Cover Image
close this bookAgricultural Growth Linkages in Sub-Saharan Africa - Research Report 107 (IFPRI, 1998, 152 p.)
View the document(introduction...)
View the documentForeword
View the documentAcknowledgments
View the documentSummary
View the documentCHAPTER 1. - Introduction
View the documentCHAPTER 2. - Concepts, Prior Work, and Issues Pertaining to Agricultural Growth Linkages
View the documentCHAPTER 3. - Methodology and Overview of Case Studies
View the documentCHAPTER 4. - North to South in Burkina Faso
View the documentCHAPTER 5. - Southwestern Niger
View the documentCHAPTER 6. - The Senegalese Groundnut Basin
View the documentCHAPTER 7. - Eastern Province, Zambia and Gazaland District, Zimbabwe
View the documentCHAPTER 8. - Conclusions
View the documentBibliography

CHAPTER 4. - North to South in Burkina Faso

The Burkina Faso case study makes use of an extraordinarily rich household-level data set from three agroecological zones in the central interior of West Africa. It gives a detailed view of the impacts of growth on the tradable farm and nonfarm sectors and on the overall household incomes of the rich and the poor. Because the household samples are drawn from several locations, the effects on cash cropping areas, with their higher agricultural potential, can be compared with those in the more arid livestock and cereal producing regions. Detailed information is also given on farm and nonfarm production and consumption activities and sources of income.

The data come from a collaborative survey conducted by the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) and IFPRI. The main baseline survey on the production and income side, conducted by ICRISAT, covered four years, 1981-85, a period that included both good and bad harvests. The expenditure and income surveys and aggregation and cleaning processes yielding the data used in the present analysis were conducted on the same sample for 1984-85, with the main input coming from IFPRI (Reardon, Delgado, and Matlon 1992).

The original surveys were conducted in three zones of Burkina Faso: the Sahelian, the Sudanian, and the Guinean regions. The true Sahelian zone in the northwest is agroclimatically a poor area with low rainfall, poor soils, and extremely variable cropping outcomes. The Sudanian zone in the Mossi Plateau is an agroclimatically poor-to-intermediate area with low-to-medium rainfall, poor soils, and moderately variable cropping outcomes. The Guinean zone in the southwest is a moderately favored zone with medium-to-high rainfall, good soils, and relatively stable cropping outcomes (Matlon 1988).

These data have been analyzed for several purposes, particularly to examine household coping behavior in the face of income shocks (Reardon, Matlon, and Delgado 1988), to determine household income diversification behavior, and to explore the dependency of nonagricultural production on agricultural income (Reardon, Delgado, and Matlon 1992; Reardon et al. 1994). However, the present study is the first time the data have been used to estimate growth multipliers in order to explore how the consumption patterns of rural people can potentially stimulate further net rural economic growth. The study also demonstrates the relative importance of consumption growth linkages relative to production linkages.

Characteristics of the Burkina Faso Sample

Characteristics of the three sample zones and the subsamples within each zone are summarized in Table 5. Average long-term rainfall doubles going from north to south, while the variability of rainfall lessens. The density of farm workers per hectare is highest in the north (the Sahelian zone), about one-third less in the south (the Guinean zone), and even less in the middle (the Sudanian zone), where low yields under subsistence cultivation support a smaller labor input per unit of land.

The income data in Table 5 have been extensively analyzed in Reardon, Delgado, and Matlon (1992) and Reardon et al. (1994). The table shows five results that are significant for the purposes of this report. First, there are significant income disparities within rural areas. Second, in the more humid Guinean zone where cotton is grown, the poor are not much poorer than the poor elsewhere in Burkina Faso, but the "rich" in Burkinabe terms are 50 percent better off. Third, in all three zones, contrary to experiences in South Asia, the poor receive a higher share of their income from their own farms than do the rich. This relationship is evident in both the north, where nonfarm income is from outside sources and crop potential is slight, and the south, where non-farm income tends to be locally generated.

In the middle, where agricultural potential remains low but farmers have historically been less dependent on outside income than in the north, the farm income disparity between the rich and the poor is less severe than in the other zones. Fourth, as a counterpart, the rich are far more involved in supplying local services than are the poor in all three zones, though this relationship is less pronounced in the middle. Fifth, local crafts and manufactures do not account for much income of the rich in any zone and only a very small share of income of the poor in all zones. The best showing for the poor is in the cotton-growing south, where respending of cotton income allows the poor to get 5 percent of their income from local crafts.

Table 5 - Characteristics of the ICRISAT/IFPRI sample households, by agroecological zone, Burkina Faso, 1981-85

Sample characteristic

Sahelian

Sudanian

Guinean

Number of households

45

44

47

Long-term average rainfall (millimeters)

480

724

952

Coefficient of variance for long-term rainfall

0.34

0.25

0.21

Land per adult equivalent (hectares)

0.92

0.58

0.65

Annual income per adult equivalent (US$)a

145

140

191


Poorest one-third of households

86

100

99


Richest one-third of households

260

238

368

Percent of income from own farm

63

66

57


Poorest one-third of households

81

86

71


Richest one-third of households

53

74

49

Percent of income from local services

32

16

31


Poorest one-third of households

11

10

15


Richest one-third of households

44

12

39

Percent of income from local manufactures

1

0

3


Poorest one-third of households

1

1

5


Richest one-third of households

0

0

0

Source: Compiled from Table 4 in Reardon, Delgado and Matlon 1992 and Table 3 in Reardon et al. 1994.

