‐‐‐‐‐ Drivers of Change in Ghana’s Cocoa Sector Ghana Strategy Support Program (GSSP)   Marcella Vigneri September 15, 2008 IFPRI - ACCRA Ghana Strategy Support Program Postal Address: c/o International Water Management Institute (IWMI) PMB CT 112, Cantonments, Accra, Ghana Local Address: CSIR Campus (Opposite Chinese Embassy) Airport Residential Area Tel: +233-(0)21-780716 Fax: +233-(0)21-784752  IFPRI HEADQUARTERS Postal Address: 2033 K Street NW Washington, DC 20006-1002 USA Tel: +1-202-862-5600 Fax: +1-202-467-4439  http://www.ifpri.org/themes/gssp/gssp.htm    www.ifpri.org For further information: Shashi Kolavalli, Senior Research Fellow and Program Leader s.kolavalli@cgiar.org     GSSP Background Paper 13  THE GHANA STRATEGY SUPPORT PROGRAM (GSSP) BACKGROUND PAPERS ABOUT GSSP IFPRI’s Ghana Strategy Support Program (GSSP) was launched in 2005 to address specific knowledge gaps concerning agricultural and rural development strategy implementation, to improve the data and knowledge base for applied policy analysis, and to strengthen the national capacity for practical applied policy research. The primary objective of the Ghana Strategy Support Program is to build the capabilities of researchers, administrators, policymakers, and members of civil society in Ghana to develop and implement agricultural and rural development strategies. Through collaborative research, communication, and capacity-strengthening activities and with core funding from the U.S. Agency for International Development/Ghana (USAID), GSSP works with its stakeholders to generate information, improve dialogue, and sharpen decision-making processes around the formulation and implementation of development strategies. ABOUT THESE BACKGROUND PAPERS The Ghana Strategy Support Program (GSSP) Background Papers contain preliminary material and research results from IFPRI and/or its partners in Ghana. The papers have not been subject to a formal peer review. They are circulated in order to stimulate discussion and critical comment. The opinions are those of the authors and do not necessarily reflect those of their home institutions or supporting organizations. Drivers of Change in Ghana’s Cocoa Sector   Marcella Vigneri1            September 15, 2008 1 Research Fellow, Protected Livelihoods and Agricultural Growth Program, Overseas Development Institute (m.vigneri@odi.org.uk) 4 Copyright © 2009 International Food Policy Research Institute. All rights reserved. Sections of this material may be reproduced for personal and not-for-profit use without the express written permission of but with acknowledgment to IFPRI. To reproduce the material contained herein for profit or commercial use requires express written permission. To obtain permission, contact the Communications Division at ifpri-copyright@cgiar.or 5 SUMMARY Using a two year panel dataset, this paper offers an empirical investigation of the unprecedented production boom episode observed in Ghana’s cocoa sector between 2002 and 2004. We look at the technology of production underlying the years of the boom, and suggest that most of the rise was due to a more intensive use of household labor, good weather, and to some extent the increased use of fertilizer. The drivers of the recent production boom in Ghana did not, however, alter the more fundamental and long-standing problems of the sector. Cocoa yields in Ghana remain well behind those observed in other producing countries and the key constraint of the sector remains its lack of innovation. Current production technology in the cocoa sector is labor-using and land saving, whereas it is labor that is scarce to the household. This poses more serious concerns about the incentives and policies required to promote and sustain technological innovation in the sector. 6 TABLE OF CONTENTS 1.  Introduction ............................................................................................................................. 7  2.  The Role of Cocoa in Ghana’s economy ................................................................................. 9  3.   Cocoa Production in Ghana ................................................................................................... 11  “The Golden Pod”: How is cocoa produced? ........................................................................... 11  Tenure ....................................................................................................................................... 11  Labor ......................................................................................................................................... 12  Non-Labor Inputs ...................................................................................................................... 13  Post-harvest activities ............................................................................................................... 15  4.  The Micro- and Macro-Dimensions of Cocoa Production .................................................... 16  Ghana’s Cocoa Farmers: A Changing Profile? ......................................................................... 16  Is the macro evidence consistent with the micro? .................................................................... 17  5.  The Analytical Approach...................................................................................................... 19  Regional Trends in Cocoa Production ...................................................................................... 20  Cocoa Production Function Analysis........................................................................................ 24  6.  Land Size and Factor Productivity: Inverse Relationship and Allocative Inefficiency? ....... 28  The inverse relationship between farm size and yields ............................................................ 28  7.  Conclusions ........................................................................................................................... 32  References ..................................................................................................................................... 