HOUSEHOLD INCOME, LAND VALUATION AND RURAL LAND MARKET PARTICIPATION IN ECUADOR

HOUSEHOLD INCOME, LAND VALUATION AND RURAL LAND MARKET PARTICIPATION IN ECUADOR By MARIA JOSE CASTILLO A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008 1

2008 María José Castllo 2

To my famly n Ecuador 3

ACKNOWLEDGMENTS I thank my mom and dad for ther constant love and support durng ths journey away from home. I also thank my husband Santago J. Bucaram for hs love, help and wsdom, and for keepng pushng me forward durng the progress of ths dssertaton. Specal thanks I gve to my char, Dr. Carmen Dana Deere, whose valuable support made possble the completon of ths research work. I thank my cochar, Dr. Ramón Espnel, as well for belevng n me ever snce I met hm. I thank my other commttee members, Charles Moss, Plar Useche, and Grenvlle Barnes for apprecatng the effort I put nto ths endeavor. I thank the Insttuto Naconal de Estadístcas y Censos (INEC) for provdng me wth the data set that gave place to ths work at a tme when t was not freely avalable on-lne. I gve specal thanks to Ing. Jula Carrón from the Sstema de Informacón Geográfca y Agropecuara (SIG-AGRO) of the Ecuadoran Mnstry of Agrculture for provdng me wth hard to get geographc nformaton whch contrbuted to my study. I am also thankful to Dr. Jeff Burkhardt, Graduate Coordnator, for always apprecatng my work and beng one of my advocates. Also to Dr. Ronal Ward for hs wllngness to help me understand econometrc applcatons every tme I looked for hm. I also thank Dr. Andrew Schmtz for offerng me the experence of publshng two papers on an nterestng U.S. topc. The Department of Food and Resource Economcs deserve my apprecaton for the fnancal assstance they made avalable to me, wthout whch my Ph.D. would have not beng possble. Fnally, I thank my frend Lly for her fathfulness and for provdng me wth a place to stay durng the fnal stage of my dssertaton. 4

TABLE OF CONTENTS ACKNOWLEDGMENTS...4 LIST OF TABLES...7 LIST OF FIGURES...9 ABSTRACT...10 CHAPTER 1 INTRODUCTION...12 page Objectves...13 Hypotheses...14 2 THE IMPACT OF LAND INEQUALITY ON ECUADORIAN HOUSEHOLD INCOME...16 Land Problems and Rural Poverty...16 Multple Market Imperfectons and the Household Income Problem...19 Statstcal Analyss...20 Credt Access...22 Value Product per Unt of Labor...25 Labor Allocaton...26 Value Product per Unt of Land...26 The Land Rental Market...27 The Model...29 Household Income Per Capta Estmaton...35 Conclusons...42 3 UNDERSTANDING LAND RESERVATION VALUES IN THE PRESENCE OF MULTIPLE MARKET IMPERFECTIONS: THE ECUADORIAN CASE...50 Introducton...50 Data and Methodology...51 Restrcted Profts...52 Land Reservaton Prces...56 Results...60 Conclusons...63 4 RURAL LAND MARKET PARTICIPATION IN ECUADOR AND ITS DETERMINANTS...69 Introducton...69 5

Data and Methodology...72 Land Supply...72 Land Demand...78 Results...83 Supply Sde...83 Demand Sde...86 Conclusons...89 5 CONCLUSIONS...100 APPENDIX PRIMARY ACTIVITY OF FEMALE HOUSEHOLD HEADS...104 LIST OF REFERENCES...105 BIOGRAPHICAL SKETCH...109 6

LIST OF TABLES Table page 2-1 Number of farms by farm sze, Coast and Serra regons, Ecuador...44 2-2 Credt access, type of credt and operatonal farm sze...44 2-3 Mean loan terms by credt sector (all farm szes)...44 2-4 Agrcultural labor productvty and operatonal farm sze...44 2-5 Household heads prmary actvty and household s man source of ncome by operatonal area...45 2-6 Mean land productvty by category of farm sze...45 2-7 Farm sze dstrbuton of land tenants...45 2-8 Varable defnton for household ncome per capta and credt equatons...46 2-9 Summary of explanatory varables (ncome and credt regressons)...47 2-10 Credt regressons for the probablty of obtanng credt and the amount of credt...48 2-11 Household ncome per capta regresson...49 3-1 Mean and medan quas-fxed factors...65 3-2 Classfcaton of owner-tenant households by category of owned farm sze...65 3-3 Credt constraned households by owned farm sze...65 3-4 Summary of varables (land reservaton value equaton)...65 3-5 Returns to fxed factors equaton (quadratc functon)...66 3-6 Log of the land reservaton prce equaton...67 4-1 Farm sze and land sales by owned land category...92 4-2 Incdence of land rentals (landlords) by owned farm sze category...92 4-3 Mean and medan statstcs of varables n Equaton 4-1...93 4-4 Farm sze and land purchases by owned land category (pror to purchase)...94 4-5 Forms of land acquston by gender...94 7

4-6 Summary of varables n Equatons 4-2 and 4-3...95 4-7a Land rented n by category of land owned...95 4-7b Choce of rental agreement by category of land owned...96 4-8a Multnomal logt regresson results (owners decsons between farmng, sellng or rentng out)...96 4-8b Multnomal logt regresson results (owners decsons between farmng, sellng, rentng under fxed-rent or under shared-rental contracts)...97 4-9 Censored Tobt regressons (amount of land rented-out and sold)...97 4-10 Logt for probablty of purchase and censored Tobt for amount of land bought...98 4-11 Logt for probablty of rentng-n and censored Tobt for amount of land rented...98 4-12 Multnomal logt for probablty of rentng-n...99 A-1 Prmary actvty of female household heads by farm sze...104 8

LIST OF FIGURES Fgure page 2-1 Mean labor productvty by farm sze...49 3-1 Shadow land values...67 3-2 Land reservaton prces per hectare and non-prce ratoned households...68 9

