 639758 1077549 1997 835017 343827 453557 512937 517868 595117 655552 743322 844004 969724 1549824 1999 862821 330475 418579 459261 508628 570259 627462 736180 869797 973540 1687255 5. Weekly our worked for employed poue 1988 52.71 57.06 54.70 53.11 53.91 51.95 53.16 52.96 56.87 52.27 46.51 1993 48.28 47.16 46.99 48.67 49.48 49.14 49.42 49.56 48.65 48.39 45.26 1997 47.51 48.63 49.37 47.40 47.26 48.40 48.91 47.87 46.89 46.78 46.28 1999 46.00 43.37 44.69 45.91 45.00 45.78 45.60 47.94 48.55 46.67 43.79 29
6. Proportion of uband-wife familie 1988 0.90 0.72 0.83 0.89 0.88 0.90 0.93 0.93 0.95 0.95 0.97 1993 0.89 0.71 0.84 0.87 0.89 0.90 0.91 0.92 0.93 0.92 0.95 1997 0.87 0.69 0.80 0.84 0.86 0.87 0.90 0.91 0.92 0.94 0.96 1999 0.89 0.75 0.78 0.87 0.87 0.87 0.92 0.94 0.96 0.95 0.97 7. Employment rate for poue 1988 0.19 0.07 0.11 0.16 0.15 0.19 0.23 0.24 0.25 0.18 0.20 1993 0.24 0.07 0.11 0.13 0.19 0.25 0.28 0.28 0.30 0.32 0.36 1997 0.30 0.08 0.11 0.18 0.25 0.32 0.35 0.39 0.39 0.40 0.40 1999 0.24 0.10 0.12 0.16 0.20 0.25 0.27 0.30 0.25 0.30 0.37 8. Head' earning weigt 1988 0.778 0.802 0.847 0.831 0.810 0.780 0.776 0.753 0.740 0.737 0.702 1993 0.761 0.811 0.853 0.824 0.800 0.767 0.758 0.743 0.715 0.694 0.640 1997 0.717 0.793 0.823 0.783 0.753 0.721 0.707 0.692 0.662 0.650 0.592 1999 0.743 0.763 0.807 0.805 0.790 0.757 0.735 0.721 0.729 0.699 0.621 9. Spoue' earning weigt 1988 0.056 0.039 0.040 0.048 0.051 0.062 0.066 0.067 0.065 0.049 0.070 1993 0.065 0.026 0.032 0.040 0.054 0.070 0.076 0.079 0.082 0.086 0.111 1997 0.080 0.031 0.036 0.054 0.067 0.087 0.095 0.106 0.103 0.104 0.114 1999 0.070 0.038 0.042 0.049 0.057 0.069 0.078 0.091 0.079 0.086 0.114 10. Oter income weigt 1988 0.167 0.159 0.113 0.121 0.139 0.157 0.158 0.180 0.196 0.214 0.228 1993 0.174 0.163 0.114 0.136 0.146 0.163 0.167 0.178 0.203 0.221 0.249 1997 0.203 0.176 0.141 0.163 0.180 0.192 0.198 0.202 0.234 0.245 0.294 1999 0.187 0.199 0.150 0.147 0.153 0.174 0.187 0.188 0.192 0.214 0.266 11. Family ize 1988 4.01 3.51 3.58 3.69 3.80 3.94 4.08 4.18 4.29 4.42 4.62 1993 3.82 3.24 3.45 3.60 3.71 3.81 3.88 3.98 4.08 4.16 4.28 1997 3.65 3.01 3.32 3.32 3.63 3.63 3.67 3.80 3.87 3.97 4.05 30
1999 3.61 3.09 3.24 3.44 3.57 3.60 3.71 3.76 3.88 3.90 3.95 31
Figure 1 Average Montly Family Income (Won) 6000000 5000000 4000000 3000000 2000000 1000000 0 Bottom 8t 6t 4t 2nd 1988 1993 1997 1999 20000 Figure 2 Average Hourly Wage for Employed Head 15000 10000 5000 1988 1993 1997 1999 0 Bottom 8t 6t 4t 2nd 32
Figure 3 Weekly Hour for Employed Head 80 60 40 20 1988 1993 1997 1999 0 Bottom 8t 6t 4t 2nd 0.5 Figure 4 Proportion of Employed Spoue 0.4 0.3 0.2 0.1 1988 1993 1997 1999 0.0 Bottom 8t 6t 4t 2nd 33
10000 Figure 5 Average Hourly Wage for Employed Spoue 8000 6000 4000 2000 1988 1993 1997 1999 0 Bottom 8t 6t 4t 2nd Figure 6 Weekly Hour for Employed Spoue 60 50 40 30 20 10 0 Bottom 8t 6t 4t 2nd 1988 1993 1997 1999 34