Agricultural Distortions in Sub-Saharan Africa: Trade and Welfare Indicators, 1961 to 2004

Transcription

1 Dscusson Paper No Agrcultural Dstortons n Sub-Saharan Afrca: Trade and Welfare Indcators, 1961 to 2004 Johanna Croser and Kym Anderson March 2010 Unversty of Adelade Adelade 5005 Australa

2 CENTRE FOR INTERNATIONAL ECONOMIC STUDIES The Centre was establshed n 1989 by the Economcs Department of the Adelade Unversty to strengthen teachng and research n the feld of nternatonal economcs and closely related dscplnes. Its specfc objectves are: to promote ndvdual and group research by scholars wthn and outsde the Adelade Unversty to strengthen undergraduate and post-graduate educaton n ths feld to provde shorter tranng programs n Australa and elsewhere to conduct semnars, workshops and conferences for academcs and for the wder communty to publsh and promote research results to provde specalsed consultng servces to mprove publc understandng of nternatonal economc ssues, especally among polcy makers and shapers Both theoretcal and emprcal, polcy-orented studes are emphassed, wth a partcular focus on developments wthn, or of relevance to, the Asa-Pacfc regon. The Centre s Actng Drector s Dr Ernesto Valenzuela (ernesto.valenzuela@adelade.edu.au). Further detals are avalable from: CIES School of Economcs Adelade Unversty SA 5005 AUSTRALIA Telephone: (+61 8) Facsmle: (+61 8) Emal: ces@adelade.edu.au Most publcatons can be downloaded from our Home page: ISSN seres, electronc publcaton

3 Agrcultural Dstortons n Sub-Saharan Afrca: Trade and Welfare Indcators, 1961 to 2004* Johanna Croser Unversty of Adelade johlou@gmal.com and Kym Anderson Unversty of Adelade kym.anderson@adelade.edu.au March 2010 *Ths s a product of a World Bank research project on Dstortons to Agrcultural Incentves (see The authors are grateful for the dstorton estmates provded by the authors of the varous Afrcan country case studes, for fundng from World Bank Trust Funds provded by the governments of the Netherlands (BNPP), the Unted Kngdom (DfID) and Ireland, as well as from the Australan Research Councl. The vews expressed are the authors alone and not necessarly those of the World Bank and ts Executve Drectors, nor the countres they represent, nor of the nsttutons provdng the project research funds.

4 Agrcultural Dstortons n Sub-Saharan Afrca: Trade and Welfare Indcators, 1961 to 2004 Abstract For decades, agrcultural prce and trade polces n Sub-Saharan Afrca hampered farmers contrbutons to economc growth and poverty reducton. Whle there has been much polcy reform over the past two decades, the njectons of agrcultural development fundng, together wth on-gong regonal and global trade negotatons, have brought dstortonary polces under the spotlght once agan. A key queston asked of those polces s: how much are they stll reducng natonal economc welfare and trade? Economy-wde models are able to address that queston, but they are not avalable for many poor countres. Even where they are, typcally they apply to just one partcular prevous year and so are unable to provde trends n effects over tme. Ths paper provdes a partal-equlbrum alternatve to economy-wde modellng, by drawng on a modfcaton of so-called trade restrctveness ndexes to provde theoretcally precse ndcators of the trade and welfare effects of agrcultural polcy dstortons to producer and consumer prces over the past half-century. We generate tme seres of country level ndces, as well as Afrca-wde aggregates. We also provde annual commodty market ndces for the regon, and we provde a sense of the relatve mportance of the key polcy nstruments used. Keywords: dstorted ncentves; agrcultural prce and trade polces; trade restrctveness ndex JEL classfcatons: F13, F14, F15, N57, Q17, Q18 Author contact: Kym Anderson School of Economcs Unversty of Adelade Adelade SA 5005 Australa Phone Fax kym.anderson@adelade.edu.au

5 Agrcultural Dstortons n Sub-Saharan Afrca: Trade and Welfare Indcators, 1961 to 2004 In the 1960s and 1970s, governments of many Sub-Saharan Afrcan countres adopted macroeconomc, sectoral, trade and exchange rate polces that drectly or ndrectly taxed farm households seekng to export ther way out of poverty. Ths antagrcultural, ant-trade, welfare-reducng polcy stance, whch was also prevalent n numerous other developng country regons up to the early 1980s (Krueger, Schff and Valdes 1988), has snce begun to be reformed. How far has that reform effort gone n alterng the trade- and welfare-reducng characterstcs of farm and food polces n Sub-Saharan Afrca? Ths matters greatly for economc development and poverty allevaton, because 60 percent of Sub-Saharan Afrca s workforce s stll employed n agrculture, nearly 40 percent of the populaton s earnng less than $1/day, and more than 80 percent of the regon s poorest households depend drectly or ndrectly on farmng for ther lvelhoods (World Bank 2007, Chen and Ravallon 2008). There are mportant questons to be addressed about future agrcultural polcy reform n Afrca because Afrcan agrculture s currently the subject of several new agrcultural development assstance programs, as well as beng mportant n on-gong multlateral and preferental trade negotatons. A frst step n consderng possble future polces s to examne the mpacts of past polcy choces, and n partcular to ask by how much are the polces stll reducng natonal economc welfare and trade? Economy-wde models are able to address that queston, but such models are not avalable for many of Afrca s poorer countres. Even where they are, typcally they depend on myrad assumptons about parameter (for lack of econometrc estmates) and they apply to just one partcular prevous year and so are unable to provde trends n effects over tme. Ths paper provdes a partal-equlbrum alternatve to economy-wde modellng, by drawng on a modfcaton of so-called trade restrctveness ndexes to provde theoretcally precse ndcators of the trade and welfare effects of agrcultural