Note: Income terciles are based on annual household total expenditures per adult equivalent, including income in kind valued at market prices.

aEstimated at 290 CFA francs per US$1.00 in 1981-85 dollars.

Thus, unlike the usual Asian assumptions of a rich landowning class that obtains its income from the land and a poor class of landless laborers that has to rely on services, in Burkina Faso the relatively wealthy profit from the nonfarm economy more than the poor, primarily through the provision of services. Nonfarm goods tend not to be of local origin, and thus they account for little local income. Finally, the best market for local crafts and services is in the most monetized area, the cash-cropping south.

Growth Linkages and Tradability

The tradability assumptions used in the present study are based on long-term familiarity with rural consumption and trade patterns in Burkina Faso, including the sample zones. For convenience, the definitions of tradability and the assumptions made for specific groups of goods are summarized in Table 6. At the local level, most goods are tradable (goods typical of the group are frequently exported to or imported from places outside a 100-kilometer radius of the study zone). Local nontradables are services, prepared foods that are not packaged for transit (such as sorghum, beer, and millet cakes), and fresh meat and dairy products.

At the national level used for analysis, more items are classified as nontradable in the sense that they are not typically exported from or imported into Burkina Faso, nor are their prices closely correlated with similar goods that are traded.5 The most significant change is that millet and sorghum, the basic food sources in the study zones, are considered nontradables. Previous linkages work tended to assume that all foodgrains are tradables and thus solely supply-constrained. While this may be true of maize in some years, at prevailing prices millet and sorghum are rarely brought in commercially from neighboring countries or exported to them, except under extreme and rarely observed circumstances. They are therefore demand-constrained. Production of these goods rarely uses inputs other than land, family labor, hand tools, and seed saved from the previous harvest. If land and labor are going unused during the cropping season, there is scope for supply increase, although such a response will involve some juggling of labor during the bottleneck periods of mid-July and early November (Delgado and Ranade 1987).

5 This is not to say that they are never traded but that such trade within the study zone is rare at prevailing prices and costs. They are nontradable rather than just nontraded if their prices are determined primarily by local factors.

Finally, at the regional level of tradability, the only products consumed in the study area that remain tradables are rice, groundnuts, and nonlocal, nonfood commodities such as matches, batteries, bicycles, and radios. Even maize consumed in the study area (with the exception of food aid in major drought situations) is West African in origin, and distinct from world market maize.

Table 6 - Assumptions about tradability by reference market, Burkina Faso


Reference market (catchment area)

Sector

Local

National

Regional

Coarse grains


Millet and sorghum

T

NT

NT


Maize

T

T

NT

Convenience starch


Wheat products, tubers, and condimentsa

T

T

NT


Rice

T

T

T

Other food staples


Groundnuts

T

T

T


Other pulses and legumesb

T

NT

NT

Meat, milk, eggs, and fish


Chicken and guinea fowl

T

T

NT


Otherc

NT

NT

NT

Prepared foods, beverages, and colad


Bottled drinks and cola nute

T

NTe

NT


Other (such as dolo)f

NT

NT

NT

Nonfood commodities


Local rural manufactures and craftsg

T

NT

NT


Outsideh

T

T'

T

Servicesi

NT

NT

NT

Notes: T is tradable and NT is nontradable. Tradability at the local level means that the good is sometimes exported or imported within the local area (100-kilometer radius around the market). Tradability at the national level implies that the good is frequently imported to or exported from Burkina Faso or to or from its neighbors. Tradability at the regional level means that the good consumed in Burkina Faso is often exported to the world market or imported from the world market.

aWheat, macaroni, bread, prepared rice meal, pepper, onion, tomato, lettuce, okra, sorrel, sauce leaves, egg-plant, unspecified leaves, sugar, salt, cauliflower, garlic, unspecified fruit, Maggi bouillion cubes, cotton seed, sesame, honey, soumbala (a pungent condiment made of fermented locust beans and spices), oils/butter, cooking oil, Irish potatoes, cassava, yam, sweet potato.

bEarthpeas, cowpeas.

cCattle, donkey, horse, goat, sheep, pig, fish, milk, egg, other meat, dairy butter.

dCola is tradable at the national level but is not important enough to warrant a separate category.

eCola nuts, beer, soft drinks, wine.

fUnspecified meal, couscous of fonio, couscous of pearl millet, gruel, fried millet cakes, biscuit, cookies and cakes, groundnut butter, miscellaneous snacks, fried groundnut butter, cooked skewered meat, sorghum beer, coffee, leven, tobacco and cigarettes.

gWater, wood, furniture, unspecified farm inputs, livestock feed, hunting materials, bedding.

hKerosene, gasoline, motor oil, matches, soap, batteries, medicine, vehicles, electronic and photo equipment, cooking utensils, clothing, toilet articles, rope, lamps, mosquito coils.

iCereal milling, ceremonial expenses, school fees, taxes or other fees, transportation fare, vehicle repair, housing repair, labor payments, herding, bride price payment, gifts, communication expenses.

