77932 Is the informal sector constrained from the demand side? Evidence for six West African capitals* Marcus Böhme Kiel Institute for the World Economy Hindenburgufer 66 D-24105 Kiel Germany Phone: +49-431-8814-571 e-mail: marcus.boehme@ifw-kiel.de Rainer Thiele (Corresponding Author) Kiel Institute for the World Economy Hindenburgufer 66 D-24105 Kiel Germany Phone: +49-431-8814-215 Fax: +49-431-8814-502 e-mail: rainer.thiele@ifw-kiel.de 1 Summary Employing a unique dataset that covers households from six West African capitals, this paper provides new evidence on the demand for informal sector products and services. We first investigate whether demand linkages exist between formal and informal products and distribution channels. In a second step, we estimate demand elasticities based on Engel curves. We find strong demand-side linkages between the formal and informal sector, with the exception that informal goods are hardly bought through formal distribution channels. The estimated demand elasticities tend to show that rising incomes are associated with a lower propensity to consume informal sector goods and to use informal distribution channels. Keywords: Informal sector; formal-informal linkages; Engel curve estimates; Africa; West African capitals 2 Acknowledgements We gratefully acknowledge funding from the World Bank’s Multi Donor Trust Fund (MDTF) “Labor Markets, Job Creation, and Economic Growth: Scaling up Research, Capacity Building, and Action on the Ground� for the project on “Unlocking potential: Tackling economic, institutional and social constraints of informal entrepreneurship in Sub-Saharan Africa�. Special thanks are due to AFRISTAT and DIAL, in particular Constance Torelli and François Roubaud, for providing and preparing the datasets used in this paper. We also thank three anonymous for very helpful suggestions. 3 1. INTRODUCTION In urban Sub-Saharan Africa (SSA), formal employment covered by labor legislation and social protection schemes is the exception rather than the rule. By far the largest part of urban employment is generated by micro and small enterprises, and (self-) employment in those enterprises can be considered informal by almost any definition one might want to apply. The informal sector is characterized by a high degree of heterogeneity comprising both low and high return activities. If markets fail to equalize returns, this can imply substantial welfare losses where poor urban households are prevented from escaping the lower tier of informal employment. Among the constraints that may bring about persistent differentials in returns, supply side factors such as capital market failures and entry barriers have received considerable attention in the literature (e.g. de Mel et al. 2008; Grimm et al. 2011). It has been less well recognized that the evolution of the informal sector is also shaped by the demand side, in particular by the structure of final demand as well as linkages to the formal sector. The literature on the structure of demand has mainly been descriptive. It has not only distinguished informal and formal products and services, but also formal and informal customers or households, typically identified by the (main) sector of occupation of the household head (formal or informal). A core proposition of this literature has been that informal and formal products will often have an overlapping customer base (Sethuraman 1997). Such overlaps may reflect complementary or competitive product markets. The most common example of a complementary market occurs when the informal sector sells formal sector products. In competitive markets, the two sectors compete within the same product market, and the informal sector may for example retain market shares by charging lower prices. The rather limited evidence for Sub-Saharan Africa tends to confirm the notion of an overlapping customer base. Covering a sample of 13 Sub-Saharan African countries, Xaba et al. (2002) find rather strong inter-linkages in the final product market, with each sector 4 being a strong supply as well as demand base of the other sector. Similar results are obtained by Grimm and Günther (2006) for the case of Burkina Faso. By contrast, Fortin et al. (2000) suggest various reasons why working in the informal sector will raise the probability of buying products in the informal market, thereby limiting the demand overlap. For instance, those working in the informal sector may have an informational advantage about how and where to obtain informal goods and services. In accordance with this reasoning, Livingstone (1991) finds that in Kenya informal goods target mainly low-income consumers, while Reilly et al. (2006) obtain an inverse relationship between purchases from informal markets and income. For a sample of 24 African countries, La Porta and Shleifer (2011) show that informal firms mostly sell to informal clients for cash, which they attribute to differences in quality between formal and informal firms. Of crucial importance for the economic prospects of informal entrepreneurs is the elasticity of demand for their products, which in turn depends on the strength of formal-informal linkages. Again, evidence on this issue is scarce. The only exception in the African context is Lachaud (1990) who shows that rising wages lead to a lower propensity to consume informal sector goods. In a similar vein, D'Haese and Van Huylenbroeck (2005) find evidence that supermarkets create fierce competition with local agricultural sales in South Africa. With rising income households tend to purchase their goods at supermarkets because they are able to offer a broader variety and a higher quality at lower prices. Even though D'Haese and Van Huylenbroeck do not address the informal sector directly it can be assumed that local agricultural sales are often informal. This paper aims to broaden the evidence on the characteristics of demand for informal sector products and services in Africa. We extend the literature in various ways. First, by using fully comparable data for a sample of six West African countries, we provide a comprehensive set of demand elasticities based on Engel curves. Second, in contrast to previous papers, our dataset allows us to consider imported goods in addition to domestically produced formal and informal 5 goods, and to account for informal and formal distribution channels. Hence, we are able to categorize goods along two dimensions – their origin (formal, informal, imported) and their point of sale. For example, a good such as industrial bread, which is produced by the formal sector, may either be sold formally in a supermarket or informally by a market hawker. Third, we address methodological challenges such as the potential endogeneity of income and the nonlinearity of Engel curves. This has been done before (e.g. Gibson 2002; Kedir and Girma 2007), but not in