EPTD DISCUSSION PAPER NO. 27 POPULATION, LAND TENURE, AND NATURAL RESOURCE MANAGEMENT THE C

EPTD DISCUSSION PAPER NO. 27

POPULATION, LAND TENURE, AND

NATURAL RESOURCE MANAGEMENT:

THE CASE OF CUSTOMARY LAND AREA IN MALAWI

Frank Place and Keijiro Otsuka

Environment and Production Technology Division

International Food Policy Research Institute

1200 Seventeenth Street, N.W.

Washington, D.C. 20036-3006 U.S.A.

April 1997

EPTD Discussion Papers contain preliminary material and research results, and are circulated prior to a full peer review in order to stimulate discussion and critical comment. It is expected that most Discussion Papers will eventually be published in some other form, and that their content may also be revised.

ABSTRACT

This paper uses cross section-time series data on 57 communities in Malawi to determine statistically the factors determining changes in land use, tree cover, and crop yield. The econometric model is developed from a theoretical model which also endogenizes population growth and prevailing land tenure institutions within the customary sector of Malawi. The analysis reflects changes between 1971 and 1995, utilizing aerial photos taken at these dates and complementing these with field surveys. The data show a deterioration of Malawi's natural resource base: declining yields, loss of tree cover, and near exhaustion of land for agricultural expansion. Key findings are that population pressure induces land conversion but not yield or tree cover change; the matrilocal system of household residence is negatively associated with tree cover but induces agricultural conversion; and there is some improvement in management of resources as their scarcity increases. Policy recommendations include greater focus on agroforestry to increase tree cover as woodland areas are poorly managed, and increased effort to improve market integration since this benefits crop yields without adverse effects on tree cover.

CONTENTS

1. Introduction (1)

2. Customary Land Tenure Systems in Malawi (4)

3. The Theoretical and Econometric Model (7)

4. Data Collection Methods and Description (15)

Sampling Strategy (15)

Natural Resource Management Indicators (16)

Land Use and its Change (16)

Tree Cover and its Change (19)

Crop Yield and its Change (20)

Driving Factors (21)

5. Econometric Analysis (25)

Estimation Method (25)

Results (26)

Population Growth (31)

Transmission of Land Through Women (31)

Change in Non-agricultural Land Tree Density (32)

Change in Crop Yields (33)

Change in Share of Agricultural Land (34)

6. Summary and Conclusions (35)

Appendix: Definitions of Land Use Categories and Vegetation Cover (39)

References (41)

i

The authors are from ICRAF and IFPRI, respectively. We would like to thank our colleagues in *

Malawi, Mr. Bob Green, Vincent Mkandawire and colleagues at the Ministry of Agriculture, and Ken Neils,Redge Masupayi, and colleagues at APRU in Bunda College. We also thank Raisuddin Ahmed, Romeo Bautista, Gaurav Datt, Peter Hazell, Ruth Meinzen-Dick, John Pender, Detlev Puetz, Sherman Robinson, Sara Scherr, and Towa Tachibana for helpful comments.POPULATION, LAND TENURE, AND

NATURAL RESOURCE MANAGEMENT:

THE CASE OF CUSTOMARY LAND AREA IN MALAWI

by Frank Place and Keijiro Otsuka *

1. INTRODUCTION

Malawi, as elsewhere in sub-Saharan Africa, has experienced a significant change in its landscape cover. Although reliable figures are hard to come by, the Forestry Department estimates the annual deforestation rate to be 1.3 percent per year in the 1980s (World Resources Institute, 1994). This has raised concern about the future supply of fuelwood and other tree products and environmental services (French, 1986; Hyde and Seve, 1993; Dewees,1995). Much of the deforestation is believed to be linked to conversion of miombo woodlands into agricultural land. This involves expansion onto steep slopes and other fragile lands in many cases. Bojo (1994) presents data suggesting that the consequent effects on soil erosion create serious costs to the Malawian economy.