6 polcy dstortons to producer and consumer prces. By drawng on a recent comprehensve database coverng most of Sub-Saharan Afrcan agrculture, we generate annual country level ndces for the past half-century, as well as regon-wde aggregates ncludng for ndvdual commodtes and a sense of the relatve mportance of the key polcy nstruments used. In dong so we make a methodologcal advance by ncorporatng a number of key nontradable products n our estmates of the ndces, whch turns out to be mportant n the Afrcan agrcultural polcy context. Data for constructon of the ndces come from the World Bank s Dstortons to Agrcultural Incentves database (Anderson and Valenzuela 2008). The database gves consstent measures of prce-dstortng polces for 75 countres for the perod 1955 to The data for the 21 Afrcan countres n that database s dscussed comprehensvely n Anderson and Masters (2009). In ths paper we focus on 19 of those Afrcan countres, leavng asde Egypt and South Afrca because they are both large and very dfferent from the others. That sample comprses fve countres of eastern Afrca (Ethopa, Kenya, Sudan, Tanzana, and Uganda), four n southern Afrca (Madagascar, Mozambque, Zamba, and Zmbabwe), fve large economes n Afrca s western coast (Cameroon, Côte d Ivore, Ghana, Ngera, and Senegal), and fve smaller economes of West and Central Afrca for whch cotton s a crucal export (Benn, Burkna Faso, Chad, Mal, and Togo). We concentrate on the perod 1961 to 2004, snce those are the years for whch the Afrcan data are most complete. Ths paper s structured as follows: the next secton presents the methodology we use. Ths s followed by a dscusson of the data n the World Bank s Agrcultural Dstortons database. We then report our estmates of the seres of ndces before presentng our conclusons and lstng some caveats and areas for further research. Methodology There s a growng lterature that dentfes ways to measure the trade- and welfarereducng effects of nternatonal trade-related polces n scalar ndex numbers. Ths lterature serves a key purpose: t overcomes aggregaton problems (across dfferent nterventon measures and across ndustres) by usng a theoretcally sound aggregaton procedure to answer precse questons regardng the trade or welfare reductons mposed by each country s trade polces. 2

7 These measures represent a substantal mprovement on commonly used measures. The usual tools for summarzng prce-dstortng polcy trends n a country or regon (see, e.g., Anderson and Masters 2009) are measures of the unweghted or weghted mean nomnal rate of assstance (NRA) and consumer tax equvalent (CTE), the standard devaton of NRAs, and n a few nstances the weghted mean NRA for exportable versus mport-competng covered products. 1 Authors often need to report more than one measure to gan an apprecaton of the nature of the polcy regme. For example, ndcators of dsperson of NRAs gve some dea of the addtonal welfare losses that come from greater varaton of NRAs across ndustres wthn the sector (Lloyd 1974). Further, f mport-competng and exportable sub-sectors have NRAs of opposte sgn, they need to be reported separately because those polces would offset each other n calculatng the aggregate sectoral NRA. Whle those varous ndcators are useful as a set, t s often helpful to have a sngle ndcator to capture the overall trade or welfare effect of an ndvdual country s regme of agrcultural prce dstortons n place at any tme, and to trace ts path over tme and make cross-country comparsons. To that end, the scalar ndex lterature s very useful. The poneerng theoretcal work s by Anderson and Neary (summarzed n ther 2005 book), wth an mportant partal equlbrum contrbuton by Feenstra (1995). The theory defnes an ad valorem trade tax rate whch, f appled unformly across all tradable agrcultural commodtes n a country wll generate the same reducton n trade, or n welfare, as the actual cross-product structure of dstortons. 2 In recent years, several emprcal papers have provded seres of estmates of scalar ndex numbers for ndvdual countres. Irwn (2008) uses detaled tarff data to calculate the Anderson Neary Trade Restrctveness Index for the Unted States n 1859 and annually from 1867 to Kee, Ncta and Olarreaga (2009) estmate partal-equlbrum ndexes for 78 developng and developed countres for a sngle pont n tme (md-2000s). Lloyd, Croser and Anderson (2010) estmate ndexes for 1 The OECD (2009) measures smlar ndcators to the NRA and CTE, called producer and consumer support estmates (PSEs and CSEs). The man dfference, apart from the CSE havng the opposte sgn to the CTE, s that the NRA and CTE are expressed as a percentage dvergence from undstorted (e.g., border) prces whereas the PSEs/CSEs relate to the dvergence from actual (dstorted) prces. 2 Other ndces defne an ad valorem trade tax rate whch, f appled unformly across all tradable products, wll generate the same government revenue (Bach and Martn 2001), or the same real natonal ncome and general equlbrum structure of the economy (Anderson 2009a), as the actual cross-product structure of dstortons. 3

8 75 developed and developng countres n the World Bank s recently released Dstortons to Agrcultural Incentves database (Anderson and Valenzuela 2008) over the perod 1955 to In addton to beng useful to summarze the agrcultural and food polcy regme n an ndvdual country, the Anderson-Neary scalar ndex measures can be usefully adapted to summarze two other aspects of agrcultural polcy: they can be computed for ndvdual polcy nstruments, to show the relatve contrbutons of dfferent polcy nstruments to reductons n trade and welfare (Croser and Anderson 2010); and they can be computed to measure the trade- and welfare-reducng effects of polcy n a sngle global or regonal commodty market (Croser, Lloyd and Anderson 2009). In ths paper we utlse the methodology to estmate all three types of ndexes. In dong so, we extend the theory and analyss to nclude nontradables, whch have not been addressed n prevous studes. Country level trade- and welfare-reducton ndexes To capture dstortons mposed by each country s border and domestc polces on ts economc welfare and ts trade volume, we adopt the methodology from Lloyd, Croser and Anderson (2010). Those authors defne a Welfare Reducton Index (WRI) and a Trade Reducton Index (TRI) and estmate them by consderng separately the dstortons to the producer and consumer sdes of the agrcultural sector (whch can dffer when there are domestc measures n place n addton to or nstead of trade measures). As ther names suggest, the two ndexes respectvely capture n a sngle ndcator the (partal equlbrum) welfare- or trade-reducng effects of all dstortons to consumer and producer prces of farm products from all agrcultural and food polcy measures n place. The WRI and TRI thus go somewhat closer to what a computable general equlbrum (CGE) can provde n the way of estmates of the trade and welfare (and other) effects of prce dstortons, whle havng the advantage of provdng an annual tme seres. Fortutously, estmates of the actual prce dstortons are avalable n the NRAs and CTEs of the World Bank s Dstortons to Agrcultural Incentves database. The dervaton of the two ndexes for n mport-competng ndustres leads to the expressons for the TRI and WRI for the mport-competng sector of a country shown n Box 1. 4