the context of the informal sector. The overall objective of the paper is to investigate the extent to which the informal sector is constrained from the demand side in the sense that the demand for informal goods and distribution channels is income-inelastic. In combination with supply side factors such as credit constraints, inelastic demand would severely limit the prospects of those working in the informal sector. If informal goods are mainly bought by poor informal households via informal distribution channels, due to lower information costs or quality differences, one would expect low demand elasticities throughout. If, by contrast, formal-informal linkages play a significant role, for instance because formal customers value certain characteristics of the informal sector such as higher flexibility in terms of longer opening hours, the picture could be much more nuanced, with certain segments of the informal sector benefiting from strong demand by richer households who earn their income in the formal sector. The remainder of the paper is structured as follows. Section 2 introduces the dataset employed in the empirical analysis and presents descriptive evidence on whether there are demand-side linkages between the formal and informal sector. Section 3 derives some hypotheses concerning the demand elasticities and describes the Engel curve methodology, while Section 4 discusses the estimated elasticities. Section 5 summarizes our main results and provides some concluding remarks. 6 2. DATA AND DESCRIPTIVE ANALYSIS (a) The dataset We use data provided by the “Enquêtes 1-2-3�. This survey was carried out between 2001 and 2003 in seven economic capitals of the West African Economic and Monetary Union (WAEMU). It consisted of three integrated phases for a representative set of households. The first phase of the survey was constructed as a general labor force survey, interviewing formal and informal workers and entrepreneurs. It provides detailed information about individual socio-demographic characteristics and employment. In identifying informal activities, the 1-2-3 surveys follow international statistical guidelines, which suggest that informal sector employment should be defined in terms of characteristics of the enterprise or production unit such as size and different legislative criteria (Hussmans 2004).1 Specifically, the 1-2-3 surveys define informal enterprises as small production units that (a) do not have written formal accounts and/or (b) are not registered with the tax administration. The second phase of the survey interviewed a sub-sample of the informal production units identified in phase one. The focus of this phase was on characteristics of the entrepreneurs and their production units. It also contains information on input use, investment, sales, profit as well as the unit’s forward and backward linkages. The third phase, on which the subsequent analysis will mainly rely, collected data on household expenditure including the point of sale. Expenditures were recorded based on a classification of 315 different products and services. The technique of registration varied according to the periodicity of the purchase. While food expenditures were registered on a daily basis for 15 days, for other types of expenditure such as clothing, housing, health, transport and education a retrospective module was used. All expenditure aggregates are recorded at the household level, annualized and given in local currency units. A two-stage random sample design was applied based on an updated general population census of each country (Amegashie et al. 2005). Area codes were used as the primary 7 sampling unit, of which 125 were selected for each city. Households were the secondary sampling unit, of which 20 (24 in Benin) were drawn from each primary unit. Data was then collected for each household member. The 4200 households included in phase three constitute a representative subsample drawn from the 15300 households of phase one. The data collected in phase one permitted an additional stratification based on income and sector of activity of the household head in phase three. This constitutes one major advantage of using the integrated 1-2-3 survey, because it allows us to distinguish formal and informal households and thus to test whether these two groups differ in their demand patterns as suggested by Fortin et al. (2000). A further strength of the 1-2-3-surveys is that they used exactly the same questionnaire and were conducted more or less simultaneously in the seven economic capitals of the WAEMU, rendering the datasets fully comparable. Finally, being coordinated by AFRISTAT and DIAL and financially supported by the European Commission, the French Ministry of Foreign Affairs and the World Bank, the surveys were elevated into the status of official data, which should add to the credibility of the results based on them. Table 1 shows summary statistics of selected socio-demographic household characteristics for the six West African capitals under consideration.2 Most interestingly in the context of this paper, between 40 and 60 percent of the household heads receive their primary income from informal sector activities, classifying the respective households as informal. The share is lowest in Dakar, the capital of Senegal, the richest country in the sample. The primary wage income of informal households is on average roughly 70 percent of the income of formal households. At the country level, informal households have throughout less income than formal households; this difference is significant (except for Mali) but varies in size across countries. Table 1 about here 8 The other major household characteristics listed in Table 1 also exhibit a considerable degree of variation across countries. Average household size, for instance, is by far highest in Dakar due to a strikingly large number of adult members.3 The mainly Christian capitals of Togo and Benin are characterized by high rates of primary school completion among household heads and a rather high share of female-headed households, whereas the opposite is true for the mainly Muslim capitals of Burkina Faso, Mali and Niger. (b) Structure of demand To analyze the structure of demand for goods and services, we aggregate annual expenditures in two different ways. First, we apply a conventional sectoral classification, where expenditures are allocated to eleven different categories.4 This classification has the advantage that it can directly be matched with the survey information and closely resembles what has been done in previous studies, allowing a comparison of our results. This has to be weighed against one important disadvantage, namely that the sectors only roughly correspond with the distinction between formal and informal goods we are interested in. One can argue, for instance, that Food and Non-Alcoholic