Attaining sustainable use of woodland resources and sustainable growth of agriculture are important for Malawi because the economy is agricultural-based and of its approximately 12 million people, nearly 87 percent live in rural areas (World Bank, 1995). The population density is about 100 persons per square kilometer which is high among southern African countries and places a great pressure on agriculture which benefits from only a single rainy

season. Primary commodities still account for 94 percent of all export revenues in Malawi, though agriculture's share in GDP is about 40 percent. Growth of agriculture, recorded to be 4.4 percent in the 1970s, has fallen to 2.1 percent between 1980 and 1993 resulting in declining food production per capita.

Malawi policy makers, having few resources at their disposal, must make critical choices concerning the type of land use patterns that will prevail in the future. Unfortunately, while much information has been generated about current land use, there is little understanding of the dynamic process leading up to the current land utilization pattern nor to related effects on productivity and the stock of natural resources. This study will provide new evidence as to how communities have managed their land and tree resources and what factors seemed to be most important in their decisions. This information is valuable to policy makers who continue to struggle with the twin objectives of alleviating poverty in the short run and in preserving the natural resource base in the long run so that future generations may have access to high quality income generating assets.

Evidence from other countries in sub-Saharan Africa is scanty. A study in Nigeria shows that deforestation is neither inevitable nor linearly related to population growth (Cline-Cole et al. 1990). In a study of 64 sites in Uganda, Place and Otsuka (1996) found large differences in land use change and tree cover on-farm and off-farm between 1960 and 1990. These differences were largely attributed to differences in population growth and density, but also to access to roads, land tenure regime, and to the responses of communities and individuals to initial tree density levels and hence their scarcity value.

In this study we are concerned with the influence of population pressure and land tenure on land use and tree resource management in Malawi. Regarding land tenure, a few

points unique to this country need to be mentioned. There are several tenure regimes governing the use and management of land and trees in Malawi, including customary or indigenous lands and the estate sector. While the estate sector has generated a great deal of policy debate in Malawi (Kydd and Christiansen, 1982; Dickerman and Bloch, 1989; Sahn and Arulpragasam, 1991), our study focuses on the customary sector which occupies about 70 percent of land area in Malawi. In this sector, cultivators are small peasants with few exceptions and many of them suffer from severe poverty. Uncultivated land is usually owned by the community and the permission from the village headman is required to clear the land for agricultural production (Nothale, 1982; Mkandawire, 1983/84; Hirschman and Vaughan, 1984). While the use rights over agricultural land are well established, the rights to transfer are limited, as in the case of other areas in Africa (Ault and Rutman, 1979). Hence, no land market exists and transfer of agricultural land is primarily through inheritance. Traditionally, matrilineal inheritance was practiced by the majority of the rural population, but the patrilineal system has increasingly been adopted in many areas. The objective of this study is to identify statistically how the customary land tenure institutions and their changes affect land use and management of woodland resources in the context of the customary sector of Malawi.

The remainder of the paper is organized as follows. We begin with a description of customary land tenure in our study sites which leads into our major hypotheses. This is followed by the presentation of the theoretical and econometric model used to test the hypotheses and then a brief description of data collection methods. The remaining sections describe and discuss the results of our analysis first at a descriptive level and then based on our econometric analyses. Finally, the conclusions and policy implications are presented to end the paper.

2. CUSTOMARY LAND TENURE SYSTEMS IN MALAWI

The customary sector consists mainly of smallholders and a key distinction here is between matrilineal and patrilineal ethnic groups. When land was abundant both systems vested land in chiefs and village headmen. The village headmen in turn would cede rights over specific tracts of land to families and family leaders. New lands could be opened through requests to family leaders and village headmen.