9 Box 1: Expressons for the TRI and WRI TRI T = { Ra + Sb}, wth WRI W R a S b 2 2 1/2 = { + }, wth n n R = ru and S = sv = 1 = 1 where R n 1 2 = [ r ] 2 u and = 1 *2 C *2 C * * u = p dx / dp / p dx / dp = ρ ( p x ) / ρ ( p x ) *2 P *2 P * * v = p dy / dp / p dy / dp = σ ( p y ) / σ ( p y ), S = n 1 2 [ s ] 2 v = 1 a= p dx /d / p d m /dp, and *2 C *2 p b= p dy /d / p d m /dp. *2 P *2 p Varable defntons: T Trade Reducton Index; W Welfare Reducton Index; R ndex of average consumer prce dstortons; S ndex of average producer prce dstortons; R Consumer Dstorton Index; S Producer Dstorton Index; s the rate of dstorton of the producer prce n proportonal terms; r rate of dstorton of the consumer prce n proportonal terms; u weght for each commodty n R and R, whch s proportonal to the margnal response of domestc consumpton to changes n nternatonal free-trade prces and can be wrtten as a functon of the domestc prce elastcty (at the protected trade stuaton) of demand ( ρ ); v weght for each commodty n S and S, whch s proportonal to the margnal response of domestc producton to changes n nternatonal free-trade prces and can be wrtten as a functon of the domestc prce elastcty (at the protected trade stuaton) of supply, ( σ ); p * P border prce; p = p * (1 + s ) dstorted domestc prce; C p = p * (1 + r ) C dstorted domestc consumer prce; x = x( p ) quantty of good demanded (as a functon of own P domestc prce); y = y ( p ) quantty of good suppled (as a functon of own domestc prce); a (b) weght of consumpton (producton) n the WRI or TRI, whch s proportonal to the rato of the margnal response of domestc demand (supply) to a prce change relatve to the margnal response of mports to a prce change. Source: Lloyd, Croser and Anderson (2010). Essentally the mport-competng TRI and WRI are constructed from approprately weghted averages of the level of dstortons of consumer and producer prces. The TRI s a mean of order one, and the WRI a mean of order two, but they use the same weghts. Because the WRI s a mean of order two, t better reflects the welfare cost of agrcultural prce-dstortng polces because t recognzes that the welfare cost of a government-mposed prce dstorton s related to the square of the 5

10 prce wedge. It thus captures the dsproportonately hgher welfare costs of peak levels of assstance or taxaton, and s postve regardless of whether the government s agrcultural polcy s favourng or hurtng farmers. The TRI and WRI can each be extended so as to add the exportable and nontradable sub-sectors of agrculture (see Appendx). Dstortons to exportable ndustres are nputted nto the TRI as negatve values because a postve (negatve) prce dstorton n an exportng ndustry has a trade-expandng (-reducng) effect, and thus decreases (ncreases) the TRI. Dstortons to nontradable ndustres are nputted nto the TRI as zero values because a domestc prce dstorton n a nontradable ndustry by defnton has nether a trade-expandng nor trade-reducng effect because of assumed prohbtvely hgh trade costs. Ths extenson of the TRI and WRI to nclude nontradables s a methodologcal contrbuton of ths paper, 3 and s of practcal sgnfcance n the case of Sub-Saharan Afrca where nontradables account for a non-trval share of the gross value of agrcultural producton (dscussed below). The expressons for the TRI and WRI weghts above show that estmates of prce elastctes are requred to compute the ndexes. In lne wth Lloyd, Croser and Anderson (2010), n the absence of elastctes we adopt some smplfyng assumptons n ths paper. We assume that domestc prce elastctes of supply (demand) are equal across commodtes wthn a sub-sector. Ths powerful smplfyng assumpton allows us (n the emprcal secton below) to fnd approprately weghted aggregates of dstortons on the producton and consumpton sdes smply by aggregatng the change n consumer (producer) prces across commodtes and usng as weghts the sectoral share of each commodty s domestc value of consumpton (producton) at undstorted prces. We expect ths smplfyng elastcty assumpton to have a very small mpact on the reported ndces. Ths s because elastctes appear n both the numerator and denomnator of the weght expressons, and therefore cancel each other out to some extent. Further, Kee, Ncta and Olarreaga (2009) show that the TRI and WRI can be decomposed nto three components and the elastcty only enters nto one of the three components, whch n practce s a very small component relatve to the other two. Ths transparent assumpton also makes sense n the context of computng tme seres of ndces for 3 Anderson and Neary (2005), chapter 12 dscusses the possblty of extendng ndces to nontradable sub-sectors and ncludng domestc dstortons n ther general equlbrum framework. 6