Abstract of Dssertaton Presented to the Graduate School of the Unversty of Florda n Partal Fulfllment of the Requrements for the Degree of Doctor of Phlosophy HOUSEHOLD INCOME, LAND VALUATION AND RURAL LAND MARKET PARTICIPATION IN ECUADOR Char: Carmen Dana Deere Cochar: Ramón Espnel Major: Food and Resource Economcs By María José Castllo August 2008 Ths research provdes an economc analyss of agrcultural land access at the household level and ts relatonshp wth rural markets and poverty n Ecuador. We fnd that land nequalty and land market mperfectons have a drect effect on household ncome per capta and that there s a synergy between these and mperfectons n the labor and credt markets, whch magnfy the effect of land nequalty on rural household ncome. In addton, the presence of multple market mperfectons ntensfes the quas-fxty of factors other than land, whch affects the contrbuton of land to profts and land values. The labor advantage of small farmers explans the remarkable dfference n reservaton prces per hectare between small and medum and large farmers. However, ths effect s reduced for credt constraned households. Lack of land ttles s not found to dscourage nvestments n land or to cause land values to be smaller than for households wth tled land. Consstent wth these fndngs, we also observe that the demand for land by small farmers s sgnfcantly larger than the supply of land by large landowners both n the land sales and rental markets. Small farmers are found to be more actve than larger farmers on both sdes of the land markets and sharecroppng arrangements are found to be especally common among the 10

land poor. Land ttles have a sgnfcant and postve effect on the lkelhood to sell and smlarly, credt access on the lkelhood to purchase and rent n land. We conclude that, gven the dffcultes that prevent desred land transfers from large landowners to the rural poor, t seems mprobable that the market be able to acheve an optmal dstrbuton of landownershp wthout assstance from the government. Also, that for the potental benefts for rural development of ncreased land access to be realzed, such an ncrease must be accompaned by better access to servces so as to mprove the compettveness of the rural poor. Polces regardng the lberalzaton and stmulaton of land rental markets and ncrease n the supply of credt n the rural sector are recommended. 11

CHAPTER 1 INTRODUCTION Latn Amercan countres, ncludng Ecuador, are known for ther severe ncome and land nequalty. Ths explans the persstent nterest by the regon s governments as well as nternatonal development organzatons n land redstrbuton and n enhancng land productvty n Latn Amerca. In Ecuador, accordng to the 2000 agrcultural census, the Gn coeffcent for land was 0.8, the same as for Latn Amerca as a whole. Land reform n Ecuador, whch took place durng the perod 1964-79, dd lttle to mprove land dstrbuton. More mportantly, access to land va the rental market s very lmted as well. Accordng to the agraran census, only two percent of the farms are under fxed or share-rent tenancy and 16% are under mxed tenancy (owner occuped combned wth leasng or sharecroppng). Ths s due n part to current Ecuadoran legslaton whch mpedes the free development of land rental markets. Though agrculture remans an mportant contrbutor to natonal ncome and a source of employment for about 30% of Ecuador s workng populaton, agrcultural polces have been defcent and unstable. In 2000, about 70% of the country s rural populaton earned ncomes under the poverty lne. It s commonly argued that an mportant reason for rural poverty s the lmted access to land of the rural poor. On the one hand, large landholdngs do not use ther land ntensely enough so as to generate suffcent employment opportuntes for the resource poor. On the other hand, poor landowners must explot ther land more ntensvely than envronmentally desrable, whch worsens sol degradaton and lowers productvty. Furthermore, not only s land ownershp and access to land unequally dstrbuted, but access to captal, technology and product markets s as well. The land market also suffers from segmentaton, where the rch trade land among the rch and the poor among the poor and where ethncty and knshp play an mportant role n land 12

transfers, especally n the Serra (Lambert and Stanfeld, 1990). Consequently, land, credt and other market mperfectons (as n the labor market) affect farm ncome generaton and land prces, fnally determnng who can partcpate n the land market (buyng, sellng, rentng n or rentng out). In turn, low farm ncome and hgh land prces for the poor strengthen market mperfectons and nequalty. Snce the 1994 Land Law was approved land redstrbuton efforts have been left to the market. Gven current restrctons on rentng land, and the market mperfectons just descrbed, how well can the land market perform ths task? Ths dssertaton explores ths queston through the study of a natonal household survey data that covers all coastal and hghland provnces of Ecuador. Objectves The objectves of the dssertaton are a) to understand quanttatvely the role of land nequalty and rural market mperfectons on rural poverty, through an analyss of the effect of these factors on the level of household ncome; and b) to dentfy the key varables that explan the persstence of land market mperfectons and nequalty. Research queston 1: What s the effect of land nequalty, land rental market restrctons, and holdng unttled land on captal access and rural ncome generaton? Research queston 2: In the survey year, what was the role of the land sale and rental markets (fxed or share tenances) n land dstrbuton and access? a. To what extent do land reservaton values reflect land qualty and productvty, as opposed to neffcences n land and related markets (credt, labor, techncal assstance) or other non-productve factors (such as holdng land for status)? b. What varables determne partcpaton n the land sales and rental markets and the extent of partcpaton? 13