In a matrilocal system (practiced by the Chewa and Yao in our study sites), where husbands move to live in the wife's village, land had traditionally been passed from mother to daughter or from family leader to female family members (Mkandawire, 1983/84). The couple often resided permanently in the wife's village. This system is akin to matrilineal systems observed in some parts of Asia, such as Sumatra (Otsuka et al., 1997). In matrilineal systems where the couple moves to or otherwise resides in the husband's village, the husband traditionally acquired land from a village headman or family leader in his village. This land would then pass from uncle to nephew or niece (perhaps through the family leader). Inheritance of land from uncle to nephew is common in other areas of Africa, such as western Ghana. These traditions appear to have changed over time and new couples search for land from several sources and locations. Precious little evidence is available on this process and this paper hopes to document the extent of this change.

A patrilineal system, common in the north of Malawi, is similar to those elsewhere in sub-Saharan Africa in that men claim land ownership and can pass on all land developments to his children. It has been customary among the Ngoni and Tumbuka at our sites to favor

sons over daughters in inheritance practices.

The various modes of transfer and tenure arrangements might have different incentives on farm and tree management due to differences in tenure security (Besley, 1995). Land that passes from parents to children or merely from mother to daughter seemingly offer the appropriate incentives to households to make long-term investments (labour and capital) on the land. While these appear to be the majority of cases in matrilineal systems, three types of situations are of concern. The first is where husbands reside in the wife's village (uxorilocal marriage) only on a temporary basis. In some of these cases, the husband will have rights to land in his own village (and may have more than one wife). This situation can reduce the incentives for a man to work in his wife's village. A second situation is where land passes from uncles to nephews or nieces, bypassing the children. In such cases, couples may neglect the long-term quality of the land especially as they grow older. This case, however, is seldom observed in our sample. A third case concerns rights to land following death or divorce of a spouse. In the case where the widow or widower resides in the deceased's village, continued rights to land are not at all guaranteed. Where either a death or divorce becomes more likely, the non-resident spouse may likely increase activities which enhance short-term returns at the expense of long-term returns.

Although most farms in Malawi are small, constrained by the inability of household to cultivate much land with low levels of technology, some individual customary holdings in the northern half are quite large. These are farms in villages where 'communal' lands had been partitioned to families. The resulting large farms contain both cultivated and uncultivated areas. The latter are primarily miombo woodlands and are largely open-access land for village members (Lowore et al., 1995).

There are also communally held lands, held by the clan or village headman. Informal interviews during recognizant work found that these are virtual open-access resources, with few rules on user group membership or use rates (see also Coote et al., 1993a, 1993b). One notable exception is the Village Forest Area system initiated in the 1920s, rekindled recently by the Forestry Department, in which communities demarcate woodland areas to be placed under special management rules (which are always conservation oriented).

A first area of investigation will be the impact of population growth on changes in tenure systems. One of the hypothesized effects is that population growth will lead to a conversion from matrilineal systems of land acquisition to more non-traditional sources such as inheritance through the male's family (Mkandawire, 1983/84). This is likely to occur when sufficient land is unavailable in the wife's village due to the exhaustion of village land.

Secondly, we hypothesize that increased population pressure will lead to conversion of woodlands to privately held lands resulting in loss of tree cover. However, the relationship may not be linear and some woodland area may be maintained even at very high population densities. The speed of conversion may be conditioned by other factors such as the value of woodlands (e.g., values might be higher where livestock holdings are large) or the possibilities and incentives for intensification on existing agricultural land.