11 Afrca, where there s a dearth of relable and consstent elastcty estmates across tme for all our focus countres and ther covered agrcultural products. Polcy nstrument trade and welfare reducton ndexes The above country-level TRI and WRI measures are the aggregate of the trade- and welfare-reducng effects of all the polcy measures n place. The varables s and r, as domestc-to-border prce ratos, can theoretcally encompass dstortons provded by all trade tax/subsdy and trade non-tax/subsdy measures, plus domestc prce support measures (postve or negatve), plus drect nterventons affectng farm nput prces. Furthermore, where multple exchange rates operate, the measures can encompass an estmate of the mport or export tax equvalent of that dstortonary regme too. Whlst t s desrable to have such an aggregated country level ndcator that s so encompassng, agrcultural polcy analysts are sometmes nterested n the relatve contrbuton of dfferent polcy nstruments to reductons n trade and welfare. To provde ths nsght, t s possble to use the Anderson-Neary framework to construct ndcators of polcy dstortons at the nstrument level to facltate ths comparson. 4 To capture dstortons mposed by each Afrcan country s dfferent polcy nstruments on ts economc welfare and ts trade volume, we adopt the methodology from Croser and Anderson (2010). These authors defne an Instrument Welfare Reducton Index (IWRI) and an Instrument Trade Reducton Index (ITRI), whch can be estmated by consderng the dstorton from a sngle polcy nstrument to the producer and consumer sdes of the economy. The methodology n Croser and Anderson (2010) dentfes four types of border dstortons (mport taxes and subsdes, and export taxes and subsdes), for whch ndvdual ITRI and IWRI seres can be estmated. In addton to the border measures, the seres for domestc dstortons n the form of producton, consumpton and nput taxes and subsdes can be estmated. To estmate the trade-reducng effect of an ndvdual nstrument, those authors derve expressons for the change n mport volume from the ndvdual polcy measures, whch are used as the bass for dervng 4 We note that most of the seres of TRI and WRI ndcators n the lterature are for sngle nstruments anyway. For example, Irwn (2008) uses only mport tarffs; and Kee, Ncta and Olarreaga (2009) report two seres of ndces one based on tarffs only; and the other on tarffs plus NTBs. However, we are unaware of other studes that use the Anderson-Neary framework to drectly compare the dstortonary effects of dfferent nstruments on trade and welfare. 7

12 ITRIs. To estmate the welfare-reducng effect of ndvdual nstruments, the authors make an assumpton about the allocaton of the total welfare loss from the combnaton of ndvdual polcy nstruments. The authors assume that border measures are appled frst, and that ths may be supplemented by addtonal domestc dstortons. Thus the domestc dstorton s welfare reducton s the resdual from subtractng the border measures effects from the total welfare reducton of all polcy measures. Ths allocaton assumpton provdes a lower-bound on welfare losses from border measures and an upper-bound on welfare losses from domestc measures. The dervaton of the ITRI and IWRI follows essentally the same steps as those for the country-level ndces whch encompass all forms of dstorton. The dfference n the algebrac methodology s to specfy separate ndces for the nne dfferent types of polcy nstrument. Smplfyng prce elastcty assumptons can be made n the absence of relable estmates, and agan these assumptons have a mnmal effect on the estmates. Commodty market trade and welfare reducton ndces In addton to constructng country-level and nstrument-specfc ndces, ths paper makes use of another methodology wthn the Anderson-Neary framework to analyse a dfferent aspect of agrcultural polcy n Afrca s poorest natons. We construct ndces that show the extent to whch Afrcan markets for ndvdual farm commodtes are dstorted relatve to others. We employ the methodology n Croser, Lloyd and Anderson (2010) for ths purpose. Ths methodology s novel because whereas all prevous work wthn the trade restrctveness ndces lterature has focused on constructng ndex numbers of dstortons from the perspectve of a sngle country, ths methodology nstead takes a regonal vew of ndvdual commodty markets. The commodty TRI (WRI) s equal to the unform trade tax that has the same effect on regonal trade volume (welfare) as the exstng set of dstortons n the regon s natonal commodty markets. The measures are constructed n the same way as those for ndvdual country ndces, except that nstead of summng across dstortons n dfferent ndustres for a sngle country, the measures are constructed by summng across dstortons n dfferent countres for a sngle commodty. 8

13 The ndces are computed usng data on the domestc producton and consumpton sdes of the regon s natonal commodty markets, and the measures account for all forms of border and domestc prce dstorton n each country for the commodty market of nterest, as well as ncorporatng mport-competng and exportable countres nto the measure. In the absence of elastcty estmates, we make smplfyng assumptons analogous to those made for natonal ndexes. Croser, Lloyd and Anderson (2010) demonstrate that these assumptons have a mnmal mpact on the estmated seres when constructng ndces for global markets. Dstortons to Agrcultural Incentves database Ths study makes use of the World Bank s Dstortons to Agrcultural Incentves database (Anderson and Valenzuela 2008). The database came out of a global research project seekng to mprove the understandng of agrcultural polcy nterventons and reforms n Asa, Europe s transton economes, Latn Amerca and the Carbbean as well as Afrca. The database contans annual estmates of nomnal rates of assstance (NRA) (postve or negatve) for key farm products n 75 countres that together account for between 90 and 96 percent of the world s populaton, farmers, agrcultural GDP, and total GDP. There are 21 Afrcan countres n the database. We concentrate on the sample of 19 Sub-Saharan Afrcan countres lsted n the ntroducton, but exclude relatvely affluent Egypt and South Afrca whch together account for between one-thrd and one-ffth of Afrca gross value of producton at undstorted prces over the perod under analyss. For the 19 Afrcan focus countres, the database contans around 6000 consstent estmates of annual NRAs to the agrcultural sector and the same number of CTEs between 1955 and Country coverage up to 1960 s much less than from 1961, so the seres of estmates presented n ths paper begns n that latter year. The estmates of NRA and CTE n the database are at the commodty level and cover a subset of 41 agrcultural products n Afrca. These so-called covered products account for around 70 percent of total agrcultural producton over the perod studed. The data dentfes each year the extent to whch each commodty n each country s consdered an mportable, exportable or nontradable, a status that 9