Hypotheses Hypothess 1: Market mperfectons and land rental market restrctons sustan land nequalty and consequently rural poverty n Ecuador. a. In Ecuador, farm sze affects household ncome drectly and ndrectly through ts effect on credt access and labor allocaton. b. Land ttles contrbute to farm ncome prmarly by facltatng access to credt. c. Insecurty of property rghts and restrctons on rentng land out contrbute to segmentaton n the land rental market, hence lmtng the amount of land the rural poor can access and consequently contrbutng to rural poverty. Hypothess 2a: Land reservaton values are negatvely affected by restrctons n the credt market, hence lowerng the compettveness of the rural poor n the land market. a. Provded that small farmers are more productve (hgher value product) per unt of land, the contrbuton of land to restrcted profts (shadow land values) s decreasng n operatonal area. b. Gven credt market mperfectons, land reservaton prces per hectare are lower for credt constraned households. land poor. Hypothess 2b: Large landowners do not make land avalable though sales or rentals to the a. Due to land sales after the agraran reform, and a lkely process of reconsttuton of latfunda (Jordan, 2003), small farmers are more actve n the land market as sellers and large landowners as buyers. b. The land rental market s frendler than the land sales market for the rural poor n Ecuador. Ths study s dvded nto three man chapters. In Chapter 2 we test hypothess 1 through the use of a theoretcal model whch shows the effect of one more unt of land on household ncome when there are multple market mperfectons. Descrptve and econometrc analyses add to the dscusson and corroborate such effects. Hypothess 2a s tested n Chapter 3 by analyzng the confguraton of land values. Frst, ths chapter presents the estmaton of a restrcted proft functon and land shadow values. It then categorzes households nto credt constraned and 14

credt unconstraned and uses these peces of nformaton n the estmaton of land reservaton prces per hectare. Graphc analyss helps understand how land values per hectare vary wth farm sze. Chapter 4 studes hypothess 2b usng descrptve analyss of household partcpaton n the land sales and rental markets and econometrc estmatons whch hghlght the varables that nfluence the lkelhood to sell, purchase, rent out or rent n land (as well as the amount of land nvolved n each type of transacton). Here we analyze the role of the land market as a channel for land redstrbuton. Fnally, Chapter 5 offers general conclusons as well as polcy recommendatons. 15

CHAPTER 2 THE IMPACT OF LAND INEQUALITY ON ECUADORIAN HOUSEHOLD INCOME Land Problems and Rural Poverty Accordng to the last agrcultural census (2000) n Ecuador, the Gn coeffcent for land was 0.8 (wth 1.0 beng equal to perfect nequalty), smlar to that of Latn Amerca as a whole, the regon wth most unequal land dstrbuton n the world. 1 The agrcultural census shows that 64% of the total 843,000 agrcultural producton unts n Ecuador are of less than fve hectares n sze and farm only 6.3% of Ecuador s total cultvable land. On the other hand, 6.4% of all productve unts each hold 50 or more hectares for a total of 61% of the agrcultural land. Land reform n Ecuador, whch took place durng the 1964-79 perod, dd lttle to mprove land dstrbuton (Otáñez et al., 2000; Chrboga and Rodrguez, 1998). 2 More mportantly, access to land va the rental market s very lmted as well. Accordng to the agrcultural census, only 2% of the farms are under fxed or share-rent tenancy and 16% are under mxed tenancy (owner occuped combned wth leasng or sharecroppng). 3 Ths s due n part to current Ecuadoran legslaton whch prevents the free development of land rental markets. 4 Though agrculture remans an mportant contrbutor to natonal ncome and a source of employment for about 30% of Ecuador s workng populaton, agrcultural polces have been defcent and unstable. In 2000, about 70% of the country s rural populaton earned ncomes 1 Hayam and Otsuka (1993) show that the Gn coeffcent for operatonal farmland dstrbuton n Latn Amerca s hgher than 0.8, much larger than the coeffcent for developng countres n Asa. Also, de Ferrant et al. (2003) note that ncome as well as asset nequalty s hgher n Latn Amerca and the Carbbean than n Asa, Eastern Europe, and the 30 countres of the Organzaton for Economc Cooperaton and Development. Average Gn coeffcents from 1966 to 1990 (obtaned by Dennger and Olnto, 2000) are 0.81 for Latn Amerca whle those for the Mddle East, North and Sub-Saharan Afrca, and East and South Asa are all lower than 0.7. 2 Chrboga and Rodrguez (1998:16) note that, compared wth the agraran reform experence of other countres n the regon, Ecuador s among those wth the least redstrbutve results. 3 In contrast, n Asa some 20 to 30% of the land s rented, n the Unted States, 40%, and n Belgum, 67% (FAO, 2002). 4 Legal and normatve types of lmtatons n the land rental market are addressed n Secton IV below. 16

under the poverty lne. More specfcally, the 1998-1999 Lvng Standard Measurement Survey shows that 50% of agrcultural households n the sample had an annual ncome equal to or smaller than $1,300; 75% reported ncomes smaller than $2,500 and 99% ndcated ncomes of less than $15,000. The average annual ncome of agrcultural households was close to $2,000, whch represents only 46% of the mean annual ncome of the total sample of households (5,816 households). 5 Snce the agrcultural porton of the sample (1,898 households) represents mostly non-urban households (90%), ths s an ndcator that rural poverty s more severe than urban poverty, 6 a fndng that s common n developng countres. It s commonly argued that an mportant reason for rural poverty s the lmted access to land of the rural poor. On the one hand, large landholdngs do not use ther land ntensely enough so as to generate suffcent employment opportuntes for the resource poor. On the other hand, poor landowners must explot ther land more ntensvely than envronmentally desrable, whch worsens sol degradaton and lowers productvty. Furthermore, not only s land ownershp and access to land unequally dstrbuted but access to captal, technology and product markets s as well. In addton, the land market n Ecuador suffers from segmentaton, where the rch trade land among the rch and the poor among the poor and where ethncty and knshp play an mportant role n land transfers, especally n the Serra (Lambert and Stanfeld, 1990). Accordng to the lterature, land market segmentaton s encouraged by land nsecurty or lttle protecton of property rghts (Marcours et al. 2005; FAO, 2002) and by hgh effectve land prces 5 The year of our survey data (October 1998- September 1999) represents a tme of severe economc crss n Ecuador, just before the dollarzaton of the economy (whch offcally took place n January, 2000). Yet, that should not be consdered an abnormal year n terms of poverty snce poverty had already been on the ncrease and contnued ncreasng n the followng years. 6 Smlarly, a FAO country profle for Ecuador based on year 2004 reports an agrcultural per capta GDP (agrcultural GDP/agrcultural populaton) that represents only 40.5% of natonal per capta GDP. 17