A third area of examination involves the effect of land acquisition mechanism and associated incentives on land use, tree planting and other natural resource investments. Greater investment in tree planting or preserving woodland resources may be associated with improved tenure security for males if they have greater power over natural resource management decisions. If this is the case, better management of woodlands and tree resources is expected in patrilineal rather than in matrilineal systems and in non-matrilocal

Mkandwire (1983/84) considers that males are heads of households in the matrilineal communities 1

of the Chewa ethnic group.rather than in matrilocal systems. The opposite could be the case, however, if women have greater decision-making power.1

3. THE THEORETICAL AND ECONOMETRIC MODEL

In this section, we construct a simple model of land use change in order to identify key variables affecting the farmer's decision to convert woodland to farm land. Conversion of land is one strategy households can pursue to increase production and income. We focus initially on the decision to convert because it is a good entry point into addressing related issues of deforestation and productivity. Specifically, we consider a farming household which faces the choice of preserving woodland or converting it to farm land. If it is preserved, it generates a net benefit of πat time t for a given area. For simplicity, we assume that the net Nt benefit grows at a constant rate of β. Then, the following relation holds:

π = π e ,(1)

Nt N0 β t where π is the net benefit in the current period. If woodland is managed sustainably, β is N0likely to be positive, because the value of those resources will increase over time. If woodland resources deplete sufficiently rapidly, however, β will be negative.

If woodland is converted to farmland, it accrues a net benefit of πat time t. Again At assuming a constant growth rate, π can be expressed as

At π = πe ,(2)

At A0 α t

Note that we have assumed away any positive externalities from woodlands which would affect 2

profits from agriculture. In the absence of steep slopes and irrigation, this seems reasonable.where π is the net benefit in the current period and α is its expected growth rate. The net A0benefit is likely to decline if crop farming is unsustainable and, hence, crop yield is declining.On the other hand, if production methods improve, α will be positive.2

Future benefits are discounted due to time preference and the risk of losing rights over future benefits from land. Under the matrilineal cum matrilocal system, if a wife dies or a marriage breaks down, the husband loses rights to use land and must return to his original village. Similarly, under the patrilineal cum patrilocal system, if a husband dies or a marriage breaks down, the wife often loses the right to receive benefits from land and must leave the husband's village. For simplicity, we assume that the discount rate combines the rate of time preference, r, and the expected rate of change in the probability of retaining land rights, ρ,both of which are constant. The magnitude of ρ will depend on the instability of marriage,mortality, and the bargaining positions of husband and wife. This rate may be small under the patrilocal system, if the husband has sole decision making power and disregards his wife's concern about the loss of her land rights in the assessment of his future benefits. In contrast,this discount factor may be positive and large if the husband is a major decision maker under the matrilocal system and is concerned with the chance of losing his land rights upon death of his wife or divorce in the future.

Under the assumptions made above coupled with the assumption of an infinite time horizon, the net present values of farm land and woodland can be shown, respectively, as

A similar specification of net future benefit is adopted by Mendelsohn (1994) in his model of land 3

use by squatters who possess insecure land rights.π/(r + ρ ? α) and π/(r + ρ ? β). Then, the benefit of converting woodland to farm land

A0 N03(V) can be expressed as the product of the converted area (?X) and the difference in the net future benefits between farm land and woodland:

V = ?X [π/(r + ρ -α) - π/(r +ρ - $)] .(3)A0 N0If V is negative, no woodland is converted to farm land. If V is positive, some woodland area may be converted to farm land, depending on any conversion costs incurred. As in the model of Ehui et al. (1990), an increase in farm productivity, which will increase π, leads to larger At conversion of woodland to farm land in the model.

The major cost of investment in land conversion is labor cost and, hence, investment cost (C) depends on the scarcity of labor, which will decline with increases in population density (N /X), and access to urban labor markets, but will worsen with increases in distance 0to urban centers (D). The investment cost also depends on the difficulty of land conversion,which is assumed to be reflected in the proportion of land area already converted to farm land (X /X). Thus, the cost function for investment may be expressed as:

A0 C = C(?X, N /X, D, X /X) ,(4)0 A0where both the first and the second partial derivatives with respect to ?X are assumed to be positive.

Woodland area is usually owned by the community and its use is controlled by the village headman. Thus, a community member who wants to clear woodland needs to obtain

permission from the headman. In reality, however, permission is easily granted and woodland area is essentially open access to the community members. We implicitly assume that woodland is an open-access area.