33  7 1. Introduction This paper examines the sources of a major growth episode in Ghana’s cocoa production that occurred between 2002 and 2004.2 The aim of the study is threefold. First, the paper aims to put this major production boom episode in the context of the more general trends observed in the sector over a longer time period. Secondly, it identifies the determinants of the production change, differentiating the share of production increase attributed to technology change from that attributed to land productivity. Thirdly, based on these findings it speculates on whether this growth episode is a result of idiosyncratic external and domestic policy conditions that occurred over the time period considered, or is rather the outcome of more structural changes in production and productivity-enhancing measures adopted by farmers. We identify four possible sources of cocoa production change: 1. The technology of production. This includes both cultivation constraints such as climate and type of land available for cultivation and the intensity of input use. 2. The institutional arrangement defining the right to use land for the cultivation of the crop. The prevailing customary land rights and a land tenure system which both have important implications for farmers’ land investment choices such as replanting, rehabilitation, and expansion decisions. 3. The national policy framework. In particular, this includes the marketing arrangement, the Government support policy and incentive programs – to include input subsidies, and mass spraying interventions. 2 Throughout the discussion the calendar years mentioned refer to the second half of the crop year for which total production refers to. For example 2002 refers to crop year 2001/02, 2004 refers to crop year 2003/04. 8 4. The structure of the international market. We look particularly at Ghana’s role in the global supply of cocoa, and the share of the free on board (f.o.b.) price that it commands. Moreover, the trends in the world cocoa price and how these are reflected in the producer price setting regulated by the state owned marketing company, the Ghana Cocoa Board (COCOBOD). Although all four sources described above are jointly relevant to explain the dynamics of the sector, the analytical focus of this study is on the first source of change listed above: the technology of production. The rest of the paper is organized as follows. Section II begins with a general overview of the macroeconomic trends of cocoa production in Ghana to describe the sector developments in the years leading to the boom episode. Section III briefly describes how cocoa is produced through: i. the use of land; ii. the relative contribution of labor and non-labor inputs to production; iii. the role of land tenure incentives; and iv. a brief account of how the crop is traded beyond the farm gate. Section IV provides a preliminary inspection of the trends observed in aggregate cocoa production-related variables in order to illustrate how the changes in the micro data mirror the changes in the macro data. Section V then discusses the findings of production functions estimating the key determinants of growth in the period investigated. Section VI discusses labor and non-labor inputs productivity and their contribution to the production growth episode. A discussion of the key findings and their policy implications concludes the paper in section VII. The data used in this paper is the Ghana Cocoa Farmers Survey (GCFS), a comprehensive farmer and farm level data set featuring a baseline run in 2002 and a follow up component in 2004. The GCFS covers modules on production, land and sales, and a detailed 9 section on farmers’ profile. Its central value rests in the panel element linking 428 farmers between the two years (see Teal, et al. 2006). 2. The Role of Cocoa in Ghana’s economy In a recent paper on the role of cocoa in Ghana’s future development (Breisinger et al. 2007), the recent performance of the sector is described as an example of what “favorable external conditions and internal reforms” can do to renovate traditional exports. Ghana has maintained over time a leading position among cocoa producing countries and exporters, despite the noted concern that its continued dependence on traditional export crops might lock the economy in a dependency trap from raw commodities (of which cocoa makes up the bulk of the country foreign exchange earnings together with gold and timber). Serious concerns also arise over the future sustainability of the sector, as recent research findings clearly indicate that past and present cocoa growth episodes have been driven by land expansion and by the intensive use of labor, rather than by a rise in land productivity (Gockowski 2007; Vigneri 2005). Further, while the state-owned marketing board, COCOBOD, has been given some credit for its role in the recent surge in cocoa production, there are contradictory indications that the rise in cocoa prices since 2000 has played a central role in stimulating farmers’ production incentives. There is no doubt that such international and domestic conditions contributed to the trends observed in the sector in more recent years. These include higher producer prices due to international prices and the partial liberalization of the internal marketing system and a number of government-sponsored rehabilitation and spraying programs, fertilizer credits, and the privatization of input distribution. However, whether these conditions facilitated farmers’ adoption of production and yield-enhancing practices and led to a doubling of production volumes over three crop years remains an open question that this study proposes to address. 10 11 3. Cocoa Production in Ghana “The Golden Pod”: How is cocoa produced? Cocoa is a perennial tree crop with a life-cycle of twenty-five to thirty years. In the initial stage of land cultivation (two to three years from planting depending on the tree variety) it is intercropped with staple food crops like maize, plantain, cassava and cocoyam, which shade the young trees until they grow and form a closed canopy, at which point they are left to stand alone. Cocoa trees typically take between three to five years from plantation before they start bearing the first crop, and full production capacity is only reached after ten years from first planting.3 The ideal climatic zone under which the crop is grown is the tropical rainforest zone. Cocoa trees grow under shaded conditions with a climate characterized by relatively high temperatures (between 18-21 and 30-32 degrees C) and plentiful rainfall. Cocoa production also depends heavily on the pattern of rainfall; the average distribution of monthly rains throughout the year is more important than the annual total. Dry spells where rainfall is less than 100mm per month should not exceed three months. Annual rainfall in excess of 2,500mm may lead to a higher incidence of fungus diseases, the most common known as the Phytophtora pod rot which causes the black pod disease, and the cocoa swollen shoot virus (ICCO, 2000; Wood and Lass, 1985). Tenure Under the existing land tenure system in Ghana there is a conceptual separation between the land and what grows on it: it may not be difficult to access land for farming but actual ownership rights cannot be easily acquired outside kinship relations. Sharecropping contracts 3 Although some new hybrid varieties may reach full bearing as early as five years from planting. 