Parameter Requirements and Estimation

Estimation of Marginal Budget Shares

As outlined in Chapter 3, the modified Working-Leser model of Hazell and R (1983) is used to obtain estimates of MBS (see equation [4]). Household characteristics included are the number of livestock per adult equivalent and dummy variables for household ethnicity (Bwaba, Fulbe, or Fulani), for access to a road, and for market group (coincident with agroecological zone).

The equations are estimated separately for each of the 12 groups of goods laid out in Table 6 using the overall sample. MBSs are then obtained for subsamples, such as income terciles or geographic zones, by using mean subsample values for the data in the estimating algorithm for the MBS (equation [5] in Chapter 3). These 12 MBSs are aggregated into farm tradables, farm nontradables, nonfarm tradables, and nonfarm nontradables. MBSs for the four sectors are calculated for the overall sample, the three income groups, and the three agroecological zones. Separate results are presented for different aggregations stemming from different definitions of catchment area, but the preferred interpretation is at the national level.

Value Added, Technology, and Savings

By assumption, technology parameters (the as and vs) do not differ over agroecological zones or over income terciles. Consistent with the model, fixed-coefficients (Leontief) technology is maintained. The value of the savings ratio out of household income is also assumed constant.

Technological parameters are derived from three sources. Values for intermediate deliveries of farm nontradables to farm tradables and nontradables (aan.at, aan.an) and value added from farm tradables and nontradables (vat, van) are obtained from average farm budgets calculated from the larger data set. Values for intermediate deliveries for farm nontradables to nonfarm tradables and nontradables (aan.mt, aan.mn) intermediate deliveries for nonfarm nontradables to farm tradables and nontradables (amn.at, amn.an), and value added from nonfarm tradables and nontradables (vmt, vmn) are obtained from calculations using data from a social accounting matrix (SAM) for Niger (Dorosh and Nssah 1991). Values for intermediate deliveries for nonfarm nontradables to nonfarm tradables and nontradables (amn.mt, amn.mn) are guesses based on values taken from SAMs for Niger and Cameroon.6 The savings rate is the overall sample average of the ratio of savings to total income for each household in the 1984 harvest year. The parameter values summarized in Table 7 are close in orders of magnitude to estimates of similar parameters for Sierra Leone in Haggblade, Hammer, and Hazell 1991 and Haggblade and Hazell 1989. The sensitivity of results to the Burkina Faso assumptions is reported farther on.

6 Niger estimates are from the SAM in Dorosh and Nssah (1991), and Cameroon estimates are from the SAM in Ganthier and Kyle (1991).

Table 7 - Parameter assumptions for model estimation, Burkina Faso


Intermediate deliveries from nontradables sector (technical coefficients)

Sector

Value-added shares

Farm

Nonfarm

Savings ratio

Farm tradables (AT)

0.85

0.055

0.06

0.06

Farm nontradables (AN)

0.93

0.036

0.03

0.06

Nonfarm tradables (MT)

0.49

0.010

0.10

0.06

Nonfarm nontradables (MN)

0.69

0.030

0.20

0.06

Sources: Values for intermediate deliveries, value-added shares for farm items, and savings ratios are calculated from the ICRISAT/IFPRI survey data for Burkina Faso, 1981-85. Values for nonfarm items are specified using data from a social accounting matrix for Niger (Dorosh and Nssah 1991).

Average and Incremental Household Consumption Patterns

Average and Marginal Budget Shares by Category of Goods

The household ABSs for the entire sample for 12 groups of goods are computed directly from the data and presented in Table 8. The corresponding MBSs estimated econometrically by the procedure described in Chapter 3 are also given. The ratio of the MBS to the ABS (not shown) is the expenditure elasticity computed at sample means; an MBS smaller than an ABS implies that the relative importance of a goods group is falling as incomes rise (income-inelastic demand). This is the case for millet and sorghum and for pulses and legumes. The reverse is elastic demand, which is the case for meat and dairy products, bottled drinks, nonfood commodities, and services. Convenience starches such as wheat products, rice, and tubers maintain their relative shares as income increases.7

7Table 8 suggests surprisingly income elastic demand for maize. This is probably an anomaly of the sample period, when massive imports of newly harvested Ghanaian maize flowed into the markets of Burkina Faso beginning in May 1985 at the height of the great drought. At the same time, households were scrambling to feed themselves, with the wealthier households increasing their purchases of maize faster than the poor.

Marginal Budget Shares by Sector

The additive properties of MBSs and ABSs allow the discrete components to be reaggregated into different composite groups of goods. Thus the goods and services are sorted by tradability characteristics at the local, national, and regional levels and divided into farm and nonfarm tradables and nontradables. Farm goods include crops and livestock, while nonfarm products include prepared foods and services in addition to the usual nonfarm goods.

