Winners and Losers from Trade Reform: Case Studies for China and Morocco. Martin Ravallion

Transcription

1 Wnners and Losers from Trade Reform: Case Studes for Chna and Morocco Martn Ravallon 1

2 Does trade ncrease nequalty? Some argue yes some no Conflctng evdence on mpacts of trade openness on nequalty wthn countres Dollar and Kraay: Trade has no effect on nequalty Lundberg and Squre; Mlanovc: Trade s nequalty ncreasng (more so n poor countres) A cloud of doubt on emprcs Methodologcal problems; e.g., dentfcaton wth lags Nose n nequalty and poverty data over tme Few attempts to replcate/reconcle 2

3 Does trade promote growth? Also much debate More sgns of postve mpacts Manstream vew: trade expanson s dstrbuton neutral but growth promotng; hence poverty reducng 2/3 of Sala--Martn et al., x-country regressons suggest that trade openness promotes growth But trade volume s not n ther robust 18 predctors. Issues n debate Specfcaton of control varables/error structure Esp., concerns about endogenety of trade and dentfcaton strateges Implcatons for trade polcy are uncertan 3

4 A debate between two shps passng n the nght? One sde of the globalzaton debate ponts to aggregate dstrbuton-neutralty; on average nequalty nether rses or falls wth growth or openness. The other sde ponts to the losers amongst the poor or vulnerable Maybe both are rght. 4

5 Evaluaton ssues: 1. Economy-wde reforms 2. Horzontal mpacts of reform 5

6 1. Welfare mpacts of an economy-wde reform There can be no control group for an economy-wde reform. We have to use a structural model of the economy to nfer the counterfactual of what t would be lke wthout the reform. Or (for ex ante reforms) we use the model to smulate the mpact of the reforms usng the status quo as the counterfactual.

7 2. Horzontal mpacts of reform Dfferences n mpact amongst people who are ex ante equal n terms of ncome Stemmng from heterogenety n net tradng postons n relevant markets Wnners and losers, at all level of lvng 7

8 Precedents n economcs Pgou (1949): horzontal nequalty created..a sense of beng unfarly treated n tself an evl. Auerbach and Hassett (2002):..large dfferences [n tax rates] amongst smlar ndvduals, regardless of ther source, mght be vewed as ntrnscally arbtrary, and therefore costly to the socal fabrc. Surely the same arguments hold for other areas of publc polcy, ncludng trade reform. 8

9 Relevance to country polcy? Horzontal mpacts reflect dverse net tranng postons n relevant markets. These need not be weghted correctly by conventonal nequalty aggregates. Old poor vs. new poor? Horzontal mpacts => churnng. Potental mplcatons for socal protecton.

10 Two examples: Trade reforms n Chna and Morocco Chna: mpacts of WTO accesson (both ex ante and ex post). Morocco: ex ante assessments of mpacts of de-protectng cereals. 10

11 Methods of mpact assessment for trade reforms Natonal surveys used to measure the welfare mpacts of prce and wage changes attrbuted to trade reform. Prce changes estmated usng a CGE. Sequental not ntegrated. Welfare mpacts are frst-order approxmatons of a money metrc of utlty, based on a household model ncorporatng ownproducton actvtes, calbrated to the household-level data mposng mnmum aggregaton. 11

12 12 Indrect utlty functon of household : ] ), ( max[ ],, [ ), ( d d d L q d L w q p L q u w p v d π +π = = where )], ( max [ ),, ( ), ( o s o d s s L z d s L z f q w L z p q p w p p o = = π q d, q s : quanttes consumed and suppled z: quanttes used as producton nputs p d, p s : prces (demand and supply) w: wage rate vector; L: labor supply vector Standard model of agrcultural household