12 used to be common among migrant farmers, for whom tree cultivation implies acquiring what amounts to permanent tenure. This implies that cultivators have a de facto ownership of the land, which only lapses back to the owner when land remains unused or the migrant dies (Berry, 1993; Quisumbing et al., 2001). In Ghana there are two predominant sharecropping contracts: these are called abunu (division into two parts) - and abusa (division into three parts). The abusa tenants typically manage already-established cocoa farms and perform tasks such as weeding, spraying and harvesting. In return for these services they are “paid” a share contribution of one third of the total cocoa harvested. Under the abunu contract, tenants are responsible for all farm tasks, from clearing land to harvesting, and receive one half of the harvest as part of the contract agreement. Labor Family members, especially spouses and children, represent the most important source of labor in cocoa farming. There is a gender difference in the assignment of tasks: male labor is required for clearing and tree felling. Female labor is used for less physically demanding tasks such as weeding and harvesting. Annual labor is a comparatively cheaper way to maintain a farm, as payment can be deferred until harvest. Yet, the precarious state of farmers’ finances means that many have become reluctant to enter into such contracts, and it is daily wage contracts that are the most frequently used. Labor is contracted for specific tasks such as weeding and preparing new farms and, although the payments demanded by laborers are higher than the minimum wages, daily wage contracts allow greater contractual flexibility, as farmers are able to hire labor as long as they are able to pay the going wage rate. 13 Non-Labor Inputs The most important non-labor inputs in cocoa production are fertilizer, insecticide, and seedlings for cocoa replanting. Until 1993, input supply was heavily subsidized by the marketing board on credit schemes (Shepherd and Farolfi, 1999), but this system often worked inefficiently due to inadequate and delayed provisions which regularly failed to reach the remotest areas of cocoa production (due to the very poor or non-existing road networks). Since the complete removal of input subsidies in 1996, the private sector has taken over input supply (Agrisystem, 1997). Table 1 below illustrates household level data from 1992 to 2004 collated from the Ghana Living Standard Survey and the Ghana Cocoa Farmer Survey. These figures clearly show an unambiguous increase in the quantity of fertilizer and insecticide used by cocoa farmers, though with a variable trend in adoption rates. Between 1992 and 1997 there was a lower application rate of fertilizer (with the exception of Western region) but a higher frequency of farmers using it (with the exception of Ashanti). The reverse trend occurred between 1997 and 2001: more fertilizer was applied across the surveyed areas, though by a lower percentage of farmers (a likely consequence of the complete removal of input subsidies). In the time interval between 2001 and 2004 the amount of fertilizer used as reported in survey data has increased by a factor of nine, and adoption rates have increased by 37 percentage points. The data collected in 2002 and 2004 (on which the core analysis of this paper is based) also show that over this three year period farmers have visibly improved their crop protection practices: on average farmers have made more than four annual 14 insecticide spraying applications, a higher frequency than what is recommended. About half of these applications were carried out under the “mass sprayings” organized by COCOBOD. TABLE 1 QUANTITIES OF INSECTICIDE AND FERTILIZER APPLIED (1990/91 –2003/2004) Crop Year Ashanti Brong Ahafotern† Total No of farmers 1990/91 112 71 137 320 1997/98 132 54 227 413 2001/02 108 94 226 428 2003/04 108 94 226 428 Fertilizer (50kg bags) 1990/91 0.28 0.13 0.03 0.14 (Adoption rate %) (13%) (8%) (6%) (9%) 1997/98 0.10 0.06 0.10 0.09 (Adoption rate %) (10%) (13%) (19%) (15%) 2001/02 0.35 0.17 0.74 0.52 (Adoption rate %) (5%) (7%) (12%) (9%) 2003/04 4.17 4.39 6.10 5.24 (Adoption rate %) (57%) (52%) (41%) (47%) Insecticide (litres) 1990/91 2.19 0.62 1.88 1.71 (Adoption rate %) (59%) (14%) (44%) (42%) 1997/98 1.66 0.98 3.76 2.73 (Adoption rate %) (39%) (25%) (64%) (51%) 2001/02 5.83 15.99 12.93 11.81 (Adoption rate %) (65%) (64%) (78%) (73%) 2003/04 7.84 5.54 11.47 9.13 (Adoption rate %) (96%) (94%) (94%) (94%) Author’s calculations from GLSS3, GLSS4, GCFS 2002, and GCFS 2004. †The two Western regions discussed in the rest of the paper are aggregated in this table for comparability with the GLSS data. Since 1999 the government developed a strategy to increase production from nearly 335,000 tons to about 500,000 tons by 2004-2005. One of the key elements of the strategy was the “Hi Tech” input program initiated in 2003. 4 Under this program COCOBOD promoted the use of fertilizers through a subsidy scheme which has now been taken up by the private sector.5 4 The Hi Tech Programme, initiated in 2003, involved the application of fertilizers, improved planting materials and the application of insecticides and fungicides on cocoa farms (Quartey, 15 Post-harvest activities As the annual distribution of rainfall is typically very uneven in coastal West Africa, planting must be done during the wet season, growth is often discontinuous, and cropping is seasonal. Cocoa trees typically have two harvest seasons in the year, the main crop (which begins in October and ends in March) and the smaller or mid crop season (between May and August). After fermentation and drying (which typically takes between 8-10 days), the beans are collected into mini or maxi bags of 30 kgs and 62.5 kgs respectively, and are sold to local buying agents who are distributed throughout the cocoa growing regions. They are then weighed, graded, and bought at prices fixed by COCOBOD. 2007). Under this program nearly all farms were sprayed in 2003/04 (up from 81 percent in 2002). The GCFS indicates that farmers reported an average of 4.22 spraying applications, of which 46 percent were carried out by the government. This program encouraged cocoa farmers to apply fertilizers to a minimum of two bags per acre. The fertilizers under this program were supplied on credit to the beneficiary cocoa farmers in the initial stages. Payments were to be made during the ensuing harvesting season by installments. Unfortunately, this policy intervention though has enjoyed maximum participation from the farmers, was bedeviled with high indebtedness from the beneficiary farmers. The program has now been repackaged and only farmers in a cooperative society or an association can benefit from the credit distribution after payment of an initial deposit. 