The household will determine the optimum amount of land conversion so as to maximize the net benefit of investment, V - C. Assuming the existence of unique interior maximum, the investment function can be derived as:

?X = ?X (π/(r + ρ ? α) ? π/(r + ρ ? β), N/X, D, X/X) .(5)

Α0Ν00A0

This function is not directly estimable, as we do not possess data on π, π, ρ, α, and β.

Α0Ν0 Therefore, in actual estimation, we use proxies and variables which are expected to affect those key variables. Firstly, we assume that the profitability of farming per unit of land in the

base period (π) is positively and closely correlated with yield of maize, which was estimated Α0

to occupy over 72 percent of smallholder cultivated area (in purestands and mixes) in Malawi in 1980 (Malawi Government 1984). Secondly, we assume that the net benefit from

woodland in the base period (π) can be captured by the average woodland tree cover.

Ν0

Thirdly, we assume that α and β are proportional to the rates of change in yield and tree cover, respectively. We note that population pressure on land may affect α and β by changing the scarcity of foods and woodland products, even if population pressure does not affect the physical farm output and the amount of woodland resource extraction. Finally we assume that ρ is conditioned by the land tenure system, which is reflected in the prevalence of the matrilineal cum matrilocal system or the proportion of land received from wife's family and the village headman of the wife's village (M). It is, however, difficult to determine the effect

of M on ρ or ?X on an a priori ground. Our own observation, as well as the observation of Ng'ong'ola et al. (1996), is that husbands are often the major decision makers with respect

to land use change and, hence, the risk of losing their land use rights is a major concern under the matrilocal system. Thus, we may hypothesize that ρ is a positive function of M. Even so, the effect of M on the relative profitability of land conversion is unclear, as it is conditioned by the relative magnitudes of α and β (see equation (3) or (5)). If crop farming is less sustainable than woodland resource extraction, so that α < β holds, an increase in ρ will increase the relative profitability of crop farming, and vice versa. In other words, an increase in ρ favors the choice of less sustainable land use system.

Assuming that each household maximizes the net benefit of investment and that all households in the community have common expectations about α and β, the aggregate area converted from woodland to agricultural land at the community level is the sum of the converted areas by households. Thus, we postulate that the land conversion at the community level is affected by the following set of explanatory variables:

?X/X = F(Y, G, T, G, M, N/X, G, D, X/X) ,(6)

0Y0T0N A0

where the dependent variable is normalized by total area of community (X) to adjust for

differences in the total size of communities; Y and T denote yield and woodland tree cover

00

in the initial period, respectively; G and G represent their expected growth rates; and G

Y T N

stands for the population growth rate. While Y, T, N/X, D, and X/X are either

000A0

predetermined or exogenous, G, G, M, and G should legitimately be considered as

Y T N

endogenous variables. For simplicity, we do not change notation for the converted area (?X)

at the individual level (equation (5)) and at the aggregate level (equation (6)).

If the Boserupian hypothesis of intensification of the farming system takes place in response to population pressure (Boserup 1965), or if the Hayami-Ruttan thesis of induced technological innovation takes place in response to changing factor scarcities (Hayami and

Ruttan 1985), G will depend on population density and on access to output and input Y

markets. Although maize yield had been stagnant or even declining in Malawi until the late 1980s (Kydd 1989), delayed green revolution in maize has taken place in some areas (Smale 1995). If induced farming intensification took place, we may expect that the green revolution was particularly successful in land scarce areas close to urban markets. Farmers may also adopt soil conservation practices to prevent declining yield in such areas by investing in land improvements including tree planting (Ndiaye and Sofranko 1994). Land tenure may matter in the growth rate of maize yield, to the extent that yield-enhancing investment plays a significant role. Thus, we assume the following functional relationship in the determination of yield growth:

G = G (Y, T, M, N/X, D, X/X, E) ,(7)

Y Y000A0

where E is a vector of exogenous environmental factors, such as rainfall, which would affect yield growth.