Table 8 - Household consumption patterns, rural Burkina Faso, 1984/85

Sector

Average budget shares

Marginal budget shares


(percent)

Coarse grains


Millet and sorghum

48.4

42.1


Maize

10,3

10.4

Convenience starcha


Wheat, tubers, condiments

4.9

3.6


Rice

1.9

0.9

Other food staples


Groundnuts

6.6

4.4


Other pulses and legumes

2.5

0.7

Meat, milk, eggs, and fish

2.9

2.5

Prepared foods, beverages, and cola


Bottled drinks, cola nut

3.7

7.5


Other (such as dolo)

3.7

2.6

Nonfood commodities


Local nontradables

0.4

0.6


Nonlocal

8.2

13.4

Services

6.5

11.3

Source: See text for estimation procedure, using expenditure data from IFPRI/ICRISAT household surveys of 122 households in three agroecological zones in 1984/85.

Note: Dolo is indigenous sorghum beer made by local artisans.

aThis category includes high-priced staples that are convenient to prepare, such as bread, macaroni, rice, potatoes, and so forth.

The first column of Table 9 shows that for the overall sample, using the local definition of tradability, an additional $1.00 of income is spent as follows: $0.62 on farm tradables, $0.03 on farm nontradables, $0.22 on nonfarm tradables, and the remaining $0.14 on nonfarm nontradables. However, switching to the national definition of tradability, only $0.19 of each extra dollar is spent on farm tradables, while the share of farm nontradables rises to $0.45. This switch is largely due to reclassifying millet and sorghum as nontradables. Adopting the national definition of tradability reclassifies large shares of household spending from incremental income on items that are demand-constrained from the perspective of the country as a whole.

Sectoral MBSs were calculated for subsamples using overall sample expenditure elasticities and subsample ABSs for the groups concerned, using the Working-Leser procedure in Chapter 3. For all three definitions of tradability, the MBS of farm items declines and that of nonfarm items increases as income increases, consistent with the results in Table 8. For example, using national tradability, as was typically done in earlier linkages studies, the MBS allocated by the highest income tercile to farm tradables in Table 9 is 2.1 percentage points lower than the poorest one-third of households, while the share going to nonfarm nontradables is more than 12 percentage points greater. The direction of change with increasing income is similar using the regional definition of tradability, except that the decline in share of farm nontradables is much more pronounced given that the main foodgrains are nontradables under this catchment area. Using the national definition of tradability, moving from the Sahelian north to the Guinean south shows a more than 10 percentage point decline in the absolute marginal share of farm tradables, a nearly 28 percentage point decrease for farm nontradables, and a 19 percentage point increase for nonfarm nontradables.

Table 9 - Marginal budget shares by sector, income, and ecological zone, Burkina Faso, 1984/85


Income tercile

Ecological zone

Sector/Catchment

Overall

Poor

Middle

Wealthy

Sahelian

Sudanian

Guinean


(percent)

Farm tradables (AT)


Local

62.2

73.3

63.3

47.5

87.5

63.6

44.5


National

19.3

21.1

17.2

19.0

24.3

21.7

13.9


Regional

5.3

7.8

4.1

4.2

1.1

9.3

4.9

Farm nontradables (AN)


Local

2.5

2.1

2.3

3.4

0.1

1.8

4.6


National

45.3

54.3

48.4

31.9

63.2

43.7

35.3


Regional

59.4

67.6

61.5

46.6

86.4

56.1

44.2

Nonfarm tradables (MT)


Local

21.5

15.1

21.6

28.9

9.7

20.7

30.0


National

13.4

8.2

13.4

20.5

4.3

8.8

23.2


Regional

13.4

8.2

13.4

20.5

4.3

8.8

23.2

Nonfarm nontradables (MN)


Local

13.8

9.5

12.8

20.2

2.8

14.0

20.9


National

21.9

16.4

21.0

28.7

8.2

25.8

27.6


Regional

21.9

16.4

21.0

28.7

8.2

25.8

27.6

Source: See text for estimation procedure, using data from IFPRI/ICRISAT household surveys of 122 households in three agroecological zones in 1984/85.

Notes: Subsample MBSs were estimated at subsample means. See footnote 4. Income terciles are determined by ranking the samples in ascending order based on the total annual household expenditure per adult equivalent, including income in kind valued at market prices.

Growth Multipliers

Farm Growth Multipliers for the Overall Sample

The parameters shown in Tables 7 and 9 were used in the model to yield the growth multipliers reported in Table 10. As an example, using the national definition of tradability, the overall growth multiplier for $ 1.00 spent on farm goods is $2.88. Thus, the initial $1.00 induces a net additional increase of $1.88 of income, through net increases in intermediate demands and new consumption of nontradables. In Table 10, the local definition of tradability yields a farm growth multiplier of 1.31, implying that the initial tradable income shock leads only to an extra $0.31 in net new income through respending on nontradable consumer items and intermediate inputs. This amount is notably similar to estimates for West African multipliers offered in Haggblade, Hammer, and Hazell (1991). The much larger multipliers shown using the regional assumption stem solely from the fact that at the regional level of tradability, most of what people spend incremental income on is nontradable, and leakages to tradables and savings are low. Thus, increments to income cycle in the economy, being re-spent over and over again, stimulating demand-constrained activities.