for the poor, beyond the productve ablty of the land, drven by mperfect captal markets and the exstence of hgh transacton costs n the land market (Carter and Salgado, 2001; Carter and Zegarra, 2000). In addton, low farm ncome and hgh land prces for the poor strengthen market mperfectons and nequalty. As argued by de Janvry et al. (2001), under multple market mperfectons lke these, mprovng land access for the poor can mprove both welfare and effcency. The objectve of ths chapter s to understand quanttatvely the role of land nequalty on the level of rural household ncome, at the same tme as we explore the hypothess that market mperfectons and land rental market restrctons sustan land nequalty and consequently rural poverty n Ecuador. The development lterature has already addressed n dfferent ways the effects of farm sze and unrestrcted land rental markets on farm ncome. Ths has been emprcally analyzed n several countres of the developng world; however, the results tend to vary from country to country, or even wthn the same country, dependng on the varables ncluded n the analyss. Moreover, prevous studes of these ssues n Ecuador have been only partal, lmted to certan sub-regons or varables. Ths study, whch covers a sample of farm households from all provnces of the coastal and hghland regons of Ecuador (an area that represents 77% of Ecuador s cultvable land), wll provde a comprehensve analyss of these ssues. Ths chapter s organzed as follows: Frst, we summarze the mportance of market mperfectons n household utlty maxmzaton. Then, n order to test for the exstence of a relatonshp between farm sze and 1) credt access, 2) value product per unt of labor, 3) labor allocaton and 4) value product per unt of land, we perform par wse analyzes based on our household data. Subsequently, we provde a bref descrpton of the land rental market stuaton 18

n Ecuador and ts possble nfluence on poverty. Then, we model the effect of addtonal land on ncome for an average agrcultural household, followed by an econometrc estmaton of household ncome per capta to test for the effect of farm sze and credt access among other relevant varables. We offer conclusons n our last secton. Our results suggest that land nequalty and land market mperfectons have a drect effect on household ncome per capta but also that there s a synergy between these and mperfectons n other markets such as credt and labor, whch are essental for agrcultural producton and productvty. These mperfectons magnfy the effect of land nequalty on household ncome. Multple Market Imperfectons and the Household Income Problem The agrcultural sector of developng countres suffers from multple market mperfectons, ncludng the credt, nsurance, labor and land markets. Development economsts (Sngh, et al., 1986; Bardham and Udry, 1999) have observed that multple market mperfectons nvaldate the classcal proft maxmzaton approach used n order to fnd (or understand) optmum nput allocaton of rural households. More specfcally, models of proft maxmzaton assume that producton decsons are ndependent of consumpton decsons such that nput choce depends only on nput and output prces and the avalable technology (Bardham and Udry, 1999). Under ths condton, the farm household s problem s separable and can be solved recursvely: producton decsons are made frst (proft s maxmzed) and consumpton decsons afterwards (thus utlty of consumpton can be maxmzed subject to a budget constrant that ncludes takng maxmzed profts as a gven). Ths mples that producton decsons affect consumpton decsons but not vce versa. In other words, preferences (e.g. between consumpton and lesure), household endowments (e.g. assets and labor), non-farm ncome and prces of consumpton commodtes do not affect producton decsons (Sngh, et.al., 1986, Bardham and Udry, 1999). 19

Ths structure, however, apples only when markets are complete or when there s only one market mperfecton, not when there are multple market mperfectons (Bardhan and Udry, 1999). In order to reach equlbrum, households must equate demand and supply for each commodty. The non-exstence or ncomplete presence of markets mpede ths equlbrum to happen at market prces, nstead t happens at what are called vrtual or shadow prces whch are dfferent from market prces and are endogenous to the household (Sngh et at, 1986). Vrtual prces and consequently the farmer s maxmzaton problem wll be a functon of household endowments, such as land and famly labor; market prces; off-farm labor market characterstcs and non-farm ncome, among other factors. Therefore, the approprate agrcultural household model under multple market mperfectons s that whch jontly consders producton and consumpton choces (Bardham and Udry, 1999). Accordng to ths analyss, land nequalty or rather an mprovement n land dstrbuton would nfluence nput choce (ncludng allocaton of famly labor), hence havng the potental of affectng farm productvty and household ncome as a result. The model below develops ths dea followng Fnan et al. (2005); frst, we perform par wse descrptve analyzes wth our household data n order to observe f and how sgnfcantly farm sze affects credt access, labor allocaton and labor and land productvty n the case of Ecuador. Statstcal Analyss We use data from Ecuador s Lvng Standard Measurement Survey (LSMS) 1998-99 provded by the Natonal Insttute of Statstcs and Census of Ecuador, n order to emprcally observe the effect of land access on credt and farm labor. The sample ncludes 5,816 urban and rural households from the Coast and Serra regons of Ecuador; 1,898 observatons of agrcultural households are used n ths chapter. Relevant sectons of the survey nclude questons on economc actvtes of the household members, credt access, land tenure, 20

agrcultural producton, farm labor, varable nput expenses, and ownershp of machnery and equpment as well as household demographcs. Here, we perform par wse analyses of credt access, labor allocaton and labor and land productvty wth respect to farm sze. The analyss below wll classfy farms n four categores based on farm sze. The frst category conssts of operatonal holdngs of less than 1 hectare. These are consdered mnfundos snce such small farm szes hardly allow for the subsstence of a household. These farms are treated as a separate category here gven ther predomnance n the sample (the natonal agraran census also reports these farm szes as a separate category). The next category, farms of one to less than fve hectares are stll consdered small and, as noted by Lopez and Valdez (2000), If not rrgated and ntensvely farmed ths amount of land cannot support levels of consumpton above the extreme poverty lne wthout other sources of ncome (Lopez and Valdez, 2000: 203). Farms of fve to less than 40 hectares are regarded as medum sze gven ther hgher probablty of beng medum captalzed unts, that s, unts belongng to farmers who have been successful n agrculture and have been able to accumulate land (or access more land) and other assets over the years. Fnally, farms larger than 40 hectares are treated as large. Table 2-1 summarzes ths nformaton. Smlar to the data gathered by the natonal agraran census, Table 2-1 shows that the largest category s made up by mnfundos and that the majorty of farm households operate less than 5 hectares. Moreover, the medan operatonal holdng s less than 2 hectares. Of the total number of farms n our sample, 73.8% are reported as owner-operated; 15.5% are partally owner-operated, and 10.6% are held by tenants only. 21