14 may change over tme. In the 19 Afrcan focus countres, nontradable products account over tme for between 40 and 55 percent of the gross value of producton of all covered agrcultural products (last column of Table 1). The range of polcy measures ncluded n the NRA estmates n the Dstortons to Agrcultural Incentves database s wde. By calculatng domestc-toborder prce ratos, the estmates nclude assstance provded by all tarff and nontarff trade measures, plus any domestc prce support measures (postve or negatve), plus an adjustment for the output-prce equvalent of drect nterventons on nputs. Where multple exchange rates operate, an estmate of the mport or export tax equvalents of that dstorton are ncluded as well. The range of measures ncluded n the CTE estmates nclude both domestc consumer taxes and subsdes and trade and exchange rate polces, all of whch drve a wedge between the prce that consumers pay for each commodty and the nternatonal prce at the border. Anderson and Masters (2009) note several patterns that emerge n the dstortons to agrcultural ncentves n the 21 focus countres. In the 1960s and 1970s, many Afrcan governments had macroeconomc, sectoral and trade polces that ncreasngly favoured the urban sector at the expense of farm households and favoured producton of mport-competng farm goods at the expense of exportables. The polcy regme was characterzed as pro-urban (ant-agrcultural) and pro-selfsuffcency (ant-agrcultural trade). Snce the 1980s, Afrca has reduced ts antagrcultural and ant-trade bases, but many dstortons stll reman. For the 19 countres n ths study, Table 1 and Fgure 1 llustrate those patterns. The weghted average NRA for the 19 countres s almost always below zero, ndcatng that together agrcultural prce, trade and exchange rate polces have reduced the earnngs of farmers n these countres. The average rate of drect taxaton (negatve NRA) of Afrcan farmers rose untl the late 1970s before declnng by more than half over the next 25 years. Meanwhle, assstance to nonagrcultural sectors rose (thereby makng farmng less attractve n relatve terms) and then declned slower than for agrculture, as reflected n the Relatve Rate of Assstance (RRA) estmates n Table 1. Table 2 reports the country-level NRAs for covered products for each of the 19 countres n ths sample. It reveals the consderable dversty wthn the sample. In some countres such as Cameroon, Ghana, Senegal, Uganda, Tanzana, and Madagascar there was a reducton n taxng farmers snce the regonal peak n 10

15 , whle n other countres such as and Cote d Ivore, Zamba, and Zmbabwe hgh levels of agrcultural taxaton persst. The country level aggregate measures hde the degree of varaton n NRA estmates wthn countres. Anderson and Masters (2009) report the standard devatons around the weghted mean NRA for covered products n each country, showng that the varaton s sgnfcant. An ndcaton of the extent of varaton between groups of products s seen when comparng the average NRAs for mportcompetng and exportable product groups, whch reflects the anttrade bas (Fgure 1). Notwthstandng the valuable contrbuton of the measures reported n Anderson and Masters (2009), sectoral averages of NRAs and RRAs can be msleadng as ndcators of the aggregate extent of prce dstorton wthn the sector. They can also be msleadng when compared across countres that have varyng degrees of dsperson n ther NRAs (and CTEs) for farm products. We therefore now turn to consderaton of the TRI and WRI seres estmated for ths paper, and the addtonal nsghts these measures can provde. Trade and welfare reducton ndex estmates The regonal aggregate TRI for the 19 Afrcan focus countres for all covered products s postve and of a sgnfcant magntude over the perod under analyss (Fgure 2). The postve TRI ndcates that overall agrcultural polcy n Afrcan countres resulted n reduced trade. The extent of that has decreased over tme, however, wth the fve-year TRI averages of between 20 and 25 percent n the frst two decades of data fallng to around 10 percent n the most recent decade. The TRI has the opposte sgn to the NRA because the TRI correctly aggregates polces that reduce trade volume, regardless of whether the NRA s postve or negatve. The mportance of the dfference n these aggregatons of the trade-reducng effect of polces can be seen n the early-1960s, for example, when the average NRA was around zero but the TRI was qute hgh (capturng the trade-reducng effect of both mport taxes and export taxes, whch offset one another n the NRA estmate). Smlarly n the late 1980s, the NRA trends from around -15 to -10 percent at a tme when the TRI ncreases from 20 to 30 percent. The aggregate NRA gves the 11

16 mpresson that polces are becomng less dstorted n ths perod but, because the upward trend n the NRA s caused by an ncrease n mport taxes, the TRI correctly reveals that agrcultural polces are n fact becomng more restrctve n ths tme perod. The WRI seres for all covered products s necessarly postve and everywhere les above the TRI seres (Fgure 2). The WRI seres correctly demonstrates the negatve welfare consequences that flow from both negatve and postve prce dstortons. Furthermore, the WRI seres provdes a better ndcator of the welfare cost of dstortons than the average level of assstance or taxaton, due to the ncluson n the WRI of the power of two. A weghted arthmetc mean does not fully reflect the welfare effects of agrcultural dstortons because the dsperson of that support or taxaton across products has been gnored. By contrast, the WRI captures the hgher welfare costs of hgh and peak levels of assstance or taxaton. That s, the WRI reflects the dsparty ssue dscussed n Lloyd (1974): the larger the varance n assstance levels, the greater the potental for resources to be used n actvtes whch do not maxmze economc welfare. The aggregate Afrcan results mask country-level dversty n the TRI and WRI seres. Some countres such as Cote d Ivore, Ethopa, Sudan, Tanzana and Zmbabwe persstently restrct trade (n aggregate) throughout the perod under analyss (Table 3). Other countres such as Kenya, Zamba and Mozambque have had perods n whch polces n aggregate have expanded agrcultural trade slghtly. In terms of the WRI, there s less dversty across countres, snce WRI measures are all necessarly postve (Table 4). The extent to whch agrcultural polcy reduced aggregate welfare does dffer across countres, however. Some countres have low reductons n welfare, ncludng Uganda and most cotton-exportng countres. Fgure 3 provdes a snapshot for of the dversty n the WRI and TRI for each of the 19 countres, wth the weghted Afrcan average n the mddle. A useful way of understandng the overall welfare reducton for Afrca from agrcultural polcy s to compute the country contrbutons to the WRI for the 19 Afrcan focus countres as a whole. Contrbutons can be found by computng dollar values of the welfare reducton for each country (by multplyng the WRI percent by the average of the gross value of producton and consumpton at undstorted prces). Such contrbutons wll therefore take account of the magntude of natonal WRIs as well as the sgnfcance of each country n terms of ts share of the gross value of 12