Table 10 - Farm and nonfarm growth multipliers for rural Burkina Faso, 1984/85


Local

National

Regional

Tradability sector

Farm

Nonfarm

Farm

Nonfarm

Farm

Nonfarm


(US$)

Overall

1.31

1.40

2.88

3.07

4.33

4.62

Zone


Sahel

1.16

1.23

3.31

3.53

9.07

9.68


Sudano

1.30

1.39

3.01

3.22

4.37

4.67


Guinean

1.45

1.54

2.58

2.76

3.19

3.40

Income


Poorest third

1.25

1.34

3.18

3.39

4.89

5.22


Middle third

1.30

1.38

3.04

3.25

4.54

4.84


Richest third

1.41

1.51

2.45

2.62

3.50

3.74

Source: Results of the model.
Note: Income terciles are based on annual household total expenditures per adult equivalent.

This illustrates the crucial role of the assumptions concerning the size of the catchment area and nontradability, with embedded assumptions about the elasticity of supply response. Such major assumptions are not unique to growth multiplier analysis. The usual practice of using border prices f.o.b. West African ports as reference prices for Sahelian destinations in project evaluation and trade analysis is another example of an embedded assumption. If the regional definition of tradability adopted here somewhat overstates the lack of tradability of major items such as coarse grains, the more usual local definition clearly understates it. The national definition seems a reasonable compromise.

The 2.88 multiplier at the national level is much higher than the conventional view of low growth linkages in Africa, even allowing for a 30 percent overestimation of the multiplier, which is presumably the result of an overly optimistic view of the elasticity of supply of nontradables. While a multiplier of 1.90 is less than one of 2.88, it is still remarkably close to the Asian multiplier of 1.80 cited in Chapter 2.

Nonfarm Growth Multipliers

One of the novelties of the model used here is that the impact of shocks on the rural nonfarm tradables sector can be estimated separately from agricultural growth linkages. Some examples of such shocks would be discovery of mineral wealth or the opening up of new export markets for handicrafts. Nonfarm growth multipliers are also displayed in Table 10. The directions of change for both the overall sample and the subsamples closely parallel the results for agriculture, except that nonfarm multipliers are consistently about 7 percent greater than corresponding farm growth multipliers. The difference is primarily the result of different intermediate demands and the valued-added structure of nonfarm tradables relative to other sectors.

The conclusion is that the stimulus to rural nonfarm tradables is at least as efficient at stimulating rural growth as the stimulus to farm tradables, provided it is possible to stimulate growth initially. Which one to emphasize, however, depends on the comparative advantage of farm tradables versus nonfarm tradables in Burkina Faso. At the present time, farm tradables are thought to be more likely to exhibit comparative advantage in trade with neighboring countries and with the rest of the world than non-farm tradables.

Differences in Multipliers by Agroecological Zone

Going from the Sahelian north to the Guinean south in Table 10, farm growth multipliers increase somewhat, using the local definition of tradability, but they decrease sharply using the regional definition. The national definition yields a 22 percent decline going from north to south. This difference stems from consumption patterns in the north that are more oriented to nontradables and production patterns that make relatively little use of intermediate inputs. Stimulating demand in the north (through increased livestock exports, say) would stimulate little overall demand in the economy if the resulting increased grain consumption in the north just encourages more grain imports into West Africa. However, if increased demand stimulates demand-constrained West African production, then the growth linkages could be large (with an estimated upper limit of an additional $2.30 for each initial $1.00). The choice between the two alternatives depends on one's view of where extra grain consumption in the north is likely to come from and what substitution effects it is likely to produce regionally.

Differences in Multipliers by Income Group

Going from the poorest to the richest one-third of sample households in Table 10, farm growth multipliers estimated under the national definition of tradability are highest for the low-income group, with the lowest multipliers being calculated for the richest one-third of households. This is because the consumption pattern of the poor is largely made up of basic foods that are nontradable with respect to the world market. This confirms Harriss's (1987) intuition that broadening the catchment area beyond the local level will have the effect of improving the income multiplier for income shocks targeted to the poor. Positive income shocks to the agriculture of the poor still produce $0.73 more net income than similar shocks to the rich. Conversely, negative income shocks to the poor are even more damaging to overall rural income than negative income shocks to the rich, because income shocks affect the consumption spending of the poor more severely than they do the consumption spending of the rich.

The Share of Growth Multipliers Attributable to Consumption and Sensitivity Analysis

The model assumes constant technology and commodity composition by sector, across agroecological zones and income groups, while allowing for actual changes in the commodity composition of consumption.8 These changes are manifested as different MBSs depending upon the degree of tradability, agroecological zone, and income group. Thus differences between groups in the present analysis are driven entirely by differences in consumption patterns.