Credt Access The dependency of credt access on land wealth s a well known constrant n the developng world where dualstc structures 7 characterze the rural sector (Bardhan and Udry, 1999; Carter and Zegarra, 2000; Feder and Feeny, 1993). Besdes the fact that collateral s usually necessary n order to access the formal credt market, and that land s the most desrable type of collateral gven ts characterstcs, land ownershp s a sgn of economc -and at tmes poltcal- power whch facltates market mmerson and partcpaton (de Ferrant et al., 2003). We explore the exstence of a relatonshp between credt access and farm sze n the case of Ecuador. Table 2-2 shows the proporton of farm households classfed by operatonal area who obtaned credt for a postve nterest rate. 8 The proporton of loans receved from the formal sector and those from the nformal sector are also reported together wth the respectve credt amounts and nterest rates. The credt varable s total credt receved by the household, whch ncludes credt for agrculture, for a famly busness and/or for consumpton (purchase of durable goods, house buldng/remodelng, sckness, etc.). The reason for ncludng all types of credt receved by the household s that snce credt s fungble t can be used on any household need regardless of the purpose for whch the loan was obtaned. Besdes, ownershp of/access to land (whch s our focus here) s a sgnal to lenders as to how much debt responsblty a household can acqure. Thus we expect to fnd a relatonshp between farm sze and credt access even f we nclude 7 A dualstc structure n the rural sector refers to low productve small famly farms coexstng wth captalst farmers who hre labor and where the moblty of farm operators between the two sectors s severely lmted (Berry and Clne, 1979; Bardhan and Udry, 1999). 8 Households who receved credt for a null nterest rate are omtted for purposes of ths analyss snce such cases often nvolve small loans provded by relatves or frends, or credt receved n-knd by nput supplers or NGOs. 22

loans for famly busness or consumpton. The nterest rate (r%) s the average nomnal nterest rate faced by the household ncludng all types of credt. The ch-square statstc for the hypothess of ndependence between operatonal farm sze and credt access (null hypothess) reveals that there s a statstcally sgnfcant relatonshp between the two varables (the null hypothess s rejected at 5% level of sgnfcance). The null hypothess s also rejected (at 10% level of sgnfcance) when owned farm sze s used nstead of operatonal farm sze. The same test was performed for the case of formal credt, n whch the ndependence hypothess between farm sze and credt access was rejected at a 10% level of sgnfcance. We also tested the hypothess for access to nformal credt but ths tme there was a falure to reject. Ths result suggests that, as expected, non-nsttutonal lenders pay less attenton to farm sze than formal lenders snce the former tend to be much more famlar wth ther borrowers, hence, facng lower levels of mperfect nformaton. Also, statstcal analyss of the relatonshp between category of farm sze and the type of credt obtaned (formal vs. nformal) reveals that such a relatonshp s sgnfcant for mnfundstas (5% sgnfcance) and small farm szes (1% sgnfcance), wth the odds of gettng nformal credt beng hgher than the odds of accessng formal credt. 9 For medum and large sze farmers formal and nformal sources of credt are more equally accessble (and/or preferred) than for mnfundstas or small farmers. In addton, when analyzng the dfference n the amount of credt obtaned from formal and nformal sources, usng a t-statstc we fnd that, although the amounts of formal credt are greater for all operatonal szes, the dfference s only statstcally sgnfcant for the small sze 9 The data ndcate that a mnfundsta s 75% more lkely to get nformal credt than formal credt. The lkelhood for small farmers s 164%. 23

category. However, takng all szes together (Table 2-3), the mean dollar amount of formal credt s sgnfcantly larger than the mean dollar amount of nformal credt. Analyss of the nterest rates reveals that, except for the large farm sze category (Table 2-2), nformal credt nterest rates are sgnfcantly hgher than formal credt nterest rates. Ths s n lne wth what was expected gven the theory and typcal emprcal fndngs (for example see Andersen and Malchow-Moller, 2006). A somewhat ntrgung fndng n Table 2-3 s that only 20% of all formal loans requred real estate as collateral (compare ths to 58% n Peru n 1997 as reported by Gurknger and Boucher, 2005). Ths, however, can be explaned by observng the structure of the formal credt market n our sample. The bulk of formal credt s offered by prvate banks (33%) and cooperatves and assocatons (49%) and the rest by governmental nsttutons (11%) and NGOs (7%). The latter nsttutons typcally not requre borrowers to put real estate as collateral. Smlar s the case of cooperatves and assocatons. Fnally, although prvate banks would be expected to act dfferently than the other lenders, askng for valuable collateral such as real estate, the evdence ndcates that the loans offered by banks n the rural sector are n general small compared to those offered n the urban sector (Espnel, 2002); ths could explan the lttle need for ths type of collateral. 10 Ths result together wth our fndngs of a statstcally sgnfcant relatonshp between farm sze and credt access suggests that land ownershp s not necessarly functonal as collateral for formal credt but t s also a sgn of economc power whch facltates credt access. 10 The average amount of credt receved from prvate banks n the sample s US$2,143. 24