5 In 2006 a private fertilizer company, Wienco (Gh) Limited, established a scheme for cocoa farmers called Cocoa Abrabopa to follow the principle underlying the former COCOBOD- sponsored Hi Tech Programme. Cocoa Abrabopa is an association of cocoa farmers who receive a package of high tech cocoa inputs sufficient to cover two acres of mature cocoa farm on credit with the promise to repay the total amount of the credit facility after harvest. The objective of the initiative is to target up to 50,000 farmers handling 300,000 acres, which will have a massive impact on Ghana’s cocoa industry. 16 4. The Micro- and Macro-Dimensions of Cocoa Production Ghana’s Cocoa Farmers: A Changing Profile? Cocoa is characterized by a production technology requiring the use of working capital mainly to hire labor for clearing and weeding the land, and to purchase the chemicals needed to spray cocoa farms for the control of pests and diseases. A recurrent feature in all descriptions of the post-reform scenario in rural Ghana is the movement of young people out of the cocoa growing areas, resulting in an increase in the average age, and a general decrease in the education level of the farmers. Table 2 illustrates some characteristics of the current profile of cocoa-farmers. On average, there is a clear dominance of male producers, in their early fifties, and with on average six years of education (so having just completed primary school). The average size of cocoa farms – reported above in its median value – has remained constant over the period 2002 to 2004, ranging around three hectares. One critical change that is observed in the data is the large reduction in the share of hired labor on cocoa farms. This decreased by 17 percentage points in the survey data employed in the present analysis, a point to which we return below. The data also show a high percentage of revenue from sales of the crop in cocoa- households’ income. This proportion suggests that cocoa production remains the major source of income for farmers. The revenue from cocoa sales can increase through two channels: (i) an increase in the selling price of cocoa (induced by national or international policy interventions affecting the sector), (ii) an increase in the amount of cocoa produced, which farmers can generate – in the short term – by varying the intensity of cultivation on existing cocoa farms, or – in the longer term – by clearing new land for the cultivation of cocoa trees. 17 TABLE 2 CHARACTERISTICS OF COCOA-FARMING HOUSEHOLDS Characteristics 2002 2004 HH head Gender (% male) 0.82 0.81 (0.38) (0.39) HH head Age 51 53 (14.65) (14.86) School years completed by HH Head 6.64 6.38 (4.72) (5.01) Household size 6.94 5.74 (2.62) (2.47) Farm size^ (ha) 3.44 2.66 (4.93) (6.18) % Of hired labor 0.56 0.39 % Revenue from cocoa (0.32) (0.27) 0.75 (0.22) 0.79 (0.21) Producer Price Changes* (Cedis/Kg.) Producer Prices (nominal) 6,200 9,000 Producer Prices (real) 6,200 6,428 Notes: Source is GCFS, 2002 and 2004. ^These are median values derived from full bearing cocoa farms only. *Real prices are derived using CPI 2001=100. Standard errors in parentheses. In Table 2 producer prices are shown to have increased in nominal terms; however the real increase in purchasing power terms was only 3.67 percent. We now look at how production inputs might have been used more intensively to engineer the production increase described above. Is the macro evidence consistent with the micro? Table 3 describes the sector trends in the decade from 1994 to 2004 by looking at three related agricultural macroeconomic indicators: production levels, land harvested, and cocoa yields. 18 TABLE 3 COCOA PRODUCTION, YIELD AND AREA HARVESTED6 Crop Year Production (metric tonnes) Area (hectares) Yields (Kg/ha) 1994 288,075 686,531 419.61 1995 403,900 1,000,000 403.90 1996 403,000 1,050,000 383.81 1997 322,490 1,074,970 300.00 1998 409,360 1,364,530 300.00 1999 434,200 1,300,000 334.00 2000 436,600 1,500,000 291.07 2001 389,591 1,350,000 288.59 2002 340,562 1,195,000 284.99 2003 497,000 1,500,000 331.33 2004 737,000 2,000,000 368.50 Tot change in decade 1.56 1.91 -0.12 Tot change in 2002 to 2004 1.16 0.67 0.29 Source: FAOSTAT Database The data, obtained from FAO official statistics, show an increase in both total area under cocoa cultivation (191 percent) and in the level of production (156 percent) between 1993/94 and 2003/04, while the figure on cocoa yields points to a 12 percent drop in the relevant indicator over the same ten year period. If we narrow the analysis to the two years of the production boom, from 2002 to 2004, the growth rate in the volume alone was 116 percent, and just twenty five percent of this could have been potentially imputed to increases in yields (given that the reported aggregate statistics marks a 29 percent increase in land productivity over the period). Although accurate data on land under cocoa cultivation is not readily available, the production increase occurred over the longer time interval is considered to be largely a result of the expansion of the area cultivated. In the Ghanaian case this would have happened by means of 6 Crop production figures include the quantities of the commodity sold in the market (marketed production). When the production data available refers to a production period falling into two successive calendar years and it is not possible to allocate the relative production to each of them, the production data refers to the year in which the bulk of the production falls (http://www.fao.org/waicent/faostat/agricult/pr_ele-e.htm). This rule applies to cocoa. As explained earlier, the production period is divided into two seasons, with the bulk of the harvest being sold between September and February of any given crop year. We therefore calculated the average across two calendar years in order to make a meaningful comparison with the GLSS household data.   19 the westward extension towards unoccupied virgin forests of Western and Ashanti regions, the two areas that are still considered suitable for their production potential as of 2003/04 (Breisinger, 2008). What other different patterns of production expansion can explain the short term increase observed in the 2002-2004 period? 