8The assumption of constant technology (common a's and v's) across income groups is straightforward. The assumption of a common commodity composition of sectors across agroecological zones, and thus of common technology across zones, is also reasonable, even if in the north production of farm tradables is likely to involve more groundnuts and less cotton than in the south, and the two in fact have a different set of intermediate demands. At all levels of tradability in the sensitivity analysis, the composite set of a's and v's used reflected the commodity composition at the national level of tradability for lack of a better alternative, even though changing tradability means changing the commodity composition of sectors.

Following Haggblade, Hammer, and Hazell (1991), an estimate of the pure production multiplier is obtained by setting all the MBSs to zero in equations (40) and (41) in Chapter 3. This yields a farm growth multiplier of 1.13 and nonfarm multiplier of 1.20. In other words, for the overall sample at the national level of tradability, an income stimulus of $1.00 to the farm tradables sector produces an extra $0.13 of income from new intermediate demand for nontradable inputs.

These pure production multipliers make up only a small part of the overall growth multipliers listed in Table 10. With the national definition of tradability, the consumption effects that remain after removal of the pure production effects account for 93 percent of the farm growth multiplier and 90 percent of the nonfarm growth multiplier (using the overall sample). Ignoring these effects, as is the case when only backward and forward production linkages are considered, would lead to a severe underestimate of growth multipliers. It also suggests that correct estimates of the consumption parameters are far more important to results than the value-added shares and technology coefficients, at least in an economy such as that of Burkina Faso.

This proposition was tested more directly through sensitivity analysis on the parameters (Table 11). The sensitivity analysis was necessarily selective and focused primarily on individually testing the effects of 10 percent changes in the entire set of relevant as, vs, and bs for both farm and nonfarm growth multipliers. The effect of the changes on the farm component per se of farm growth multipliers is also shown.

The elasticity of change in the farm multiplier with respect to a change in all the technology coefficients (as) is on the order of 0.1 (or 1.04/10). The farm component of the overall farm growth multiplier is the component of the total multiplier that is due to increased demand for farm nontradables. This has an elasticity of 0.16 with respect to a change in all the as. There is considerable curvature in the nonfarm multiplier for the as. Thus a 10 percent increase across-the-board in the as is associated with a 2.0 percent increase in the nonfarm multiplier, while a 10 percent decrease in the as is associated with a 1.6 percent decrease in the nonfarm multiplier. The effect of an across-the-board increase of 10 percent in all the value-added shares (vs) without changing anything else is to lower multipliers by about 8 percent (for an increase in vs) or to raise them by about 9 percent (for a decrease in vs). Changing the MBSs (the bs) for nontradable goods provokes an elastic response in the multiplier estimates, ranging from an 18 to 29 percent increase in multipliers for a 10 percent increase in the bs to a 13 to 21 percent decrease in multipliers for a 10 percent decrease in bs.

Table 11 - Sensitivity analysis for parameters (percentage change in multiplier from a 10 percent change in parameter)


Farm multiplier


Parameter

Total

Farm nontradable

Nonfarm multiplier


(percent)

Eight intermediate demand coefficients


Increase of 10 percent in all aan.j and amn.j

1.0

1.6

2.0


Decrease of 10 percent in all aan.j and amn.j

-1.0

-1.6

-1.6

Four value-added shares


Increase of 10 percent in vj

-8.3

-7.9

-7.8


Decrease of 10 percent in vj

9.0

9.4

8.5

Two nontradable marginal budget shares


Increase of 10 percent in ban and bmn

18.4

29.1

18.6


Decrease of 10 percent in ban and bmn

-13.5

-21.3

-13.4

Source: Calculated from model results, with incorporation of changes indicated relative to the baseline. Each change is in isolation.

Notes: In the notation, the sectors j = at, mt, an, mn. The changes in the multipliers are for simultaneous changes in, respectively, the eight aij, then the four vj and finally the two bs with as and vs at original values, that enter each multiplier. Since aat.j + amt.j + aan.j + amn.j + vj £ 1, a uniform fixed percentage increase in intermediate demand coefficients is compensated in the calculations by a decrease in value-added shares, so that the above relation continues to hold, and similarly, a uniform fixed percentage increase in value-added shares is compensated by a decrease in intermediate demand coefficients. The relationship between multipliers and parameters is nonlinear, such that increases and decreases of parameters do not necessarily produce symmetric effects.

In sum, the choice of as is of relatively minor significance for the final elasticity estimates. The vs matter much more, but still only half as much as the MBSs for nontradables, which truly drive the results. This confirms the importance of getting tradability assumptions right, because erroneously classifying an important set of goods as supply-constrained tradables rather than demand-constrained nontradables severely reduces multiplier estimates. The reverse also holds: erroneously assuming a good to be a nontradable greatly inflates multiplier estimates.