17 producton and consumpton at undstorted prces. Table 5 shows that Ngera and Sudan are the two countres that domnate the regon s contrbutons, wth Sudan becomng more mportant over tme (as ts WRI seres trends upwards). Ethopa accounts for up to 10 percent of the focus regon s welfare reducton. The last column of Table 5 reports country contrbutons to the declne n the regonal WRI from ts value of 44 percent n to ts value of 27 percent n Once agan, Ngera and Sudan domnate the overall reducton, together accountng for around 80 percent of the fall n the WRI. However, Uganda, Cameroon, Senegal and Madagascar have a slghtly offsettng effect on the regonal fall n the WRI over that perod. It s useful to compare the TRI and WRI seres reported above for all covered products, wth those for just covered tradables n Afrca. 5 In Table 6, t can be seen that the TRI and WRI for all covered products s sgnfcantly lower than that for covered tradables. Ths s because nontradables account for a large share of the gross value of producton and consumpton. The TRI estmates for all covered products are roughly half, and WRI estmates are roughly two-thrds, what there were wth nontradables ncluded. Another pont to note from Table 6 s that the country sample matters for the reportng of TRI and WRI results. For comparson, we report the results from Lloyd, Croser and Anderson (2010), whch computed TRI and WRI seres for an alternatve sample of Afrcan countres n the Dstortons to Agrcultural Incentves database. Ther sample s the same 19 countres n ths paper, wth the addton of Egypt and South Afrca, and excludng the fve cotton countres of Benn, Burkna Faso, Chad, Mal and Togo, because these countres only have one covered tradable product (cotton). In general, the 19 focus countres n ths study have hgher TRI and WRI 5-year averages. Ths s drven by the excluson of Egypt and South Afrca, whch had low country-level TRI and WRI estmates. The excepton to the general pattern s the tme perod , where the Lloyd, Croser and Anderson (2010) estmates are hgher, owng to very hgh protecton n Egypt n that fve-year average perod (when nternatonal food prces collapsed just as Egypt rased ts prevously very low domestc food prces). 5 Note that Lloyd, Croser and Anderson (2010) report measures for covered tradables only. 13

18 It s also useful to compare the TRI and WRI results for the 19 focus countres to the TRI and WRI estmates for other developng country regons, whch are reported n Lloyd, Croser and Anderson (2010). The 19 Afrcan focus countres had the most welfare reducng polces over tme, and generally the most trade-dstortng. All three regons have shown a trend towards less trade and welfare reducng agrcultural polces n recent years, however (Fgure 4). Polcy nstrument results We now turn to the natonal decompostons of the TRI and WRI to the polcy nstrument level. Fgure 5 provdes a summary of the estmates of the contrbuton to the weghted average WRI seres for the 19 Afrcan focus countres of four dfferent border measures: taxes and subsdes on both mports and exports. The fgure demonstrates the very substantal role that export taxes have played n the reducton of welfare n the regon. On average, more than half the welfare reductons have come from ant-agrcultural export taxng polces over the perod studed. Notwthstandng ther sgnfcant dstortonary contrbuton, export taxes have also been the area n whch there has been most reform n recent decades. The contrbuton of export taxes to the reducton n the WRI over the perod to s 93 percent. Import taxes reduced the overall WRI by 34 percent; whle there were offsettng ncreases n the contrbuton of export subsdes (13 percent) and mport subsdes (15 percent) to the WRI. The contrbutons to TRI and WRI estmates for the 19 Afrcan countres from domestc dstortons are small, never accountng for more than 2 percent of the overall regonal TRI or WRI. Commodty TRI and WRI results The TRI and WRI estmates for ndvdual regonal commodty markets provde a dfferent perspectve on the level of dstorton n the 19 focus countres over the perod under analyss. Table 7 reports the fve-year average WRI estmates for the focus regon for ndvdual commodty markets. The table reveals consderable dversty n the dstortons n dfferent commodty markets. Frut and vegetable commodty markets, whch tend to have a hgh share of nontradable producton, have low WRI estmates on average, whereas traded commodtes such as tropcal crops, 14

19 olseeds and lvestock tend to have more welfare-reducng polces n place. Grans, whch comprse a mxture of tradable and nontradable products, had hghlydstortonary polces n place n the 1960s on average, but these have been reduced over tme. Fgure 6 gves a snapshot of the dversty across commodty markets n the regonal TRI and WRI for Sugar and cotton markets contnue to have hghly dstorted polces n terms of both the trade and welfare effects of polces. Soybean, by contrast, has trade-expandng polces n aggregate, but the polces are nevertheless welfare reducng. Conclusons, caveats and areas for further research Reform of agrcultural polcy n Afrca s topcal at present. Recently announced nternatonal nvestment programs, domestc polcy reforms, and the negotaton of nternatonal and regonal trade agreements are on the agenda. To assess each of these polcy ntatves, measurement of nterventon levels s requred. Certanly, economywde models can measure the welfare and trade (and other) effects of polcy n a partcular country or market. But such models requre relable data on the structure of the economy, and econometrc estmates of myrad parameters, nether of whch can be easly found for the poorer countres of Afrca. Even where economy-wde models are avalable, they may be calbrated to a partcular year and so ncapable of provdng a long tme-seres of estmates of the regonal effect of dstortonal polces over tme. Scalar ndex measures, by contrast, can meanngfully summarze the welfare and trade effects of polcy nterventon n agrculture n poorer countres. As demonstrated above, these ndces can be estmated usng already avalable prce, quantty and dstortons data, and so are relatvely nexpensve to generate. The estmates n the paper reveal several mportant fndngs. The natonal level TRIs and WRIs ndcate that although there has been polcy reform n Afrcan agrculture over the past 50 years, the overall trade- and welfare-reducng effects of current polces reman sgnfcant. Some ndvdual commodty markets n Afrca are more dstorted than others, sugar and cotton beng two of the most dstorted. The scalar ndex numbers reported n ths paper provde a better measure of polcy nterventon than wdely-used NRA-type measures because they correctly 15