As was pointed out earlier, woodland resources have been depleted in Malawi, because no explicit management rules have been adopted (Coote et al. 1993a, 1993b; Lowore 1995). Even so, management efficiency may depend on the homogeneity of community members, since this affects the cost of organizing community actions and enforcing agreements (Clarke et al., 1996; Matose and Wily, 1996), and on the prevailing land tenure,

as this affects the extent of myopic behavior. Aside from the cost of mobilizing collective

action, the rate of resource extraction will depend on the demand for woodland resources, which is affected by population density and access to markets, and the supply, which is governed by the initial endowment of woodland and the ability to regenerate woodland. It is also important to note that the Malawi government has attempted to promote tree cover by designating village forest areas and by implementing tree planting projects. These considerations lead to the formulation of the following functional relationship:

G = G (Y, T, M, N/X, D, X/X, H, A, GP) ,(8)

T T000A0

where H represents the degree of homogeneity of community members, which may be measured by the proportion of the major ethnic group and the proportion of cattle-owning households using woodland for grazing; A represents an ability for woodland regeneration, which may be measured by the incidence of tree species with coppicing ability; and GP corresponds to government policies, such as protection of village forest areas and implementation of tree planting projects. In the estimation, we use the woodland tree cover

in the initial period for T and percentage changes in tree cover for G.

0T

We do not have clear-cut theory to explain secular changes in matrilineage cum matrilocal system towards patrilineal cum patrilocal system. In Malawi, ethnic groups under the patrilineal tradition reside mainly in northern and north-central regions, whereas other areas are occupied by ethnic groups practicing matrilineal descent traditions. Needless to say, such traditions affect the prevalence of the matrilineal system at present. Hirschmann and Vaughan (1983, 1984) argue that when land becomes scarce husbands in matrilineal societies work off farm and often migrate to other areas to find jobs. This suggests that ample

opportunities for off-farm employment help preserve the matrilocal system. If sufficient land is unavailable in the wife's village and off-farm employment opportunities are limited, a couple may seek land from the husband's family and reside in his village if land is available, thereby transforming a matrilineal system to a patrilineal system (Mkknadwire 1992). Therefore, we postulate that the tradition of ethnic groups, scarcity of land, and off-farm employment opportunities, among other things, affect the prevailing land tenure institutions measured by M, i.e.,

M = M (Y, T, N/X, G, D, X/X, EH) ,(9)

000N A0

where EH indicates dummies for ethnic groups distinguished by matrilineal and patrilineal traditions.

The population growth rate reflects both natural growth and net migration. We assume that population growth will be affected by the following set of variables: (1) ethnic traditions, (2) population density in the initial period, (3) access to markets, (4) productivity of farming, (5) proportion of already cultivated land, and (6) the prevalence of polygamy. The last variable is included as the degree of polygamy is expected to be a positive indicator of out-migration (of males) from the area and may also be related to lower birthrates per woman. Thus, we specify the following function to explain the population growth rate:

G = G (Y, T, N/X, D, X/X, EH, PO) ,(10)

N N000A0

where PO denotes the proportion of households under the polygamy system.

Aerial photographs for 60 sites were analyzed for land use and tree cover change, but three were 4

not visited for field surveys due to logistical difficulties.4. DATA COLLECTION METHODS AND DESCRIPTION

SAMPLING STRATEGY

The data for this study come from 57 enumeration areas throughout Malawi. There 4are about 6,500 enumeration areas in Malawi which were created for census purposes. This is the smallest unit for which boundaries are known and can thus be transferred onto maps with relative ease. The boundaries often follow physical land features and they do not run through villages. The next largest well defined area, an extension planning area, was often too large to ensure collection of accurate information about communities. Enumeration areas are designed to have roughly the same population and thus are different in size depending on the density of population. Our sample had a range of size from 180 to 5,580 hectares, with the mean being 1,271 hectares (or about 4km by 3km). Excluded from the sample were enumeration areas in the extreme north and south (Karonga and Shire Valley), and those corresponding to towns and gazetted state land. The remaining enumeration areas were selected by a stratified random process in which 15 each from the north, north-central, south-central, and south of Malawi were selected. This stratification was made to ensure a variation in population density and tenure type, both of which are known to vary from north to south in Malawi.