5. The Analytical Approach To address this issue this study estimates a standard Cobb-Douglas production functions with three conventional inputs: land, labor, and non-labor inputs (fertilizer, insecticide, and agricultural equipment), a number of household characteristics which we expect to have important effects on the level of cocoa production, and rainfall. 1 2 3 4 5 6 7) ) ) ) )ln(cocoa = ln(farmsize + ln(Input + ln(Labour + farmer age+ farmer sex+ ln(rain + Tβ β β β β β β γ+ Where: cocoa = Kilos of cocoa produced farm size = Total hectares of cocoa farms cultivated by each household Input = Amount of non labor input use (differentiate in regression between fertilizer and insecticide) Labor = Man-days of labor (both household and hired) Farmer age = Age of farmer in years Male farmer = Dummy =1 if household head is male Rain = District level amount of rainfall (measured as annualised monthly millimetres of precipitation) T = Time trend equals 1 for crop year 2004 (to proxy for total factor productivity) The functional form in the expression above is used to estimate by OLS the determinants of the volume of cocoa production, first for each year separately, and then pooled across time (column 3) with a time dummy to proxy for total factor productivity (TFP). In order to remove the potential bias of OLS parameters due to the effect of unobserved characteristics of the 20 variable inputs (for example effort of labor, and land quality), the pooled production function is also estimated using a fixed effect model. Finally, the last column shows the results of a Two Stage Least Squares regression which was run to attempt a more accurate estimate of the contribution of fertilizer to production as explained in detail further below. Regional Trends in Cocoa Production We begin with a description of the regional distribution of production related indicators to contrast the two periods under investigation. Table 4 presents the key variables relevant for cocoa production analysis in levels and in their logarithmic transformation. Because few large outliers tend to dominate their distribution, the changes observed over time are discussed based on the logarithmical data (as seen in the second half of the table). The far right column shows the statistical changes across the regions: (i) average production has increased by 31 percent, (ii) average size of land holdings has increased by 21 percent, (iii) cocoa yields on average have increased by 8 percent, (iv) insecticide use has risen by 27 percent, and (v) fertilizer inputs have increased nearly ten times fold (and largely by means of higher fertilizer applications). The latter (and to a much lesser extent labor change) is the single most important element of growth in variable production inputs observed between 2002 and 2004. The total averages above mask substantial inter-regional variation. Output increase in the area sampled has primarily occurred in the Western Sefwi region, with Western Wassa and Brong Ahafo accounting for a much lower share (about one third) of this production growth. Yields have risen prominently in the Sefwi sub-region of Western, and significantly decreased in Ashanti (the traditional cocoa area sampled), while remaining constant elsewhere. Total labor use increased significantly across all regions (in the aggregate by 213 percent), while the hired 21 labor component has fallen by 18 percentage points. In looking at the composition of labor input (top half of the table), the large increase in the household component dominates and drives the increase in the combined variable. The rise in output and the large increase in labor input explain the large decrease in labor productivity of 58 percent (with regional level statistics uniformly confirming this downward trend). The intensity of input use was variable across regions, as discussed in Section 3. 22 TABLE 4 REGIONAL MEANS OF COCOA PRODUCTION VARIABLES Data in Levels N. observations /yr Ashanti 108 Brong Ahafo 94 West. Sefwi 107 West. Wassa 119 Total 428 Cocoa harvested (kg) 2002 1,037 1,004 1,625 1,388 1,272 2004 1,070 1,216 2,746 1,698 1,725 % change 0.03 0.21 0.69 0.22 0.36 Cocoa farm size (ha)a 2002 3.64 4.05 6.48 4.26 4.45 2004 4.13 4.90 7.08 4.96 5.38 % change 0.13 0.21 0.09 0.16 0.21 Cocoa yield (kg/ha) a 2002 182.75 154.44 154.44 196.62 180.18 2004 178.82 164.78 236.80 206.19 205.92 % change -0.02 0.07 0.53 0.05 0.14 Insecticide 2002 6.09 15.74 12.07 13.55 11.78 (litres) 2004 8.12 5.64 12.19 10.56 9.31 % change 0.33 -0.64 0.01 -0.22 -0.21 Fertilizer 2002 16.74 8.11 36.40 36.48 25.14 (Kilos) 2004 214.08 219.01 258.53 337.97 262.99 % change 11.79 26.00 6.10 8.27 9.46 Tot lab days 2002 268.53 354.66 416.13 272.64 325.29 (Yearly person/days) 2004 788.20 586.21 858.34 648.66 718.41 % change 1.94 0.65 1.06 1.38 1.21 Household lab days 2002 84.14 98.19 109.07 89.35 94.91 (Yearly person /days) 2004 350.53 361.09 384.85 352.29 361.92 % change 3.17 2.68 2.53 2.94 2.81 % of hired labor 2002 0.55 0.49 0.60 0.59 0.56 2004 0.39 0.35 0.42 0.36 0.38 % change -0.16 -0.14 -0.18 -0.23 -0.18 Household labor 2002 86.47 100.89 108.83 87.55 95.50 (Yearly person /days) 2004 367.33 353.68 385.98 348.10 363.38 % change 3.25 2.51 2.55 2.98 2.81 Labor productivity 2002 7.66 10.17 8.13 9.43 8.82 (kg cocoa/ person-days) 2004 2.67 3.83 6.02 3.77 4.14 % change -0.65 -0.62 -0.26 -0.60 -0.53 Insecticide/ha 2002 1.80 1.98 1.77 2.50 2.02 2004 2.32 1.34 1.60 1.57 1.68 % change 0.29 -0.32 -0.10 -0.37 -0.17 Fertilizer/ha 2002 5.24 1.94 6.07 3.69 4.28 2004 52.16 36.44 25.44 40.92 38.23 % change 8.95 17.81 3.19 10.10 7.93 Person-days lab./ha 2002 77.67 65.19 54.83 50.03 61.61 2004 216.25 139.86 133.83 132.09 152.28 % change 1.78 1.15 1.44 1.64 1.47 23 Table 4 Cont’d Data in Logarithmic b Ashanti Brong Ahafo W. Sefwi W. Wassa Total Cocoa harvested (kg) 2002 6.48 6.45 6.90 6.79 6.66 2004 6.44 6.62 7.52 6.99 6.93 % change -0.03 0.19 0.86 0.23 0.31 Cocoa farm size (ha) 2002 1.25 1.41 1.84 1.45 1.48 2004 1.44 1.59 2.00 1.62 1.67 % change 0.21 0.20 0.17 0.19 0.21 Cocoa yield (kg/ha) 2002 5.23 5.04 5.06 5.34 5.17 2004 5.00 5.03 5.52 5.37 5.25 % change -0.20 -0.01 0.59 0.03 0.08 Insecticide 2002 1.33 1.49 1.79 1.98 1.66 (litres) 2004 1.84 1.66 2.14 1.92 1.90 % change 0.66 0.19 0.42 -0.06 0.27 Fertilizer 2002 0.24 0.31 0.67 0.47 0.42 (Kilos) 2004 3.37 2.98 2.24 2.73 2.80 % change 21.92 13.54 3.81 8.63 9.76 Tot lab days 2002 4.98 5.08 5.25 4.97 5.07 (Yearly person/days) 2004 6.23 5.99 6.44 6.16 6.21 % change 2.49 1.48 2.26 2.28 2.13 Labor productivity 2002 1.50 1.37 1.64 1.81 1.59 (kg cocoa/person-days) 2004 0.22 0.64 1.08 0.83 0.72 % change -0.72 -0.52 -0.43 -0.62 -0.58 Insecticide/ha 2002 0.60 0.64 0.42 0.77 0.61 2004 0.27 -0.03 0.05 0.28 0.14 % change -0.28 -0.49 -0.31 -0.39 -0.38 Fertilizer/ha 2002 3.87 3.01 2.59 3.29 3.03 2004 3.97 3.79 3.83 4.35 4.00 % change 0.11 1.16 2.46 1.89 1.64 Person-days lab./ha 2002 3.73 3.67 3.41 3.52 3.58 2004 4.79 4.40 4.44 4.54 4.53 % change 1.89 1.07 1.79 1.75 1.59 Source: Author’s calculation based on GCFS 2002 and 2004. a) Median values. b) The percentage of log-differences in the bottom half of the table is computed using the formula: Exp (log-difference)-1. In summary, a preliminary inspection of the data suggests a unique combination of changes: the increase in farmers’ cocoa output occurred in conjunction with a major rise in labor input (mainly as a result of the drop in the percentage of hired labor), a noticeable increase in the average size of land holdings, and a substantial expansion in the use of non-labor inputs. A regional breakdown of these trends suggests where the largest changes on production-related indicators occurred. 