Beverages includes mainly informal products, whereas Transportation and Communication supplies mainly formal goods and services, but at the same time one has to acknowledge that within categories there are notable exceptions such as informal taxi services. Second, we distinguish four types of expenditures: on domestically produced formal goods, domestically produced informal goods, imported goods and services. In principle, this option is superior to the sectoral classification as it directly captures the distinction between formal and informal goods. However, the households in the survey were not asked whether they bought any specific product from the formal or informal sector. The best available alternative was to use certain characteristics of the products that households purchase to sort them into the two different categories. Specifically, agricultural, artisan and traditional products were assumed to be produced informally, whereas capital intensive, technologically advanced and industrial products were 9 assumed to be produced formally. The first category mainly comprises food products such as traditionally made bread, while the second category consists mainly of electricity, fuel, construction materials, household articles, clothing and footwear, as well as certain food products such as industrially produced bread or canned meat. With this procedure, we cannot rule out that specific products are sorted into the wrong categories, but are confident that the aggregates constitute a meaningful representation of informal and formal goods. Still, findings based on this categorization have to be interpreted cautiously. In contrast to domestic formal and informal goods, demand for imports and services can readily be identified from the survey. Households can with an acceptable margin of error name the country of origin of goods purchased due to the packing and labeling. One central characteristic of services is the quasi-concurrence of production and purchase. Hence, knowing the distribution channel is sufficient to distinguish between formal and informal services. The questionnaire asked consumers about the location of their purchases such as explicitly formal enterprises, supermarkets and the public sector, which are assumed to be formal points of sale, and household production, mobile traders and public markets, which are assumed to be informal points of sale. This allows us to distinguish purchases via formal and informal distribution channels for each of the demand categories defined above. As can be seen in Table 2, food products and non-alcoholic beverages account for roughly 30 percent of annual household expenditures throughout the sample. If non-frequent purchases are excluded, i.e. only monthly expenditures are used this share rises to over 70 percent for all West African capitals. Housing and Transport and Communication constitute the next-biggest positions, accounting for 10-18 percent and 10-16 percent of total expenditures respectively. The structure of demand does not seem to vary in a systematic way across the sample countries. We do, for example, not find higher shares of food expenditures in poorer countries such as Mali and Niger than in richer countries such as Senegal and Benin. While one might argue that this finding works 10 against Engel’s law, our interpretation is that even the richer countries in the sample are still too poor to exhibit significantly lower food expenditure shares. Table 2 about here This interpretation is also in accordance with our within-country finding that the expenditure share of food and non-alcoholic beverages even rises slightly from the first to the second income quintile and only then starts to drop in accordance with Engel’s law (Table 3).5 In all countries, expenditure shares tend to fall across quintiles for Housing6 as well as Hotels and Restaurants, while they tend to rise substantially for Health and Education as well as Transport and Communication. From Table 4 it appears that the informal sector is the dominant point of sale.7 The only notable exception is Health and Education where services are almost exclusively distributed through formal channels. For Gas, Electricity and Water and Transport and Communication we see a rather balanced relationship between formal and informal expenditure shares. Households turn to the formal sector for purchases of capital intensive products such as private vehicles and for the use of public transportation, and to the informal sector for rather inexpensive transportation-related goods such as bicycles and spare parts. In the food and beverages sector, formal outlets such as supermarkets appear to be virtually non-existent. Tables 3 and 4 about here Turning to the classification by expenditure categories, a clear pattern emerges where the informal distribution channel predominates for all four types of expenditures (Table 5). This is most obviously the case for informal goods, which are hardly ever sold via formal distribution channels. Comparing the six capitals, demand for formal goods is found to be highest in Dakar, but even there the bulk of products are distributed informally. Overall, formal-informal demand relationships do not appear to vary systematically with income per capita. For example, the demand share for informal goods is lower in the poorest country, Niger (15 percent), than in the richest country, Senegal (18 percent). This lack of a clear relationship may reflect that cross-country differences in 11 per capita income are not sufficiently pronounced. Strikingly, Togo has the lowest expenditure shares on both formal goods and goods distributed via formal distribution channels and at the same time exhibits by far the worst governance indicators among the countries under consideration. Import shares vary considerably across countries, ranging from 15 percent in Dakar to 38 percent in Niamey. Around 30 percent of all imported goods in Cotonou, Ouagadougou and Bamako are produced in other (mainly neighboring) African countries, while this figure is substantially higher in Niamey (46 percent) and substantially lower in Dakar (9 percent). Table 5 about here Table 6 reveals that expenditures by informal households as defined above are by no means restricted to informal goods and informal distribution channels. In Cotonou, for example, these households account for more than a third of total expenditures on formally distributed formal goods (36.8 percent) and formally distributed imports (34.6 percent). Irrespective of product category, distribution channel and city, their expenditure share is in no case lower than 20 percent. Conversely, the fact that the expenditure share of informal households in no case exceeds 60 percent implies that formal households are important buyers of informal goods and goods distributed through informal channels, which points to a strong