13 Money metrc of welfare mpact Money metrc of change n utlty nduced by changes n prces and labor earnngs: g du v π = m j= 1 [ p s j q s j dp p s j s j p d j ( q d j + z j ) dp p d j d j ] + n k= 1 ( w k L s k dw w k k ) Note: Ths assumes small prces changes around optmum (envelope theorem) no ratonng n goods or labor markets 13

14 Modelng the gans at household level Money metrc of change n utlty can be modeled as a functon of household characterstcs, ncludng locaton: g = x + β x k β γ D + ε k k For example: Demographcs/lfe-cycle => consumpton patterns => welfare effects of trade reform Locaton => prces, wages + producton and consumpton behavor => welfare effects of trade reform 14

15 15 Decomposton of the mpact of trade reform on nequalty Condtonal mean mpact: ) ( c y y g E g = = The overall change n nequalty (MLD) can then be wrtten: = = = n c n c y g y g n y g y g n I I 1 1 * / 1 / 1 ln 1 / 1 / 1 ln 1 vertcal component + horzontal component Vertcal component: depends on how mean mpacts vary wth prereform ncome. Horzontal component: depends on the devatons n mpacts from ther condtonal means.

16 Case Study 1: Chna s accesson to the WTO Offcal date of Chna s WTO accesson s 2001 However, the Chnese economy had already started to adapt to ths expected change. Stage 1: a lead-up perod n whch tarffs started to fall n antcpaton of WTO accesson Stage 2: the perod 2001 onwards s a plausble begnnng of the lead-up perod to WTO accesson.

17 The general equlbrum model We use the GTAP model for Chna developed by Ianchovchna and Martn. Flexble prces and wages (market-clearng). Imports and domestc goods are mperfect substtutes. Estmates of the prce changes due to WTO for and The counter-factual s not jonng WTO.

18 Lnkng the CGE model to the household survey Prce changes carred to large household survey. Prce changes are passed on fully weak nternal market ntegraton leads one to queston ths assumpton.

19 Two smulatons 1. Subtractng the estmated gans over from the 1999 ncomes at household level Ths tells us the mpacts of the prce changes durng the frst stage of the trade reform.e., what the baselne dstrbuton would have looked lke wthout the reforms 2. Addng the household-specfc gans from to the 1999 ncomes Ths tells us the mpact of the post-2001 prce changes.e., how those changes are expected to mpact on the baselne dstrbuton, lookng forward.

20 Aggregate mpacts Mean gans (Yuan/capta) Inequalty (Gn ndex as %) 1999 Mnus mpacts Plus mpacts Rural Urban Total (1.5%)* (-.04%)* Note: * gves % of mean ncome

21 Impact on rural poverty Fgure 1a: Poverty ncdence curves: rural 60.00% % people below the poverty lne 50.00% 40.00% 30.00% 20.00% 10.00% Post WTO (mddle) Pre WTO Baselne 0.00% Annual per capta ncome (Yuan)

22 Impact on urban poverty Fgure 1b: Poverty ncdence curves: urban 40.00% 35.00% % people below the poverty lne 30.00% 25.00% 20.00% 15.00% 10.00% 5.00% Pre WTO ncome Baselne dstrbuton n 1999 (mddle) Post WTO 0.00% Annual per capta ncome (Yuan)

23 Horzontal mpacts of WTO n Chna Large and partly explcable varance n mpacts across household characterstcs at gven ncome. Rural famles lose; urban households gan. Larger mpacts n some regons than others. E.g., non-neglgble welfare losses among agrcultural households n the northeast, where households are more dependent on feed gran producton. Vertcal dfferences n pre-nterventon ncomes accounted for vrtually none of the measured welfare mpact of ths trade reform. 23

24 Regressons for level (Yuan) of gan n rural areas Laonng Guangdong Schuan Coeffcent t-rato Coeffcent t-rato Coeffcent t-rato Log of household sze Age of household head Squared age Agrculture household # of employee/h'hold sze # of TVE workers/hh sze # of mgrate workers/hh sze Area of cultvated land Area of hlly land Area of fshpond land Hghest educaton level s llterate or sem-llterate prmary school mddle school hgh school techncal school college (default) Rato of labor force Rato of chldren under Rato of chldren age Rato of chldren age Rato of chldren age Constant R-square