24 We also note that in the survey data the change in yields relative to the change in the volume of cocoa production mirrors the proportional changes illustrated above in the corresponding macro variables. Though in absolute terms the production increase in the farmer data is lower than the macro one seen in the aggregated statistics (a difference which is not surprising given that the GCFS is neither representative of all cocoa producing regions, nor regionally representative of the underlying cocoa farming population), yields increases in the sample are just under a third of the production increase. This similar proportional change will allow drawing generalizable inferences from the analytical findings below. Cocoa Production Function Analysis Table 5 shows the results of estimating a cocoa production function run first by Least Squares (OLS) on the two cross sections – columns [1] and [2] for 2002 and 2004 respectively - then on the pooled sample in column [3], and finally by fixed effects (FE) in column [4]. The first noticeable feature of the regression output is the decrease in the size of both land and labor coefficients between the OLS estimates and those of the fixed effect model. In the two cross sections and in the pooled OLS regression the point estimate for land is 0.53, which is statistically significant at the 1 percent level. Although we expect land to be a major factor explaining cocoa production, we suspect the OLS coefficient to be biased upwards due to the effect of unobserved land characteristics (such as land quality). The fixed effects estimator purges (by construction) the regressor from any time-invariant unobserved effect. The point estimate on land is 0.32 in column 4 (FE model), and we interpret part of this reduction as the effect of removing the time invariant effect of unobserved land quality. 25 The OLS labor parameter may be biased as well by unobserved time invariant effects (such as farmers’ ability, or the effort of hired laborers). In the fixed effects regression the marginal productivity of labor drops (and its statistical significance disappears). The interesting point of reflection is the comparison of the labor estimates across the two OLS cross sections. The results in columns [1] and [2] indicate a significant (in both size and statistical terms) contribution of labor in 2002, but not in 2004. If we take account of the qualitative implications of these coefficients, it is possible to interpret the change as a result of the substantial increase (by a factor of two as seen in table 4) in the household component of labor between the two years, which will have pushed the marginal productivity of labor beyond the efficiency point, and hence significantly decreased the contribution of labor input increase to output growth outcomes. Insecticide and farming assets are found to be both unambiguous contributing factors of production across years, with respectively a 9 and 4 percent impact on the production growth observed between 2002 and 2004 (as shown in the FE model in column 4). Favorable rainfall made the single largest contribution to production growth, whereas fertilizer appears to be wholly non-significant. Based on the discussion in Section 3, and on the massive increase in the use of fertilizer by farmers described in Table 1, this result is somehow surprising. TABLE 5 PRODUCTION FUNCTION ANALYSIS (1) (2) (3) (4) (5) (6) Dep. Var. Log of cocoa production. OLS 2002 OLS 2004 OLS Pooled Fixed Effect First Diff. 2SLS FD Full bearing farm (ha) 0.53*** 0.55*** 0.53*** 0.31*** 0.31*** 0.29*** (0.06) (0.08) (0.05) (0.08) (0.08) (0.08) Labor (man days) 0.13*** -0.02 0.08** 0.02 0.02 0.05 (0.05) (0.05) (0.03) (0.03) (0.03) (0.04) Fertilizer (Kg)T 0.02 0.13* 0.08 0.06 0.06 0.54*** (0.10) (0.07) (0.06) (0.06) (0.07) (0.25) 26 Insecticide (litres) 0.13 0.23*** 0.19*** 0.09** 0.09* 0.02 (0.08) (0.07) (0.05) (0.05) (0.05) (0.06) Farming assets^ 0.10*** 0.08*** 0.10*** 0.04* 0.04* 0.02 (0.03) (0.03) (0.02) (0.02) (0.02) (0.02) Male cocoa farmer 0.23** 0.11 0.16** (0.10) (0.11) (0.07) Cocoa farmer’s age 0.01 0.01 0.01 (0.02) (0.02) (0.01) (Cocoa farmer’s age)2 -0.00 -0.00 -0.00 (0.00) (0.00) (0.00) Cocoa Farms sprayed 0.15 0.59 0.21* 0.12 0.12 0.17 (0.12) (0.50) (0.11) (0.13) (0.13) (0.14) Rainfall (mml) 1.45*** 0.14 1.03*** 0.75*** 0.75*** 0.66*** (0.26) (0.79) (0.20) (0.17) (0.18) (0.18) Trend Dummy = 2004 -0.29*** -0.08 (0.09) (0.08) Ashanti 0.19* 0.18 0.29*** (0.11) (0.25) (0.09) Brong Ahafo 0.54*** 0.02 0.28*** (0.17) (0.20) (0.11) Western Sefwi 0.60*** 0.38 0.53*** (0.16) (0.25) (0.10) Constant -3.30** 2.89 -1.50 1.88* -0.23* (1.39) (4.13) (1.06) (.94) (0.13) Observations 428 428 856 856 428 428 R-squared 0.52 0.50 0.50 0.21 (within) 0.14 (overall) 0.14 Hansen J test (Ho: Validity of instruments): 0.31 P-value 0.56 Hausman Test: H0 OLS efficient against IV 0.97 Partial R2 of excluded. instruments: 0.05 F Test of excluded instruments: 10.73 P-value 0.000 Notes: All variables are in logs. Dummy variables to control for farmers not using inputs (fertilizer, insecticide, agric equipment) where used in all regressions but are not reported. Western Wassa is the reference group for regional dummies. Robust standard errors are in parentheses. Statistical significance levels are marked as follows: * significant at 1%; ** significant at 5%; *** significant at 10%. ^These include cocoa farming equipment such as: drying mats, boots, cutlasses, knapsack sprayers. T Instruments used for fertilizer were: 1) Whether farmer has bank account (as a proxy for being credit constrained); 2) number of spraying applications as a proxy for good farming practices. The statistical properties of the key variables in the production functions are reported in table A.1 in the annex. Table A.2 in the annex report the first stage regression results. To further probe the effect of fertilizer associated with the OLS and FE parameter estimate, we have run an instrumental variable estimation (reported in column 6) on the first difference (FD) model reported in column [5], which mirrors the results of the FE model. The two variables used to instrument7 fertilizer were: 1. the number of spraying applications (to 7 The first stage regression is not reported, but is available from the author. Table 5 reports a number of diagnostic tests to confirm the robustness of the instrument variables selected. 