20 aggregate offsettng polces and the WRI captures the hgher welfare costs of more dsparate polces across ndustres. These scalar measures have the advantage of makng polces more transparent, whch can facltate further reforms. Notwthstandng ther contrbuton, there are a number of lmtatons to the ndces. Some are emprcal. Frst, the estmates can only take account of agrcultural products whch have commodty level data n the World Bank s database. The database has product level data for approxmately 70 percent of the 19 Afrcan focus countres farm producton value, and somewhat less of ther consumpton value. We, therefore, necessarly mss some nformaton about dstortons to producton and consumpton and therefore trade and welfare n Afrca. Furthermore, the data n the World Bank s database are not hghly dsaggregated, whch s not deal for capturng the full extent of welfare and trade dstortons from Afrcan polces affectng dfferentable processed as well as prmary products. Fnally, n the absence of relable, consstently estmated tme-seres of elastctes of demand and supply, we make smplfyng assumptons about those elastctes to compute the scalar ndex number seres. The estmates would be more precse f we had access to relable elastcty estmates, although probably not a lot dfferent, accordng to senstvty analyss conducted by Croser, Lloyd and Anderson (2010). There are also some methodologcal caveats worth notng. The methodology n the paper adopts the standard approach stll presented n most textbooks on trade polcy or welfare economcs, based on the benchmark of compettve markets. The methodology gnores the exstence of dvergences and governance problems, ncludng admnstratve costs. Thus the trade and welfare reducton ndexes reported above may be over- or under-stated to the extent that such problems exst. For example, n some cases where there s market falure, we know from second-best theory that polces that ncrease assstance to a lghtly protected ndustry may ncrease rather than decrease natonal economc welfare. In partcular, the RRA measure reported n Table 1 suggest that dstortons to non-farm tradables sectors n Afrca exst. Even so, the seres reported n ths paper are useful aggregatons of data and almost certanly gve a better ndcaton of trade and welfare effects of polcy than average NRA-type measures. 0(3): Anderson, J.E. and J.P. Neary (2005), Measurng the Restrctveness of Internatonal Trade Polcy, Cambrdge MA: MIT Press. 16

21 Anderson, K. (2009b), Fve Decades of Dstortons to Agrcultural Incentves, chapter 1 n K. Anderson (ed.), Dstortons to Agrcultural Incentves: A Global Perspectve, , London: Palgrave Macmllan and Washngton DC: World Bank. Anderson, K. and J.L. Croser (2009), Natonal and Global Agrcultural Trade and Welfare Reducton Indexes, 1955 to 2007, database avalable at Anderson, K. and W. Masters (eds.) (2009), Dstortons to Agrcultural Incentves n Afrca, Washngton DC: World Bank. Anderson, K. and E. Valenzuela (2008), Global Estmates of Dstortons to Agrcultural Incentves, 1955 to 2007, database avalable at Bach, C. and W. Martn, (2001), Would the Rght Tarff Aggregator for Polcy Analyss Please Stand Up? Journal of Polcy Modelng 23: Chen, S, and M. Ravallon (2008), The Developng World s Poorer Than We Thought, But No Less Successful n the Fght Aganst Poverty, Polcy Research Workng Paper 4703, World Bank, Washngton DC, August. Croser, J. and K. Anderson (2010), Changng Contrbutons of Dfferent Agrcultural Polcy Instruments to Global Reductons n Trade and Welfare, Contrbuted Paper for the Annual Conference of the Australan Agrcultural and Resource Economcs Socety, Adelade, February. Croser, J.L., P.J. Lloyd and K. Anderson (2010), How do Agrcultural Polcy Restrctons to Global Trade and Welfare Dffer Across Commodtes? Amercan Journal of Agrcultural Economcs Vol. 92 (forthcomng). Feenstra, R. (1995). Estmatng the Effects of Trade Polcy, n G. Grossman and K. Rogoff (eds.), Handbook of Internatonal Economcs, Vol. 3, Amsterdam: Elsever. Irwn, D. (2008), Trade Restrctveness and Deadweght Losses from U.S. Tarffs, , NBER Workng Paper No. W Kee, H.L., A. Ncta and M. Olerreaga (2009), Estmatng Trade Restrctveness Indexes, Economc Journal 119(534): Lloyd, P.J. (1974), A More General Theory of Prce Dstortons n an Open Economy, Journal of Internatonal Economcs 4(4): , November. 17

22 Lloyd, P.J., J.L. Croser and K. Anderson (2009), Welfare-Based and Trade-Based Indcators of Natonal Agrcultural Dstortons, Ch. 11 n K. Anderson (ed.), Dstortons to Agrcultural Incentves: A Global Perspectve, , London: Palgrave Macmllan and Washngton DC: World Bank. Lloyd, P.J., J.L. Croser and K. Anderson (2010), Global Dstortons to Agrcultural Markets: New Indcators of Trade and Welfare Impacts, 1960 to 2007, Revew of Development Economcs 14(2), May (forthcomng). OECD (2009), Agrcultural Polces n OECD Countres: Montorng and Evaluaton 2009, Pars: Organzaton for Economc Cooperaton and Development. World Bank (2007), World Development Report 2008: Agrculture for Development, Washngton DC: World Bank. 18

23 Appendx: Dervaton of Trade- and Welfare-Reducton Indexes Lloyd, Croser and Anderson (2010) outlne a methodology for computng ndces whch accurately capture the state of trade polcy regme n an ndvdual country n a theoretcally meanngful way. Ther methodology, whch draws heavly on the Anderson and Neary (2005) methodology, defnes partal equlbrum ndexes whch aggregate the producton and consumpton sdes of the economy separately (nstead of trade data as s more commonly done wth trade restrctveness ndexes). Ths form of ndex s well-suted to agrcultural dstortons research, where data s avalable for producton and consumpton of ndvdual farm commodtes. Ths Appendx brefly outlnes that theory for the mport-competng sector of a small open economy. Start by consderng an ndvdual country, assumng t has a small, open economy n whch all markets are compettve. The market for an mport good may be dstorted by a tarff and other nontarff border measures or by behnd-the-border measures such as domestc subsdes and prce controls.the frst measure of nterest s the effect of a country s dstortons on ts mport volume, the TRI. Ths s defned as the unform tarff rate whch, f appled to all goods n the place of all actual border and behnd-the-border prce dstortons, would result n the same reducton n the volume of mports (summed across products by valung them at the undstorted border prce) as the actual dstortons. Consder the market for one good, good, whch s dstorted by a combnaton of measures that dstort ts consumer and producer prces. For the producers of the good, the dstorted domestc producer prce, P p, s related to the border prce, p *, by the relaton, P p = p * (1 + s ) where s s the rate of dstorton of the producer prce n proportonal terms. For the consumers of the good, the dstorted domestc consumer prce, C p, s related to the border prce by the relaton, C p = p * (1 + r ) where r s the rate of dstorton of the consumer prce n proportonal terms. In general, r s. Usng these relatons, the change n the value of mports n the market for good s gven by: M = p x p y * * 19