25 Welfare mpacts of WTO n Chna A postve mpact on mean household ncome (pror to formal accesson). Vrtually no change n aggregate nequalty and slghtly lower aggregate poverty n the short term. But a szable and partly explcable varance n mpacts across household characterstcs at gven ncome. Specfc sub-groups of losers amongst the poor. 25

26 Case Study 2: Trade reform n Morocco Proposed de-protecton through substantal cuts n tarffs on mported cereals. Government of Morocco/World Bank CGE analyss of mpacts of ths trade reform. But nobody bothered to look at the household survey! 26

27 The GE analyss of GOM/WB Prce changes mpled by trade reform generated by a CGE model that was commssoned by a jont workng group of the Mnstry of Agrculture and the World Bank (Doukkal, 2003). The model was explctly desgned to assess the aggregate mpacts of de-protectng cereals n Morocco. 27

28 Interactons between agrculture and the rest of the economy represented by sx sectors Qute detaled representaton of the agrcultural sector. 16 dfferent crops or groups of crops, three dfferent lvestock actvtes, 13 major agro-ndustral actvtes, sx agro-ecologcal regons, wthn each regon the model dstngushes between ranfed agrculture and four types of rrgated agrculture. Two types of labor, both wth fxed real wage rates. 28

29 Polcy smulatons Smulatons dffer n the extent of the tarff reductons for cereals, namely 10%-100%. In all cases, the government s exstng open-market operatons (whch attempt to keep down consumer prces by sellng subsdzed cereals) are also removed. The loss of revenue from a 50% tarff cut approxmately equals the savng on subsdes. 29

30 Aggregate mpacts Baselne Polcy 1 (10%) Polcy 2 (30%) Polcy 3 (50%) Polcy 4 (100%) Natonal Poverty rate (%) Mean Log Devaton (x100) Gn ndex Per capta gan Mean % gan: weghted by mean shares Producton gan Consumpton gan Consumpton per capta

31 Impacts on poverty.6 Total Baselne Polcy 1 Polcy 4 % people below the poverty lne Annual per capta consumpton Urban.6.6 Rural % people below the poverty lne % people below the poverty lne Annual per capta consumpton Annual per capta consumpton 31

32 Polcy 1 Polcy 1 Polcy 4 Polcy absolute gan/loss per capta % gan/loss per capta Per capta consumpton percentles Per capta consumpton percentles 32

33 Decomposton for agrcultural trade reform n Morocco Impact on nequalty (MLD) Partal deprotecton (30%) => Full deprotecton (100%) => Vertcal component 58% -20% Horzontal component 42% 120% 33

34 Wnners and losers Small aggregate mpacts on mean consumpton and nequalty. But large horzontal mpacts; partly measurement errors but also systematc covarates Rural poor (and nonpoor) lose on average Majorty of rural poor are net consumers But rural poor net producers n the aggregate Urban poor and nonpoor gan Large losses n specfc regons 34

35 It s lttle consolaton to the losers to be told that nequalty has not changed wth trade reform Gans as % of consumpton from full deprotecton of cereals n Morocco Selected regon (5/16) Poorest 15% of rural households Chaoua-Ouardgha Meknes Tafl Fes-Boulemane Taza-Al Hocema- Taounate Tanger-Tetouan

36 Implcatons for socal protecton? In both Chna and Morocco, the horzontal mpacts are partly explcable n terms of observable demographc, economc and geographc characterstcs. The observable heterogenety of mpacts holds mplcatons for socal protecton polces, to help compensate poor losers from polces that are propoor overall. Unexplaned varance s suggestve of dosyncratc welfare mpacts that wll need a self-targeted safety net 36