27 proxy good farming practices); and ii. whether farmers have a bank account (to proxy for cash constrains). Column [6] shows the outcome of the procedure, the parameter estimate of fertilizer is now 0.54 and statistically significant, therefore suggesting a much more prominent contribution of the input to production once farming practices and financial constrains faced by farmers are accounted for. In sum we would argue that the fixed effect results provide a suggestive indication of the contribution of inputs to production growth. Land, insecticide, and rainfall have all sizeable effects, whereas the contribution of labor is found to be negligible. Fertilizer use, which has dramatically increased over the period investigated, is significant only once farming practices and credit constraints identify its positive effect when used efficiently by farmers. We have showed this to be the case by instrumenting the effect of fertilizer with a 2SLS model. Finally, no evidence was found of any rise in TFP – as suggested by the time trend dummy used to proxy for it in the FE model - which suggests the absence of technical innovation as an engine of production growth. 28 6. Land Size and Factor Productivity: Inverse Relationship and Allocative Inefficiency? This last section looks at the relationship between land size and productivity to determine if and how cocoa production differs across small and large farmers. The inverse relationship between farm size and yields If cocoa output per hectare (yields) and labor per hectare are related through a significant inverse relationship (IR) to farm size, this would suggest higher levels of input productivity on smaller landholdings. If supported by the data, this finding would have in turn important policy implications on how best to promote an efficient allocation of inputs in Ghana’s cocoa production. The question of an inverse relationship between land size and yields has been extensively looked at in the economic literature, especially as an observed outcome of missing or imperfectly functioning markets of agricultural inputs (Berry, 1979; Bhalla and Roy, 1988; Carter, 1984; De Janvry, 1981; Sen, 1975). This paper looks at the inefficiency in resource allocation and at the unobserved effect of land quality as possible causes for the negative relationship. Table 6 shows the results of estimating the yield regression in a similar fashion to what was done for the production functions in levels: first by OLS for each year separately, and aggregating across yearly cross section (columns [1] to [3]), then by FE (column [4]). Since the findings from yield regressions mirror those in the production function (of which the yield regression represents the intensive specification), the comments below focus on the coefficient on land size, on which the discussion on the existence of an inverse relationship is centered. 29 TABLE 6 COCOA YIELDS AND THE INVERSE RELATIONSHIP Dep. Var. Log of cocoa yields (1) OLS – 2002 (2) OLS - 2004 (3) OLS - Pooled (4) FE - Pooled ln (cocoa farm size) -0.20*** -0.21*** -0.16*** -0.45*** (0.06) (0.07) (0.05) (0.10) ln (person days/ha) 0.15*** -0.07 0.06 0.01 (0.06) (0.05) (0.04) (0.04) ln (litres fertilizer/ha) 0.18* 0.15* 0.14** 0.04 (0.09) (0.08) (0.06) (0.09) ln (litres insecticide/ha) 0.18 0.36*** 0.29*** 0.16** (0.11) (0.10) (0.07) (0.07) ln (farming assets/ha) 0.11*** 0.08** 0.11*** 0.05* (0.04) (0.03) (0.02) (0.03) Male farmer 0.09 0.10 0.08 0.00 (0.11) (0.11) (0.08) (0.00) Age farmer 0.03 -0.00 0.02 0.26** (0.02) (0.02) (0.01) (0.10) (Age farmer)2 -0.00 -0.00 -0.00 -0.00*** (0.00) (0.00) (0.00) (0.00) Ashanti -0.07 -0.28 0.04 0.00 (0.15) (0.26) (0.10) (0.00) Brong Ahafo 0.21 -0.29 0.05 0.00 (0.22) (0.21) (0.12) (0.00) Western Sefwi 0.43** 0.24 0.49*** 0.00 (0.18) (0.26) (0.11) (0.00) Ln (rainfall) 1.00*** -0.29 0.89*** 0.72*** (0.32) (0.82) (0.23) (0.22) Farmer sprayed 0.05 0.67 0.14 0.17 (0.13) (0.51) (0.13) (0.13) Farm affected by black pod -0.27** -0.05 -0.11* -0.11 (0.13) (0.08) (0.07) (0.08) Mean age all cocoa farms 0.03*** 0.00 0.00 -0.00 (0.01) (0.00) (0.00) (0.00) (Mean age all cocoa farms)2 -0.00** -0.00 -0.00 0.00 (0.00) (0.00) (0.00) (0.00) Y04 -0.21** 0.00 (0.10) (0.00) Constant -2.20 5.37 -1.25 -3.10 (1.65) (4.26) (1.19) (3.31) Observations 367 428 795 795 R-squared 0.32 0.28 0.26 0.27 Number of farmer id no. 428 Notes: Western Wassa is the reference group for regional dummies. Robust standard errors are in parentheses. Statistical significance levels are marked as follows: * significant at 1%; ** significant at 5%; *** significant at 10%. All three OLS estimates show a consistently negative and statistically significant coefficient on land, suggesting that yields are higher on smaller cocoa land holdings. Since (omitted) land quality is a recurrent source of bias in yield regression estimates, often accounted as driving the observed negative coefficient of land, a fixed effect model was estimated to test 30 this source of bias by exploiting the panel element of the data. Column [4] – which reports the results of the FE model - shows an even larger and negatively-signed coefficient associated with land, confirming that cocoa yields were consistently higher on relatively smaller cocoa farms over the three year period analyzed. It is now possible to combine the results of the findings above to suggest how Ghanaian cocoa farmers responded to changed incentives – both external (higher nominal price, higher share of fob) and domestic (the Hi Tech Programme sponsored by COCOBOD) – to double production between 2002 and 2004. We first noted that two key changes underlying this boom: i) a doubling of labor input – primarily driven by an increase in the household component, and ii) the dramatic rise in the use of fertilizer. Farmers have increased the amount of fertilizer used by a factor of ten, combined with a near doubling of adoption rates (though both these increases were characterized by a minimal initial uptake in 2002). In our estimation of the production function we found no