overlapping customer base. Overall, this section has shown that significant formal-informal linkages exist in the six West African capitals. They may well be strong enough to affect the pattern of estimated demand elasticities, to which we will turn in the next section, in a way that eludes the notion of a simple formal-informal sector dichotomy. Table 6 about here 12 3. ESTIMATION OF ENGEL CURVES (a) Hypotheses While demand estimation is often primarily concerned with quantities, consumers also face a quality choice (Deaton 1988, Blundell and Stoker 2005). Consequently the observed expenditure patterns will be conditioned by price, quantity and quality of the available products. Merella (2006) has shown that Engel Curves depend on the assumption of constant quality. With increasing quality of a product the marginal utility of this same product would not decline. Based on this assumption the author asserts that increasing income will first affect the quantity of products bought and only thereafter the shift between products due to quality differences. More generally, poor households will be concerned primarily with subsistence and therefore quantity while rich household s’ consumption is driven by quality differences in goods. This theory has implicitly been corroborated in prior studies (e.g. Wan 1996) where food and shelter – which are necessities and are expected to show an elasticity coefficient that is greater than zero but less then unity – have been found to be treated as a luxury by extremely poor households. Recently, Banerji and Jain (2007) introduced a new perspective by arguing that the dynamics of the informal sector are also driven by changes in consumer demand. At the core of their argument stands the observation that there is a marked quality difference between formal and informal goods and services. Accordingly, the informal sector caters to a consumer base that is not able to pay high prices for high quality. From these theoretical considerations a testable hypothesis can be derived. If Banerji and Jain (2007) are correct and formal and informal goods and services differ markedly in their quality, we have to expect significantly different income elasticities for the same goods produced or sold by formal and informal enterprises. If, in addition, Merella (2006) is right then we should observe smaller income elasticities for low quality products, i.e. informal sector products. To obtain a rough indication of whether quality differences exist, we compared the unit prices between informal 13 and formal points of sale following the trade literature (e.g. Fontagne et al. 2006). Out of the product and service categories we were able to compare on a disaggregated level, 46 Percent showed significantly different unit prices, and in 96 percent of these cases the mean unit price was found to be lower for the informal point of sale. While these results are suggestive of quality differences, we have to interpret them cautiously given that the aggregation and the enormous variance in prices could drive the differences. In addition, the fairly strong formal-informal linkages we find suggest that demand for informal goods and distribution channels is not exclusively driven by quality dualism, but may also reflect positive characteristics of the informal sector such as flexibility in terms of longer opening hours or less paperwork that are valued by formal consumers. The existing linkages may lead to a more nuanced picture of elasticity estimates. (b) Methodology To answer the question of whether customers behave differently vis-á-vis formal and informal products and distribution channels we estimate demand elasticities for different aggregated production sectors and different types of expenditures as defined above. As predicted by neoclassical consumer theory private demand for goods and services is a function of disposable income and prices. Since the data we work with is cross-sectional, estimations have to be simplified assuming prices to be constant across observations. The corresponding specifications, known as Engel curves, represent the evolution of Marshallian demand functions for a particular good or service category as income varies, holding the prices for all goods constant. Deciding on whether to estimate the Engel curves by means of a simultaneous equation approach or equation by equation involves a trade-off. On the one hand, estimating a demand system would allow us to account for the restrictions required by utility-based demand theory such as the adding- up criterion. A Breusch and Pagan test in a number of cases rejects the null hypothesis of no correlation among the error terms of different equations, pointing to interdependencies between 14 demand categories. On the other hand, the equation-by-equation approach has the advantage that mis-specified equations do not affect overall results, which we consider to be especially relevant given the strong indications of differences in functional forms between demand categories (see below). We therefore decided to present estimates equation by equation as our preferred approach, but also estimated demand systems for both sectors and expenditure categories. The results of these estimations are shown in the Appendix; a brief comparison of the two sets of results is given below. Since non-parametric Kernel density regressions as well as formal Ramsey RESET tests point to the absence of non-linearities in all but two cases (see below), we adopt a simple linear double- logarithmic model as our base specification.8 The equation fitted takes the following form: ∑ with: ∑ where xij is the log of expenditure of household i on product category j, xi is the log of household income, Hi is the log of household size in adult equivalents, mi are L different household characteristics including age, sex and religion of the household head, household composition variables such as a dummy for elderly members, an indicator of primary school completion of the household head, and sector of employment of the household head, and Di are district dummies. While we include the additional control variables primarily to account for household heterogeneity, the occupation variable also allows us to test the hypothesis (e.g. Fortin et al. 2000) that being employed in the informal sector raises the likelihood of consuming informal goods and services. Concerning household size, we performed robustness checks using simple household size instead of 15 equivalence scales as predictor variables, which hardly affected our results. Likewise, results do not change when we employ years of schooling rather than primary completion as an indicator of education. When estimating the Engel curves, several challenges arise. One problem is that observed income might not be the main driver of expenditure behavior. More specifically, we have to consider