24 = p dx / d r p dy / d s (1) *2 C *2 P p p where the quanttes of good demanded and suppled, x and y, are functons just C P of ther own domestc prce: x = x( p ) and y = y ( p ). Strctly speakng, ths result holds only for small dstortons. In realty rates of dstorton may not be small. If, however, the demand and supply functons are lnear over the relevant prce range, the effect on mports s gven by equaton (1) wth constant slopes of the demand and supply curves ( dx / and C dp respectvely). If the functons are not lnear, ths expresson provdes an approxmaton to the loss. P dy / dp, Wth n mportable goods subject to dfferent levels of dstortons, the aggregate reducton n mports, n the absence of cross-prce effects n all markets, s gven by: n n *2 C *2 P / p / p = 1 = 1 M = p dx d r p dy d s (2) Settng the result equal to the reducton n mports from a unform tarff, T, gves: n *2 C *2 *2 / n P / n p p = / = 1 = 1 = 1 Solvng for T, gve p dx d r p dy d s p dm dp T T = { Ra + Sb} (3a) where n R = ru wth = 1 C u = p dx / d / p dx / d p, (3b) *2 C *2 p S wth n = sv = 1 P v = p dy / d / p dy / d p, and (3c) *2 P *2 p *2 P *2 a= p dx /d / p d m /dp and b= p dy /d p / p d m /dp (3d) *2 C *2 p Evdently, the unform tarff T can be wrtten as a weghted average of the level of dstortons of consumer and producer prces ( R and S are ndces of average consumer and producer prce dstortons; they are arthmetc means). An mportant advantage of usng ths decomposton of the ndex nto producer and consumer effects s that t treats correctly the effects of NTMs and domestc dstortons that affect the two sdes of the market dfferently. In equaton 3c (equaton 3b), the weghts for each commodty are proportonal to the margnal response of domestc producton (consumpton) to 20

25 changes n nternatonal free-trade prces. These weghts can be wrtten as, among other thngs, functons of the domestc prce elastctes (at the protected trade stuaton) of supply and demand ( σ and ρ, respectvely): 6 * * * ρ ( p x ) / ρ ( p x and = ( p y ) / u = ) n v σ σ ( p y ) (4) The other ndex defned n Lloyd, Croser and Anderson (2010), the WRI, measures the effect of a country s dstortons on ts economc welfare. The dervaton follows the same steps as n the dervaton of the TRI except that nstead of startng from the loss n trade volume from a polcy, one starts from a loss of consumer and producer surplus (a welfare loss, L ). Wth n mportable goods subject to dfferent levels of dstortons, the aggregate welfare loss, n the absence of cross-prce effects n all markets, s gven by: n n 1 * 2 P * 2 ps 2 p pr = 1 = 1 L = { ( ) dy / d ( ) dx / d p } (5) The unform tarff rate, W, that generates an aggregate deadweght loss dentcal wth that of the dfferentated set of tarffs s determned by the followng equaton: n * 2 P * 2 * 2 ( ) dy /d n C ps ( ) dx /d n p pr p = ( pw ) d m/dp = 1 = 1 = 1 (6) W s thus the unform tarff whch, f appled to all goods n the place of all actual tarffs and NTMs and other dstortons, would result n the same aggregate loss of welfare as the actual dstortons. Solvng for W, we have: W R a S b 2 2 1/2 = { + } (7a) C n * where R n 1 2 = [ r ] 2 u (7b) = 1 S n 1 2 = [ s ] 2 v (7c) = 1 wth u, v, a and b as defned for equaton 3 above. W s the desred Welfare Reducton Index, whle R and S are the contrbutons to W from consumer and producer prce dstortons, respectvely. They, lke ther approprately weghted average W, are means of order two. As wth the ndex T, we can deal wth, and analyse, the producton and consumpton sdes of the sector separately. 6 These expressons can also be wrtten as functons of, among other thngs, the domestc prce elastctes at the free trade ponts. 21

26 Extenson to exportable sectors Lloyd, Croser and Anderson (2010) report how the ndexes can each be extended to nclude the exportables sub-sector. Ths s facltated by way of aggregatng the mport-competng and exportables sub-ndces where the weghts for each subsector are the share of the sub-sectors value of producton (consumpton) n the total value of producton (consumpton). The resultng measure s the mport tax/export subsdy whch, f appled unformly to all products n the sector, would gve the same loss of welfare as the combnaton of measures dstortng consumer and producer prces n the mport-competng and exportable sub-sectors. The only trck n the case of the TRI s to keep separate track of the subsets of mport-competng and exportable goods because the sgn of an NRA n exportable sector (postve or negatve) has the opposte effect on the TRI. That s, whle an export subsdy n the exportable sub-sector reduces welfare n the same way as an mport tax n the mport-competng sub-sector, the export subsdy wll ncrease trade and the mport tarff reduces trade. Extenson to nontradables sectors In ths paper we make a further methodologcal extenson to the theory. We extend ndces to nclude nontradable, as well as tradable sectors. Ths s mportant for Afrca, because n many countres the share of nontradables n the gorss value of agrcultural producton s hgh. Becasue nontradables are generally free of dstortons, an ndex that does not take nto account these sectors wll tend to overstate the trade- and welfare-reducng effect of agrcultural polcy. To nclude nontradables, we keep separate track of three subsectors of the economy: mport-competng, exportable and nontradable sub-sectors. We generate sub-sector specfc TRI and WRI ndces (as we prevously dd for each of the mport-competng and exportable subsectors). The three sub-sector ndces are then aggregated usng as weghts each sub-sectors share of value of producton (consumpton) n the total value of producton (consumpton). 22