Value Product per Unt of Labor Evdence n developng countres has also shown that there s a drect relatonshp between farm sze and value product per unt of labor (usually referred to as labor productvty). Our data conform to what s expected (Table 2-4). Whle all farm szes reported the use of non-remunerated labor, medum and large farmers hred a sgnfcantly larger amount of labor than mnfundstas and small farmers (Table 2-4). Larger amounts of hred labor reflect the captalst nature of medum and large farmers, whch s manfested n hgher labor productvty. Captalst farmers hre labor up to the pont where margnal labor productvty equals the wage rate, whle tradtonal famly farms usually have larger amounts of labor per unt of land whch, gven labor and credt market mperfectons, they must allocate less effcently to the farm (Berry and Clne, 1979). More specfcally, snce moral hazard and hence the need for labor supervson s not an ssue when usng famly labor whle t s when hrng labor, famly labor tends to be more productve than hred labor (Bnswanger et al., 1993); however, the presence of mperfectons n the labor and credt markets (.e. unemployment and credt ratonng) cause small farmers to make a less effcent allocaton of labor to the farm compared to larger farmers, resultng n lower value product per unt of labor for small farmers. The t-tests of mean dfferences ndcate that the most sgnfcant [consecutve] dfference n productvty occurs between medum and large sze farmers (10% sgnfcance). However, the dfferences n productvty between a mnfundsta and a medum sze farmer and between a small and a large farmer are hghly sgnfcant (1% sgnfcance). Dfferences n mean labor productvtes for the four dfferent categores of farm szes can be better observed n Fgure 2-1, whch shows a clear ncrease n mean labor productvty as farm sze ncreases. 25

Labor Allocaton Table 2-5 shows the dstrbuton of the prmary actvty of household heads and the composton of household ncome by agrcultural and non-agrcultural sectors for each category of farm sze. 11 Table 2-5 llustrates that as farm sze ncreases so does the proporton of household heads who prmarly work on-farm. Smlarly, consderng total household ncome, t s more lkely that farmng s the man source of ncome for the household as farm sze ncreases. Also, both mnfundstas and small farmers rely more heavly on wage ncome (ether from the agrcultural or non-agrcultural sector) than medum and large farmers. Our data thus suggests that as farm sze ncreases so does the mportance of the farm busness for the household, resultng n a hgher level of ncome obtaned from agrculture (compared to other sources of ncome of the household). Takng nto account the sex of the household heads, 16% of them are women. They are over-represented among mnfundstas and under-represented among small and medum farmers (Table A-1). Of the household heads who are not economcally actve, these are slghtly more lkely to be female rather than male (53 vs. 47%). Also, female heads are more lkely to declare agrcultural self-employment as ther prmary actvty than male heads (55 vs. 45%). The female heads most lkely to declare agrcultural self-employment as ther prmary actvty are smallholders and large farmers. Value Product per Unt of Land The hypothess of an nverse relatonshp between farm sze and the value of total product per hectare (usually referred to as land productvty), often found n the developng world, s also 11 The decomposton of rural household ncome follows Corral and Reardon (2001). 26

tested here usng par wse analyss. Table 2-6 presents the mean land productvty for each group of farm sze. The t-tests of mean dfferences n land productvty show that the dfference n mean value product per hectare between farm sze categores s statstcally sgnfcant at a 5% level of sgnfcance, ndcatng that small farmers tend to explot the land more ntensvely than large farmers. Ths result was expected gven that, as noted earler, small farmers usually have a larger labor to land endowment rato than large farmers (Berry and Clne, 1979), hence output per hectare tends to decrease wth farm sze. The Land Rental Market It has been argued that land ownershp should not necessarly be the man objectve n order to mprove the lvelhood of the poor, but that access to land va other forms of tenure, frendler to the poor, should be earnestly sought too (de Janvry et al., 2001; Sadoulet et al., 2001, and Curre 1981). In accordance wth ths dea, nternatonal development organzatons advocate for lberalzaton of the land rental markets n developng countres. Ths process could be consdered to be only half-way mplemented n Ecuador because of legal as well as normatve types of lmtatons. 12 Among the legal lmtatons are the followng: a) Sharecroppng, a form of tenure that, although regarded as neffcent by some authors, has proved effectve n overcomng mperfectons n the captal and labor markets, was abolshed n 1970 by the Law of abolton of precarous forms of labor n agrculture, and t contnues to be llegal. b) Fxed-rent tenancy s allowed by the law but the law also contemplates the possblty of prescrpton of the owner s property rghts under certan condtons, namely, 1) f the landowner does not have a vald land ttle (properly regstered), he/she can easly loose the land to ther tenant; or 2) even n the presence of a vald land ttle by the owner, f the rental agreement s not n the form of a wrtten contract properly regstered, and f the tenant has been the land operator for at least 15 years, the landowner s property rghts can be prescrbed. 12 Based on FAO (2002) study. 27

c) The Consttuton stll ncludes the possblty of expropraton based on the concept of the socal functon of the land (art. 30). 13 Ths legslaton conserves the sprt of the agraran reform era; t gves more protecton to tenants than to landlords and dscourages supplyng land to the rental market. In addton, normatve condtons that hnder the rental market are as follows: d) Hgh transacton costs dscourage land ttle regstraton by landowners; e) Prolferaton of land conflcts (due to conflctng nhertance rghts or lack of ttlng); f) Lack of knowledge about the relevant legslaton and abundance of corrupt lawyers who ncrease the costs of legal processes; g) Lack of formal enforcement of property rghts; and h) Unequal ethnc and socoeconomc relatons Ponts (d) and (e) reflect the mpedments landowners face n order to obtan land ttle and consequently to satsfy requrements of the law n order to engage n formal rental contracts. Condtons (f) through (h) pont to the hgh rsk of losng property rghts that landowners would face f they decde to rent-out ther land. Ths analyss would explan the low ncdence of land rental agreements n Ecuador reported by the agraran census (see the frst secton of ths chapter). As a consequence, t has been found that land rental markets n Ecuador are segmented (Lambert and Stanfeld, 1990; FAO-COTECA, 1995; FAO, 2002). Landowners prefer tenants they already know and can trust and vce versa, hence the more economcally powerful rent among themselves and so do the poor, wth the ethnc component beng of mportance too. 14 Ths alone suggests that the landless poor would at best be able to access land of the poor, whch 13 Causes for expropraton are precarous forms of farm labor; technologes of producton that endanger natural resources; abandonng farmng for more than two consecutve years; and lands that, whle not fulfllng ther socal functon, face demographc pressure by peasant populatons. 14 Lambert and Stanfeld (1990) and FAO-COTECA (1995) note that land sale markets are also segmented by class and ethncty. 28