the possible seasonal volatility of employment and earnings which casts doubt on the appropriateness of monthly income as a representation of disposable annual income. Thus, we use total expenditure as a proxy of income, even though our data contains information about declared monthly income from primary and secondary employment as well as income from other sources such as remittances and assets for all household members. A further justification for using total expenditure instead of total income is based on the permanent-income hypothesis. Accordingly, expenditure will be conditional upon long-run income rather than transitory income, even though it has to be conceded that consumption smoothing in West Africa is likely to be far from perfect as a result of liquidity constraints.9 While the use of total expenditure can lead to biased or even inconsistent estimation results given that it is only a proxy of income, various authors (e.g. Lewbel 1996; Gibson 2002) have argued that this bias tends to be small compared to the bias introduced by using transitory income. Using total expenditure does, however, introduce an econometric complication, namely the possible correlation of the independent variable with the error term since by definition our dependent variable, the expenditure on different product categories and types, will always be part of the explanatory variable. This possible simultaneity bias calls for the use of instrumental variables. In addition, measurement error in household survey data is a well-known problem in demand estimation (Liviatan 1961, Griliches 1974, Theil 1979, Keen 1986, Lewbel 1996, Hausman 2001). This problem can also be mitigated by using instrumental variables. Based on the classical errors-in variables assumption that presumes a correlation between the observed variable and the error term one would expect the OLS estimator to be closer to zero than the true estimator represented by a 16 valid IV coefficient. Several authors have pointed to exceptions to this rule in the context of demand analysis (e.g. Keen 1986; Liviatan 1961). Most recently, Gibson and Bonggeun (2007, p. 479) have asserted that “[…] only some form of correlated error could cause� the coefficient “to be biased downwards�. In the context of consumer behavior most of the previous studies have instrumented total expenditure by monthly income (e.g. Lewbel 1996, Kedir and Girma 2007). We also did so, but additionally employed a wealth index, given that it seems plausible to assume that household wealth will also influence the observed expenditure patterns, perhaps even more so than current income. We constructed a wealth index, using a principal components analysis along the lines of, for example, McKenzie (2003) and Filmer and Pritchett (2001). In doing so, we converted all available wealth and asset variables (housing characteristics, access to infrastructure and durable assets) into binary ones, except for the number of rooms in the household. One of the main advantages of such an index is that it reduces measurement error. Since reporting errors in household income and household wealth cannot be ruled out, we additionally considered an instrumentation strategy using variables that describe the activity portfolio of the household (number of fixed-contract wage earners in the household; share of wages/salaries from public sector, primary sector, industrial sector, commercial sector and service sector) as well as job characteristics (sector of occupation; type of job; type of contract; type of wage payment; hours worked) of the household head and the main wage earner in the household.10 We started our empirical analysis by employing ordinary least-squares (OLS) estimation techniques. The explanatory power of this simple model is rather good; in almost all cases up to one half of the observed variation can be explained. As already noted above, a non-parametric analysis of the data pointed to a non-linear relationship in some cases, which was confirmed by a formal Ramsey RESET test. Specifically, we had to reject the null hypothesis of no omitted variables and therefore our linear specification for food and non-alcoholic beverages at the 1 percent level and for 17 transport and communication at the 5 percent level for all countries in favour of a more flexible specification. To account for these test results we included a quadratic term of the log of total expenditure for these two categories. Using a Hausman test we can reject the null hypothesis of endogeneity of the log of total expenditure for all sectors apart from Food and Non-alcoholic Beverages as well as Electricity, Gas and Water. By contrast, the Hausman test points to potential endogeneity for all four expenditure categories (formal goods, informal goods, imports, services) at least in some of the countries. Accordingly, we performed two-stage least-squares (2SLS) estimations for our linear specifications when this was suggested by the test results. We found no evidence of endogeneity in the non-linear specifications for Food and Transport and therefore did not employ the 2SLS approach for this setup. To evaluate the strength of our proposed instrumental variables we applied the test suggested by Staiger and Stock (1997) and Stock and Yogo (2003). According to this test, income and the wealth index are strong instruments, The F value of the wealth index being somewhat higher throughout, whereas the indicators of activity portfolio and job characteristics generally turn out to be weak instruments. 4. ESTIMATION RESULTS Our OLS estimates at the sectoral level displayed in Table 7 are in line with the findings of prior studies for East Africa (Massel and Heyer 1969, Ostby and Gulilat 1969, Humphrey and Oxley 1976, Okunade 1985, Teklu 1996) and for other developing countries such as India and China (e.g. Tiwari and Goel 2002, Chern and Wang 1994). In particular, we find Food and Non-alcoholic Beverages as well as Alcoholic Beverages and Tobacco to be necessary goods as suggested by Engel’s law. An increase of one percent in disposable income would on average lead to an increase of expenditure on food and non-alcoholic beverages of 0.77 percent in the six countries.11 Clothing and Shoes, Furniture, Health and Education, Transport and Communication as well as Leisure and 18 Culture turn out to be luxury goods in all countries under consideration. The low demand elasticities for hotels and restaurants may appear somewhat surprising, but the previous literature is inconclusive as to whether food-away-from home - for hotels, there is no comparable evidence - constitutes a necessity or a luxury good (Byrne et al. 1995; Min et al. 2000). Table 7 about here When comparing these results with the demand system estimation shown in the Appendix, it turns out that the pattern of elasticities is very similar, even though in a number