Decomposition of Growth Multipliers by Source

The multipliers reported in Table 10 measure the net total effect of exogenous increases in income from the tradables sector and are set out as equations (26) (the farm growth multiplier) and (27) (the nonfarm growth multiplier) in Chapter 3. Both of these equations can be decomposed into three parts: (1) the initial tradable income stimulus, equal to unity by definition, (2) the net additional income from intermediate demands and consumer respending in the farm nontradables sector stemming from the initial stimulus, and (3) the net additional income from intermediate demands and consumer respending in the nonfarm nontradables sector stemming from the initial stimulus. These three components have been calculated separately and set out in Table 12 as T, A, and M, respectively. The sum of the three components is the total multiplier, also reported in Table 10.

Thus under the national definition of tradability, the total farm growth multiplier for the poorest one-third of households is 3.18, of which $1.00 is the initial exogenous shock, $1.65 stems from net new spending on farm nontradables, and $0.53 comes from net new spending on nonfarm nontradables. In relative terms, 31 percent of the total increment to income stems from the initial stimulus, 52 percent from net new re-spending on farm nontradables, and 17 percent on nonfarm nontradables.

Table 12 - Sources of income growth from linkages by catchment area, income group, and ecological zone, Burkina Faso, 1984/85


Income group

Ecological zone

Catchment area

Poorest third

Richest third

Sahel

Guinean

Farm


Local







Tat

1.00

1.00

1.00

1.00



A

0.09

0.12

0.07

0.14



M

0.16

0.29

0.09

0.31



Total

1.25

1.41

1.16

1.45


National







Tat

1.00

1.00

1.00

1.00



A

1.65

0.80

1.97

0.92



M

0.53

0.65

0.34

0.66



Total

3.18

2.45

3.31

2.58


Regional







Tat

1.00

1.00

1.00

1.00



A

3.09

1.58

7,20

1.38



M

0.80

0.92

0.87

0.81



Total

4.89

3.50

9.07

3.19

Nonfarm


Local







Tmt

1.00

1.00

1.00

1.00



A

0.06

0.08

0.03

0.10



M

0.28

0.43

0.20

0.44



Total

1.34

1.51

1.23

1.54


National







Tmt

1.00

1.00

1.00

1.00



A

1.72

0.81

2.06

0.94



M

0.67

0.81

0.47

0.82



Total

3.39

2.62

3.53

2.76


Regional







Tmt

1.00

1.00

1.00

1.00



A

3.26

1.64

7.65

1.42



M

0.96

1.09

1.03

0.98



Total

5.22

3.74

9.68

3.40

Source:

Results of the model components.

Notes:

Tat is the initial income growth for farm tradables.


Tmt is the initial income growth for nonfarm tradables.


A is the farm component of the overall growth multipliers.


M is the nonfarm component of the overall growth multipliers.


Income terciles are based on annual household total expenditures per adult equivalent.

Four trends are apparent from the decomposition of farm growth multipliers in the top part of Table 12. First, under assumptions of local tradability, the contribution of nonfarm items (M) to the total multiplier tends to be higher than the farm component (A), especially for the richer households and especially in the southern Guinean zone, which has higher agricultural potential. These results are consistent with earlier linkages work, which tended to see linkages primarily as the way in which agricultural growth stimulates nonagricultural growth.

Second, once the assumptions consistent with a bigger catchment area are substituted, the absolute and relative contributions of farm items themselves to overall farm growth multipliers increases sharply, because so many farm items become nontradable. The two remaining trends are largely related to this finding. Third, the absolute and relative contributions of farm items to farm growth multipliers tend to be higher for the poor than the rich, and this relationship escalates sharply as the catchment area expands. Fourth, the relative and absolute roles of farm items in farm growth multipliers tend to be more important in the north and the south, when the national and regional definitions of tradability are adopted.

In sum, when the local definition of tradability is used, similar to that in previous household linkages work, income shocks to the poor or to the north as a region - or both - stimulate little extra income growth through linkages. Shocks to the south and to the relatively rich do better, but they primarily stimulate the nonfarm sector. When a more satisfactory set of assumptions inherent in a broader national definition of tradability is used, basic foods begin to become nontradables. And decomposition of the estimated growth multipliers shows that the farm nontradables sector accounts for the majority of net additional income growth from exogenous increases in incomes in the tradables sector, whether or not the initial income stimulus was inside or outside of the farm sector. Income shocks directed to the poor and the north stimulate the farm sector in particular once basic foods are designated nontradables.

Conclusions

Demand is income-elastic, in increasing order of magnitude, for prepared foods and beverages, for local nontradable nonfarm items, and for nonlocal nonfoods. Nonfarm items as a whole, including prepared and manufactured foods, accounted for 23 percent of average whole-sample expenditures and 35 percent of increments to expenditures. These items are important for the future but are numerically of much smaller importance than basic staples. Basic food staples account for the majority of both average and incremental expenditures in the study zones. For the sample as a whole, coarse grains alone accounted for 59 percent of average expenditures and 53 percent of incremental expenditures. The comparable figures for the poor and the Sahelian north are even stronger. These findings are not atypical of the Sahel and suggest that the tradability assumptions made about coarse grains are central to one's view of the potential for productive demand-side stimulus to rural development in these areas.