evidence of labor contribution to production. This result is not surprising given the wide increase in the household component of labor on cocoa farms which will have decreased the marginal contribution (as measured by its estimated elasticity) of aggregate labor. We also found that rainfall has been a particularly favorable contributing factor to the production growth episode. The impact of fertilizer is found to be significant only after accounting for producers’ good farming practices (as done by means of instrumenting fertilizer use with the number of spraying applications, and with an indicator variable identifying farmers who were not financially constrained). We have rather clear evidence that the increase in yields associated with this production growth episode has been limited, and that TFP has not been rising. One possible reason for the lack of innovation in new 31 technology in cocoa production is that it remains labor intensive, and the cost of hired labor to the farmers remains too high. While this issue remains speculative at present, it is clear from the main analytical findings of this paper that the underlying problem facing the cocoa sector remains farmers’ lack of innovation in the production technology. 32 7. Conclusions Ghana’s cocoa sector remains a key engine of growth for the economy as a whole. Since 2000 production has increased steadily, and between 2002 and 2004 alone the volume of cocoa beans produced has doubled. The recent trends observed in the sector pose important policy questions on how this growth episode has been achieved, whether it is sustainable, and under what conditions it is replicable. This paper looks at the technology of production underlying the years of the boom, and suggests that most of the rise was due to a more intensive use of household labor, good weather, and to some extent the increased use of fertilizer. A number of favorable conditions, such as higher producer prices, the government sponsored mass spraying and the high tech program for subsidizing the use of fertilizer, have undoubtedly contributed to a more intensive use of land, which led to a 29 percent increase in yields in just three years (as observed in the macro data). However, Ghana’s cocoa yields remain far below the levels observed in other producing countries, and the gap between current levels and what can be achieved remains significant. The key constraint of the sector remains its lack of innovation. Current production technology in the cocoa sector is labor-using and land saving, whereas it is labor that is scarce to the household. Identifying under what conditions (not necessarily driven by price-output considerations) and for which farmers it is possible to obtain higher levels of land productivity is the critical issue for the future of the sector. 33 References Agrisystems. 1997. Study of the Cocoa Sector to Define Interventions on Behalf of Ghana's Smallholder with Particular Reference to the Framework of Mutual Obligations to be Prepared for the Stablex 1992 and 1993 Allocation, Aylesbury. Berry, R.A., and W.R. Cline. 1979. Agrarian structure and productivity in developing countries Johns Hopkins University Press, Baltimore. Berry, S. 1993. No condition is permanent: the social dynamics of agrarian change in sub- Saharan Africa University of Wisconsin Press, Madison; London. Bhalla, S.S., and P. Roy. 1988. "Mis-specification in farm productivity analysis: the role of land quality". Oxford Economic Papers; 40:55-73. Breisinger, C., Diao, X., Kolavalli, S., and Thurlow, J. (2008) ‘The role of cocoa in Ghana’s future development’ Ghana Strategy Support Program (GSSP). Background Paper No. GSSP 0011 Carter, M.R. 1984. "Identification of the inverse relationship between farm size and productivity: an empirical analysis of peasant agricultural production". Oxford Economic Papers 36:131-45. De Janvry, A. 1981. The agrarian question and reformism in Latin America. The Johns Hopkins University Press, Baltimore. Gockowski, J. (2007) ‘Cocoa Production Strategies and the Conservation of Globally Significant Rainforest Remnants in Ghana’. STCP-IITA mimeo ICCO. 2000. What Climate and Conditions are Necessary to Grow Cocoa Trees? [Online]. Available at http://www.icco.org/questions/tree.htm Quartey, E. (2007) ‘Cocobod Strategy’ COCOBOD mimeo, Research Department. Quisumbing, A.R., K. Otsuka, S. Suyanto, J.B. Aidoo, and E. Payongayong. 2001. "Land, Trees, and Women. Evolution of Land Tenure Institutions in Western Ghana and Sumatra". International Food Policy Research Institute, Washington D.C. Shepherd, A.W., and S. Farolfi. 1999. "Export Crop Liberalization in Africa: A Review". FAO Agricultural Services Bullettin 135. FAO, Rome. Teal, F., Zeitlin, A., and Maamah, H. (2006) ‘Ghana Cocoa Farmers Survey 2004: Report to Ghana Cocoa Board’ Centre for the Study of African Economies, Oxford University. http://www.gprg.org/pubs/reports/pdfs/2006-04-teal-zeitlin-maamah.pdf Vigneri M. (2005) ‘Trade Liberalisation and Agricultural Performance: Micro and Macro Evidence on Cash Crop Production in Sub-Saharan Africa.’ DPhil thesis. Oxford University. 34 Wood, G.A.R., and R.A. Lass. 1985. Cocoa. 4th / ed. Longman, London. Sen, A.K., and International Labour Office. 1975. Employment, technology and development: a study prepared for the International Labour Office within the framework of the World Employment Programme. Clarendon Press, Oxford. 35 ANNEX I TABLE A.1 STATISTICAL PROPERTIES OF PRODUCTION FUNCTION VARIABLES Level Variables Variables min max mean variance skewnes s kurtosi s Cocoa harvested (kg) 200 2 5.00 11,875.0 0 1,275.7 3 2,143,824.0 0 3.06 15.59 200 4 30.00 15,625.0 0 1,712.0 8 4,184,583.0 0 3.25 17.21 Cocoa farm size (ha)a 200 2 0.18 40.06 6.26 36.82 2.57 11.19 200 4 0.27 90.25 7.49 60.36 4.09 34.35 Tot lab days (person days) 200 2 4.00 6,114.00 326.35 376,388.80 5.70 45.25 200 4 100.0 0 9,941.00 715.69 816,965.50 5.18 42.25 Fertilizer (Kgs) 200 2 0.00 2,000.00 25.91 19,966.76 9.10 106.77 200 4 0.00 5,500.00 261.91 341,459.90 4.91 34.02 Insecticide (litres) 200 2 0.00 800.00 11.81 1,828.00 15.20 273.11 200 4 0.00 150.00 9.13 206.48 5.55 44.46 Logged Variables min max mean variance skewnes s kurtosi s Cocoa harvested (kg) 200 2 1.61 9.38 6.66 1.10 -0.53 4.64 200 4 3.40 9.66 6.91 1.21 -0.38 3.26 Cocoa farm size (ha)a 200 2 -1.72 3.69 1.49 0.71 -0.10 3.54 200 4 -1.31 4.50 1.65 0.72 0.03 2.96 Tot lab days (person days) 200 2 1.39 8.72 5.06 1.33 0.18 3.55 200 4.61 9.20 6.19 0.66 0.59 3.28 36 4 Fertilizer (Kgs) 200 2 0.00 7.60 0.44 2.12 3.27 12.32 200 4 0.00 8.61 2.78 8.95 0.23 1.23 Insecticide (litres) 200 2 0.00 6.69 1.65 1.59 0.20 2.72 200 4 0.00 5.02 1.88 0.76 0.32 3.91