The data are reflective of the customary areas in the study sites as estates were found in only six of the enumeration areas and represented a small proportion of land area. Because

of the small influence of estates and the fact that smallholders have only recently been granted access to quotas of tobacco, the most important commercial crop in Malawi, we do not believe that demand for fuelwood for curing tobacco played any significant role in land use and tree cover change in our study.

NATURAL RESOURCE MANAGEMENT INDICATORS

Land Use and its Change

For each enumeration area, aerial photographs were obtained for 1971-3 and 1995. The 1970s aerial photos are from a countrywide coverage at a scale of 1:25,000. The 1995 aerial photos, kindly made available by the British Overseas Development Agency, were at the same 1:25,000 scale which greatly facilitated the land use and tree cover interpretation. The photos were all taken during the dry season in Malawi and there were no problems with cloud cover in the study sites. At such a scale, the minimum mapping unit (i.e., area which can be effectively differentiated and given boundaries) is about one hectare.

Interpreters differentiated as many as 27 classifications for land use and tree cover (see Appendix 1). The broad land use classifications were agriculture, wetland, grassland, forests, plantations, woodlands, woody regrowth (i.e., bushland), rocky areas, marshes, built-up areas, and surface water. Within agriculture, there were several sub-categories, including upland rainfed, irrigated, and wetland cultivation. Within upland agriculture (which comprises 99 percent of all agricultural land), further differentiation was made according to tree cover with separate classes identified for under 2 percent cover, 2-5 percent cover, 5-10 percent cover, 10-20 percent cover, and 20-40 percent cover. Forests were differentiated by normally stocked (averaging of 80-100 percent) and depleted (average of 50-60 percent). Woodlands

If there was significant tree cover, then the area would have been reclassified (e.g., from grassland 5

to woodland).

It is worthy to note that the 1995 land use estimates from our enumeration areas correlate very 6closely to those from a larger study designed to be representative of Malawi (Green and Mkandawire, 1996).

Note that agriculture is a small proportion of land in the larger enumeration areas (in northern 7Malawi). Thus, the weighted share of land in agriculture is much lower, .50 in 1995, for example.and regrowth were each differentiated into five tree cover classes: 2-20 percent, 20-40percent, 40-60 percent, and 60-80 percent cover. Other land use categories were largely devoid of trees by definition.5

Natural forests and plantations were easily distinguished from each other by the pattern found in plantations. The two were also distinguished from woodlands by the density of stand and type of species. The difference between woodland and bushland is less obvious.An area is classified as woodland if the average height of woody vegetation exceeds four meters in height; it is bushland otherwise. In our analyses, the two are always grouped together. Where a tree cover estimate was difficult to discern, a grid with square sampling areas was overlayed on the area in question and the tree cover from the sampled squares was measured by hand or with the aid of available cover guides. Areas were measured with the use of a planimeter or circle grid.

Table 1 displays some statistics on broad land use categories in the 1970s and in the 1990s using simple averages across the 57 enumeration areas. Agriculture had the largest 6share of the enumerated land area in both years, averaging .52 in 1972 and .68 in 1995.7Roughly 20 percent of enumeration areas had an agricultural land share of over .80 in 1972but as many as 48.0 percent achieved the same proportion by 1995. Despite this, a few enumeration areas from the north of Malawi maintained relatively little agricultural activity during the period. The lowest share of land in agriculture among enumeration areas was .03