gven the potental effect of farm sze on poverty, would mply both that the amount of land they could access s small and that they would contnue to be poor. Ths mplcaton s strengthened by the prevous analyss on credt access and labor productvty, whch are nfluenced by farm sze. Fnally, our data (Table 2-7) ndcates that the large majorty (84%) of tenants (ether tenants only or farmers that combne farmng ther own and rented land) are ether mnfundstas or small landholders. Gven land concentraton n the hands of large landowners, t s the land poor who usually engage n the rental market. Although we lack data on whom the tenants n the sample rented from, the fndngs about land segmentaton of the studes already mentoned lead us to expect that these tenants lkely rented from small farmers. The Model The model we developed has ts roots n the agrcultural development lterature (Bardhan and Udry, 1999; Feder and Feeny, 1993; Fnan et al., 2005) and has been adapted to ft the stuaton faced by a representatve farmer n Ecuador as observed n our statstcal analyss secton. The purpose of ths model s to show the effect that an addtonal unt of land would have on total household ncome. Ths model llustrates the optmzaton process of an average agrcultural household when choosng productve nputs. Three market mperfectons are consdered here: ncomplete land markets, dependency of formal credt on land wealth, and unemployment. The frst market mperfecton reflects the fact that -as argued n the prevous secton- land markets n Ecuador are segmented, hence demand for land by the poor or margnalzed segments can only be partally satsfed. Under these condtons, land purchases by the average household and even access to land va the rental market can be regarded as unmportant for the purpose of ths model. 29

Therefore, followng Fnan et al. (2005), we gnore land transactons and consder land as exogenous. The second market mperfecton occurs because costs of nformaton and [consequently] of repayment enforcement lmt the ablty of nsttutonal credt to reach small farmers (Feder et al., 1988). As a result, these farmers are ratoned n the formal credt market, whle ths s generally not the case for large farmers. Hence, dfferent from perfect captal markets, borrower rsk s not necessarly the cause of borrower rejecton (snce nsttutonal lenders do not have nformaton on how rsky a specfc small farmer s); nstead, the borrower s farm sze s. Ths s supported by the fndngs n our statstcal analyss secton whch show that there s a statstcally sgnfcant relatonshp between farm sze and formal credt access and that the mean dollar amount of formal credt s sgnfcantly hgher than the mean dollar amount of nformal credt. Informal credt s thus gnored n ths secton. Fnally, unemployment s also a crucal problem n Ecuador. In 1998-1999 natonal unemployment reached between 11 and 14% 15 (Insttuto Naconal de Estadístcas y Censos, INEC), among the hghest rates n Latn Amerca and the Carbbean (ECLAC, 2005: Table 1.2.17). Ths s an mportant constrant for household ncome maxmzaton snce, together wth ncomplete land markets, unemployment produces neffcences n the allocaton of famly labor, whch s reflected n low farm-labor productvty (Berry and Clne, 1979). Moreover, segmentaton of land markets s encouraged by mperfectons n the credt markets snce lmted access to captal cannot easly make up for poor households lqudty constrants enough so as to purchase larger unts of land (Carter and Salgado, 2001). 15 Although 14% was the hghest unemployment rate between 1990 and 2006, rates were n average 10% annually between 2000 and 2006 (INEC). 30

The assumptons n ths model are the followng: 1) no land transactons or land rental contracts; 2) hred and famly labor are perfect substtutes, and 3) off-farm wage equals the wage earned by hred farm labor. Two lmtng condtons are of mportance: a budget constrant and a labor market constrant. The budget or cash constrant ncludes the addtonal lmtaton that the amount of credt that can be borrowed depends on the household s total land endowment. For our model, a household s expected to maxmze returns to ts fxed assets (land and famly labor) ncome that wll be used for consumpton and savngs. Income for the agrcultural household comes manly from three sources: farm profts, off-farm wage ncome, and non-labor ncome -whch would nclude remttances, governmental transfers and ncome from nvestmentsmnus the cost of borrowng 16. Max L f, Lm, H, X pq( A, L, X ; y) wh qx + wl m + R D( A) s. t. L = L f + H wh + qx wl m + R + D( A) (2-1) L m M ; L m + L f = E Where, H 0 Q = total farm output A = land endowment E = famly labor endowment Lm = off-farm labor supply L f = on-farm famly labor H = hred labor M = maxmum amount of labor hours the labor market can accept from the household X = varable nputs y = household and farm characterstcs (human captal, farm locaton, sol qualty, etc.) 16 Total debt value cancels out n the objectve functon snce the household repays the same amount of money t receves; the only thng left to affect ts ncome s the cost of borrowng. 31