of cases the point estimates differ quite substantially between the two approaches. Both approaches suggest that Clothing and Shoes, Furniture, Health and Education, Transport and Communication as well as Culture and Leisure are luxuries, whereas Food and Non-alcoholic Beverages, Housing as well as Hotels and Restaurants tend to be necessities. In the lower part of Table 7 we report selected results for the 2SLS estimations using the wealth index as an instrumental variable.12 According to the tests performed, Food and Non-alcoholic Beverages, as well as Electricity, Gas and Water are the only sectors where we are advised to perform an instrumental variable estimation. For these sectors, differences between OLS and 2SLS estimates turn out to be substantial. Food and Non-alcoholic Beverages are shown to follow Engel’s law even more clearly when using 2SLS as indicated by lower budget elasticities in all countries under consideration, while Electricity, Gas and Water switches from being a necessity to being a luxury good. Among the additional control variables, household size uniformly has a positive and significant influence on food expenditures. Its impact on expenditures is significantly negative throughout for Transport and Communication and in some countries also for Shoes and Clothing as well as Furniture and Household Maintenance. These findings are in line with economies of scale concerning these commodities. In several cases, the gender of the household head turns out to be 19 another important determinant of expenditures, pointing to gender-specific preferences and intra- household bargaining. All else being equal, households headed by a male tend to spend less on Food as well as Health and Education, and more on Transport and Communication as well as Leisure and Culture. We also find that being Catholic or Muslim has a significantly negative effect on alcohol and tobacco expenditures in all countries except Mali. The quadratic OLS estimations are shown in Table 8. Recall that the RESET test points to a non- linear specification only in the case of Food and Beverages as well as Transport and Communication. We find that the elasticity function for Food and Beverages is concave, which is in line with the saturation hypothesis. We also observe that the turning point (the maximum) is located in the 10th expenditure decile in all countries; it is comparatively higher in Benin and Senegal than in the other countries. Given that the turning point is located very close to the maximum of the income range of households, our finding arguably does not invalidate Engel’s law. For Transport and Communication the elasticity function is first falling with income and after a certain threshold point increasing again.13 In this case the turning point is a minimum, which is lowest in Benin and Togo and highest in Niger, but practically irrelevant as it is located outside the observable income range. Table 8 about here The estimates shown in Table 9 allow us to assess the hypothesis of quality dualism as they capture the distinction between formal and informal goods and distribution channels. A very clear pattern emerges for formal goods: In all countries the demand elasticity is substantially above unity for formal provision and substantially below unity for informal provision. Imported goods and services are also uniformly seen as luxuries when distributed via formal retailers, and in some countries (Mali, Niger, Senegal) the demand elasticity is above unity even in case of informal distribution. Informal goods exhibit elasticities far below unity across the board if they are distributed via informal distribution channels, whereas the few formal sales appear to be luxuries in Benin, Mali, 20 Senegal and Togo. Aggregated over the two distribution channels, informal goods are in all countries considered necessities, while the opposite is true for services. Overall, our evidence broadly supports the hypothesis of quality dualism, which is also corroborated by the demand system estimates displayed in the Appendix. Table 9 about here Graphically these results are illustrated for Benin in Figure 1. The steep slope of the fitted values of formal distribution represents clearly the higher elasticity of this distribution channel compared to the informal channel. Using a simple Chow test we find differences in slopes and intercepts between formal and informal distribution channels to be significant throughout, for informal goods at the 10 percent level and for all other categories at the 1 percent level of significance. Regarding the household characteristics, we cannot observe clear tendencies of influence, which may at least partly reflect the high level of aggregation. Most notably, employment of the household head in the informal sector is in almost all cases statistically insignificant. This corroborates the above finding that informal households reveal no particularly strong preference for informal goods and services. Figure 1 about here Up to now, we have considered the four expenditure categories only in the aggregate. This is because a further disaggregation dramatically reduces the number of observations. We nonetheless specifically looked at Food and Beverages as well as Transport and Communication, for which the number of observations is largest.14 But even in these two categories we partly run into data problems. For Food and Non-Alcoholic Beverages, only between 1 and 7 percent of all purchases were located in the formal sector, i.e. products that have been distributed by formal vendors. The data restriction becomes even more severe if one focuses on a single expenditure category such as formally produced domestic products. The corresponding results thus have to be interpreted very cautiously.15 For all three expenditure categories considered (services are not recorded), the estimated expenditure elasticities for formally distributed food products reveal a very mixed picture, 21 which arguably reflects to a large part the lack of sufficient data. By contrast, a stable pattern of elasticities below unity appears for informal distribution channels, which corroborates the findings obtained at the aggregate level. Aggregated over distribution channels, the three product categories are found to be necessities in all but two cases. For Transport and Communication, we disentangle different types of products and their distribution channels. While formal and imported goods are composed mainly of capital intensive items such as private vehicles, services are composed of public and private transport modes such as public buses and taxies. Imports and services account for the bulk of expenditures in this category. If significant, estimated demand elasticities at formal points of sale tend to exceed unity. Most notably, formally distributed formal products such as private vehicles turn out to be strong luxuries in the two richest countries, Benin and Senegal. In contrast to Food and Beverages, even the elasticities for informal distribution channels are partly above unity, suggesting that informal sales will not necessarily fall with rising incomes. 