The tendency in earlier literature to assume that these coarse grains were freely traded resulted in the implicit assumption that respending of rural incomes on these items was a "leakage," in the sense that it simply displaced exports of grain from rural areas or encouraged further imports of grain to the study zone. This assumption erroneously leads to the conclusion that the scope for additional demand-led growth from an initial supply-side shock, as in growth linkages theory, is low.

The concept of a catchment area implicit in any regional model is central to estimates of the level of growth multipliers. It is not an arbitrary choice, since only one set of assumptions is likely to be consistent with the world as it really is. If local income growth in northern Burkina Faso does truly elicit additional sorghum shipments north from southern Burkina Faso without major price increases, as implicitly hypothesized in the assumptions underlying the national level of tradability, then this should be the set of assumptions chosen. Failure to make appropriate assumptions leads to strategies that underemphasize the employment-creating potential of income growth in the rural tradables sector through growth linkages.

Defining tradability at the national rather than the local level is a reasonable compromise. While household-level data of the sort used do not capture all the linkages implicit in adopting a national catchment area, they do vastly improve estimation of how initial supply-side shocks to the tradables sector can further stimulate economies that are still semi-open, because of remoteness and poor infrastructure. The estimates of linkages for the national level of tradability in the present study are presumably too low, since some of the truly national linkages are not captured in the household data available. This only strengthens conclusions that show high multipliers.

Ignoring the central fact of the nontradability of much of agriculture under a larger catchment area can lead to growth strategies that are inefficient. Estimation of farm growth multipliers under the national definition of tradability leads to an overall estimate of $1.88 in net new income from intermediate demands and consumer respending of an initial $1.00 in additional income from the farm tradables sector. The net new income from targeting income shocks to specific subgroups ranges from $ 1.45 to $2.31.

This points to a central difference between Burkina Faso and areas where food-grains are freely tradable because of low relative transfer costs to and from the outside world. Demand stimulus in rural Burkina Faso is capable of inducing considerable employment within the farm sector itself. Provided that a sustainable way is found to achieve the initial boost in incomes in the rural tradables sector, and provided that the supply of farm nontradables, such as grain, is somewhat price-elastic, foodgrain production itself will provide substantial new employment. This could be the case if cash crop development were to both stimulate demand and, through better infrastructure and input distribution, increase the elasticity of supply of foodstuffs.

Shocks to rural tradable nonfarm incomes are slightly more efficient at boosting overall value added than income shocks to farm tradables. The estimated multipliers were robustly greater (about 7 percent) than those for shocks to farm tradables. The choice of which rural tradables sector to stimulate depends on comparative advantage and not growth multipliers. Both farm and nonfarm tradable income shocks increase net additional employment outside the stimulated sector principally through consumption linkages to the farm nontradables sector.

The tradability issue not only affects one's view of the scope for agricultural growth linkages to complement growth strategy, but also one's view of the consistency between growth and equity policies. In the model, the relative importance of the farm nontradables sector in providing additional value added from growth linkages increases greatly with the size of the catchment area considered. When only local linkages are considered, which is a mistake, income shocks to the rich create more overall income growth than shocks to the income of the poor, confirming the conventional wisdom from Asian linkage studies that there is a trade-off between growth and equity arising even from differential consumption patterns of the poor and rich.

This relationship is reversed when the more appropriate, national catchments are assumed, because the poor consume a relatively higher proportion of items that are tradable locally but not over national borders. At the local definition of tradability, the initial $1.00 stimulus to the farm incomes of the poorest one-third of households produced an additional net income of $0.25 beyond the initial shock, while the same figure for the richest one-third was $0.41. With the alternative national definition and assumptions, the net additional increment was $2.18 for the poor and $ 1.45 for the rich.

To summarize, neglect of the fact that so many rural people in countries such as Burkina Faso consume items that are nontradable because of high transport costs may have led, logically, to growth strategies that tend to ignore the potential to achieve multiplied growth through alleviation of demand constraints for farm items. The initial engine of this growth still needs to come from the rural tradables sector. For a given initial income influx, growth linkages in the farm sector are nearly as efficient as in the nonfarm sector. Supply-side policies should continue to focus on farm products that are likely to be competitive on expanded regional and world markets, with the knowledge that the true returns in terms of net value added to the economy are potentially much higher than the income shocks themselves. Realizing the growth potential offered by strong demand linkages to the farm sector, from both farm items themselves and nonfarm items, will require a price-elastic supply of those things that rural people wish to consume more of as their incomes go up. Reaping the fruits of export-led growth will also require policy attention to increasing the supply of nontradable wage goods such as coarse grains.

Ultimately the dynamic rural consumer items will be those whose demand is currently income-elastic, such as services, radios, batteries, beverages, fruits, vegetables, meat, and dairy products. However, the high average expenditure share for starchy staples suggests that - despite having slightly income inelastic demand - they can form either a prime source of growth, or a major bottleneck to it. The elasticity of the supply of staples with respect to price will determine whether demand stimulus will lead to real income growth or to the choking off of growth through rising relative prices of nontradables. In the Burkina Faso case, where nontradables are probably major wage goods such as food, this could mean that the cost of labor and therefore the basic cost of production of farm tradables would also rise.