R = non-labor ncome D = amount of credt receved by the household p, q, w, = output prce, nput prce, wage rate, nterest rate The Lagrangean s represented by the followng equaton: ζ = pq( A, L, X ; y) wh qx + w[ E L μ ] + R D( A) + { w[ E L ] + R + D( A) wh qx } + μ [ M E + L ] 1 f 2 f f (2-2) Kuhn Tucker condtons: ζ L f = pq L f w μ1 w + μ 2 0; L f 0; L f [ pql w μ1w + μ 2 ] = 0 ζ = pqx q μ1 q 0; X 0; X[ pqx q μ1q] = 0 X ζ = pqh w μ1 w 0; H 0; H[ pqh w μ1w] = 0 H ζ = w[ E Lf ] + R + D( A) wh qx 0; μ1 0; μ1 w[ E L μ1 ζ = M E + L f 0; μ 2 0; μ 2[ M E + L f ] = 0 μ 2 Snce the households beng modeled are agrcultural producers, f { ] + R + D( A) wh qx} = 0 f L f and X should be (2-3) (2-4) (2-5) (2-6) (2-7) postve. Now, f H > pqh = w(1 + μ ). 0 1 From (2-3), pq = w( 1+ μ ) μ f 1 2 L pqh = pql + μ pq f 2 μ 2 = H pql f. Snce famly and hred labor are assumed equally productve (assumpton 2), then μ = 2 0. Ths mples that when H > 0 the labor constrant s not bndng ( L m < M ); n other words, the household would not be sufferng from unemployment. In such a case and assumng the fnancal constrant s bndng, 17 the ndrect (or maxmzed) household ncome functon (F) would be as follows: 17 Ths assumpton s based on the fact that credt s scarce n rural areas. 32

F = Φ( p, q, w,, A, E, R; y) * * * { wh + qx w[ E L f ] R} * * * * * (2-8) = pq( A, L, X ; y) H w X q + w[ E L f ] + R In order to observe the effect of one more unt of land on the optmal choces of nputs and consequently on total household ncome, let us obtan the total dervatve of household ncome wth respect to land. df da = pq * * * L f H X + [ pql w(1 + )] + [ pqh w(1 + )] + [ pqx q(1 + )] (2-9) A A A A f Ths s equvalent to (from the frst order condtons): df da = pq A * * * L f H X + w[ μ 1 ] + w[ μ1 ] + q[ μ1 ] (2-9) A A A Provded that μ 0, whch s the case for credt constraned households, 18 we see that 1 > an ncrease n land endowments contrbutes to household ncome both drectly through affectng total producton and ndrectly through an effect n the optmal choce of farm labor, hred labor and varable nputs. On the other hand, f = 0 pq w < μ w. From (2-3) we have μ 1w = pq w + μ 2 H H 1 L. f Replacng μ 1 w n the prevous equaton and solvng for μ 2 we get μ 2 > pqh pq L f. Ths suggests μ 0, whch mples that the labor market constrant would be bndng ( L m = M 2 > ). In other words, the household cannot send any more members to off-farm labor, therefore t must use all ts labor on-farm and nether would t need hred labor nor could t afford t. Ths result s more approprate for an average agrcultural household n Ecuador, where the labor constrant s 18 The Lagrange multpler for the budget constrant, 1 μ, represents the shadow value of captal. For households who are credt constraned, ths shadow prce s larger than the market nterest rate (Cater and Salgado, 2001). 33

very lkely to be bndng, hence lttle or no hrng s done. 19 The maxmzed household ncome would n ths case be (agan assumng μ > 1 0 ): F = Λ( p, q, w,, A, E, K, y) * * * = pq( A, L f, X ; y) q[1 + ] X + w[1 + ] M + [1 + ] R (2-10) The effect of an ncrease n the land endowment would be df da = pq A + pq L * L f A + * X [ (2-11) { pqx q 1+ ] } A pq A * * L f X + [ w(1 + μ 1 ) μ 2 ] + q[ μ1 ] A A From the famly labor frst order condton we can see that f μ = 1 0 and μ > 0 2 μ 2 = w pq > 0 w > pq L f L f. Hence, gven the labor market restrcton, the productvty of on-farm labor s smaller than off-farm wage labor (ths s what causes no hrng to be chosen). Because of the labor market mperfecton then, even f captal markets were perfect, the shadow value of land for the household would be greater than smply the drect effect of land on farm product. Now, f both μ > 1 0 and μ 2 > 0 μ 2 = w (1 + μ1) pql > 0 w(1 + μ ) > pq f 1 L f. That s, labor productvty s smaller than the wage rate tmes one plus the shadow value of captal. Thus, one more unt of land would have an even larger effect on household ncome than f credt markets were not constraned. In Equaton 2-7 we see that one of the ways n whch one more unt of land wll contrbute to household ncome s by allowng t to relocate labor between on- and off-farm actvtes (smlar to what was observed n Table 2-5), however ths effect wll reman smaller than 19 Our data shows that less that 30% of small farmers (less than 5 hectares) dd any hrng n ther farms; less than 3% used more hred than non-remunerated labor, and less than 1% used only hred labor. 34

* L f w( 1+ μ1) as long as the labor market constrant s bndng ( μ 2 0 A > ). In other words, neffcency n labor allocaton wll contnue gven the labor market mperfecton. However, snce L m = E L, the more A ncreases, the smaller the labor market constrant for the f household (the constrant could go from bndng to non-bndng) and the closer pq to w 1+ μ ). Hence, ncreases n the land endowment should mprove labor allocaton and, L f ( 1 consequently, productvty (assumng that farm labor sklls are standard for all famly members). Our results therefore suggest that mperfectons n the credt, land and labor markets affect farm labor and nput allocaton, nducng low productvty and hence worsenng rural poverty. In turn, a poor, low productvty farmer has lmted access to credt and land markets. Snce ths type of producer makes up the great majorty n rural Ecuador, nequalty n the ownershp of land (and lack of access to land n general) s therefore strengthened and so s poverty. Addtonal land alone would not necessarly turn smallholders nto effcent, modern entrepreneurs as many land reform experences across Latn Amerca have shown. In order to rase agrcultural ncomes, reforms n complementary markets are essental. In the next secton, we proceed to estmate the effect of farm sze, credt and labor on household ncome per capta. Household Income Per Capta Estmaton From the prevous sectons we see that land s expected to affect household ncome both drectly and ndrectly through ts effect on credt access, labor and nput allocaton. Hence, we need to deal wth an endogenety problem when attemptng to estmate the effect of these varables together wth farm sze on household ncome. In order to perform a consstent estmaton, two-stage least squares (2SLS) estmaton s used, where a credt regresson s 35