5. CONCLUDING REMARKS In this paper, we have offered a descriptive overview of demand in six capitals of the West African Economic and Monetary Union as well as an analysis of budget elasticities for different sectors and distribution channels. Our main findings are that (i) there is support for linkages between the formal and informal sector regarding the channels through which goods are distributed, with the exception that informal goods are hardly bought through formal distribution channels; (ii) there appears to be a strongly overlapping customer base between the formal and informal sector; (iii) rising incomes tend to lead to a lower propensity to consume informal sector goods and to use informal distribution channels. 22 We find little systematic variation in demand structures across countries. Most notably, expenditure shares for informal goods and informal distribution channels do not appear to be consistently higher in the richer sample countries. Macro indicators such as government effectiveness or trade openness are also hardly related to cross-country differences in demand. An interesting correlation can only be observed for Togo: it exhibits by far the worst governance indicators and at the same time has the lowest expenditure shares on both formal goods and goods distributed via formal distribution channels. Our elasticity estimates imply that overall the development of the informal sector in West Africa will most likely be constrained from the demand side, which is in accordance with the hypothesis of quality dualism, with the informal sector being characterized by low quality. However, the pattern is not uniform, underscoring the notion of a heterogeneous informal sector put forward in studies of the supply side. Along expenditure categories, elasticities of the informal distribution channel are much higher – in some cases (e.g. Mali) even above unity – for imports and services than for domestically produced informal as well as formal goods, suggesting that importers and buyers of services tend to value certain characteristics of informal distribution channels and thus do not necessarily turn to the formal sector when their incomes rise. The overall demand bias against the informal sector suggests that the majority of poor informal households, for example those who produce or sell food, would be affected less than proportionately by recessions. The implications for their welfare in the longer run depend on how easily they can switch to more income responsive activities. As long as the high entry barriers previously identified for West Africa (Grimm et al. 2011) continue to limit the adjustment possibilities of informal entrepreneurs, the growth process of the urban economy is unlikely to be pro-poor. In the food sector, which accounts for a large share of informal activities in all six capitals, future competition by supermarkets may even further reduce the room of maneuvering for informal households. 23 As concerns future research, the next step would ideally involve a further disentangling of the relation of quality and quantity by using more homogenous goods and panel data. This would allow us to mitigate the well-known difficulties caused by the aggregation of broad product groups. Another interesting area for future research would be to investigate in more detail why formal- informal demand linkages exist. 24 BIBLIOGRAPHY Amegashie, F., Brilleau, A., Coulibaly, S., Koriko, O., Ouedraogo, E., Roubaud, F., & Torelli, C. (2005). La conception et la mise en oeuvre des enquêtes 1-2-3 en UEMOA, les enseignements méthodologiques. Stateco, 99, 21-42. Banerji, A., & Jain, S. (2007). Quality dualism. Journal of Development Economics, 84, 234-250. Blundell, R., & Duncan, A. (1998). Kernel regression in empirical microeconomics. Journal of Human Resources, 33 (1), 62-87. Blundell, R., & Stoker, T. M. (2005). Heterogeneity and aggregation. Journal of Economic Literature, 43 (2), 347-391. Byrne, P., Capps, O. Jr., & Saha, A. (1996). Analysis of food away from home expenditures for U.S. households, 1982-89. American Journal of Agricultural Economics, 78, 614-627. Chern, W., & Wang, G. (1994). The Engel function and complete food demand system for Chinese urban households. China Economic Review (5), 1, 35-57. Deaton, A. (1988). Quality quantity, and spatial variation of price. The American Economic Review, 78 (3), 418-430. Deaton, A. (1997). The analysis of household surveys: A micro-econometric approach to development policy. Baltimore: Johns Hopkins University Press. Diaz, L., Plat, D., & Pochet, P. (2008): Household transport expenditure in Sub-Saharan African cities: measurement and analysis. Journal of Transport Geography, 16 (1), 1-13. D'Haese, M., & Van Huylenbroeck, G. (2005). The rise of supermarkets and changing expenditure patterns of poor rural households case study in the Transkei area, South Africa. Food Policy, 30 (1), 97-113. 25 De Mel, S., McKenzie, D., & Woodruff, C. (2008). Returns to capital in microenterprises: Evidence from a field experiment. Quarterly Journal of Economics, 123 (4), 1329-1372. Filmer, D., & Pritchett, L. (2001). Estimating wealth effects without expenditure data-or tears: An application to educational enrollments in States of India. Demography, 38 (1), 115-132. Fontagne, L., Freudenberg, M., & Gaulier, G. (2006). A systematic decomposition of world trade into horizonal and vertical IIT. Review of World Economics, 142 (3), 459-475. Fortin, B., Laxroix, G., & Montmarquette, C. (2000). Are underground workers more likely to be underground consumers? The Economic Journal, 110 (466), 838-860. Gërxhani, K. (2004). The informal sector in developed and less developed countries: A literature survey. Public Choice, 120 (3/4), 267-300. Gibson, J. (2002). Why does the Engel method work? Food Demand, Economies of Size and Household Survey Methods. Oxford Bulletin of Economics and Statistics, 64 (4), 341-359. Griliches, Z. (1974). Errors in variables and other unobservables. Econometrica, 42 (6), 971-998. Gibson, J., & Bonggeun, K. (2007). Measurement error in recall surveys and the relationship between household Size and food demand. American Journal of Agricultural Economics, 89 (2), 473-489. Grimm, M., & Günther, I. (2006). Inter- and intra-household linkages between the informal and formal sector. A case study for urban Burkina Faso. In B. Guha-Khasnobis & R. Kanbur (Eds.), Informal labor markets and development (pp. 155-179). London: Palgrave Macmillan. Grimm, M., Krüger, J., & Lay, J. (2011). Barriers of entry and capital returns in informal activities: Evidence from Sub-Saharan Africa. Mimeo. Hausman, J. (2001). Mismeasured variables in econometric analysis: Problems from the right and problems from the left. The Journal of Economic Perspectives, 15 (4), 57-67. 26 Humphrey, D., & Qxley, H.(1976). Expenditure and household-size elasticities in Malawi: Urban- rural comparisons, demand projections and a summary of East African findings. Journal of Development Studies, 12 (2), 252-269. Hussmanns, R. (2004). Statistical definition of informal employment: Guidelines endorsed by 17th ICLS (2003) in Delhi Group on informal sector, February. ILO, Geneva. Kedir, A., & Girma, S. (2007). Quadratic Engel Curves with Measurement Error: Evidence from a Budget Survey. Oxford Bulletin of Economics and Statistics, 69 (1), 123-138. Keen, M. (1986). Zero expenditures and the estimation of Engel curves. Journal of Applied Econometrics, 1 (3), 277-286. Lachaud, J. (1990). The urban informal sector and the labour market in Sub-Saharan Africa. In D. Turnham, B. Salomé & A. Schwarz (Eds). The informal sector revisited (pp. 111-130). Paris: OECD Publication. La Porta, R., & Shleifer, A. (2011). The unofficial economy in Africa. NBER Working Paper 16821. Lewbel, A. (1996). Demand estimation with expenditure measurement errors on the left and right hand side. The Review of Economics and Statistics, 78 (4), 718-725. Liviatan, N. (1961). Errors in variables and Engel curve analysis. Econometrica, 29 (3), 336-362. Livingstone, I. (1991). A reassessment of Kenya's rural and urban informal sector. World Development, 19 (6), 651-670. Massell, B., & Heyer, J. (1969). Household expenditure in Nairobi: A statistical analysis of consumer behavior. Economic Development and Cultural Change, 17 (2), 212-234. McKenzie, D. (2005). Measuring inequality with asset indicators. Journal of Population Economics, 18 (2), 229-260. 27 Merella, V. (2006). Engel’s curve and product differentiation: A dynamic analysis of the effects of quality on consumer’s choice. International Review of Economics, 53 (2), 157-182. Min, I., Fang C., & Li, Q. (2000). Investigation of patterns in food-away-from-home expenditure for China. China Economic Review, 15, 457-476. Okunade, A. (1985). Engel curves in developing nations: The case of Africa. Eastern Africa Economic Review, 1 (1), 13-22. Ostby, I., & Gulilat, T. (1969). A statistical study of household expenditure in Addis Ababa. Eastern Africa Economic Review, 1 (2), 63-74. Reilly, B., Krstic, G., & Cominetta, M. (2006). Purchasing form informal markets: Evidence from Serbia. Applied Economics Quarterly, 52 (1), 239-264. Sethuraman, S. (1997). Urban poverty and the informal sector: A critical assessment of current strategies. Geneva: International Labour Organization. Staiger, D., & Stock, J. H. (1997). Instrumental variables regression with weak instruments. Econometrica, 65 (3), 557-586. Stock, J. H., & Yogo, M. (2003). Testing for weak instruments in Linear IV Regression. NBER Technical Working Paper, 284. Teklu, T. (1996). Food demand studies in Sub-Saharan Africa: A survey of empirical evidence. Food Policy, 21 (6), 479-496. Theil, H. (1979). The effect of measurement errors in the estimation of demand systems. Economic Letters, 3 (4), 373-376. Tiwari, R. S., & Goel, M. S. (2002). Migration pattern, poverty profile and consumption pattern : a study of unregistered informal sector workers in cities of Agra and Kanpur in Uttar Pradesh and Puri in Orissa. The Indian Journal of Economics, 82 (3), 311-335. 28 Wan, G. (1996). Income elasticities of household demand in rural China: Estimates from cross- sectional survey data. Journal of Economic Studies, 23 (3), 18-33. Xaba, J., Horn, P., & Motala, S. (2002). The informal sector in sub-Saharan Africa. ILO Working Paper on the Informal Economy, Employment Sector, ILO, Geneva, 848. 29 1 For a survey of different definitions of the informal sector, see Gërxhani (2004). 2 Political tensions in Côte d’Ivoire prevented the execution of the third phase in Abidjan. Hence, Abidjan is not included in the present analysis. 3 The large number of adults may partly reflect polygamy. We thank an anonymous referee for pointing us to this possible explanation. 4 These categories are: (1) Food and Non-Alcoholic Beverages, (2) Alcoholic Beverages and Tobacco, (3) Clothing and Shoes, (4) Housing, (5) Gas, Electricity and Water, (6) Furniture and Household Maintenance, (7) Health and Education, (8) Transport and Communication, (9) Leisure and Culture, (10) Hotels and Restaurants and (11) Diverse Goods and Services. 5 For Togo, we find no evidence pointing to an Engel curve. 6 Housing expenditures include imputed rents for owner-occupied housing. 7 The results of Table 4 remain comparable or become even more pronounced when we focus on the frequency of purchases. 8 See Blundell and Duncan (1998) for a detailed discussion of household expenditure and non-parametric kernel regressions. 9 For a general discussion of whether income or expenditure constitutes the preferred welfare indicator in the context of developing countries, see for example Deaton (1997). 10 We thank an anonymous referee for pointing us to this option. 11 We get virtually the same result when pooling the data from the six capitals (see last column of Table 7). 12 Robustness checks using the other two instrumentations (not shown; available upon request) left results qualitatively unaffected. 13 See Diaz et al. (2008) for a survey of transport expenditures in Sub-Saharan Africa. 14 Ideally, one would want to disaggregate even further so as to arrive at fairly homogenous items (e.g. single goods such as maize or millet) where quantity and quality aspects can be disentangled. This would, however, render the distinction between formal and informal distribution channels meaningless as one of them prevails. 15 To save space, we do not report the regression results here. Estimations are available from the authors’ upon request. 30 Table 1: Summary Statistics of Sample Households by Ccountry Burkina Benin - Mali - Niger - Senegal - Togo - Country Faso - Cotonou Bamako Niamey Dakar Lomé Ouaga Category coastal sahel sahel sahel coastal coastal Mean number of household members Infants (<6) 0.4 0.6 0.8 0.8 0.4 0.4 Children (6-15) 0.7 1.3 1.2 1.5 1.2 0.7 Adults (>15) 3.0 4.0 3.6 3.9 6.1 3.0 Mean age of household head 42.5 43.2 43.8 44.5 50.5 39.6 Sex of household head (male=1) 72.9 87.0 88.6 86.3 73.0 72.1 Completed primary education of household 83.9 54.2 53.1 48.0 58.7 83.3 head (%) Primary wage income source is public 12.6 19.0 13.1 18.9 13.4 10.9 sector (%) Primary wage income source is private 75.4 71.7 77.6 68.0 75.1 79.9 sector (%) Household head earns primary income in 54.1 46.8 50.1 47.1 38.2 61.3 the informal sector (%) Observations 573 936 956 575 567 569 Source: Authors’ calculation based on 1-2-3 Surveys 31 Table 2: National Annual Household Expenditure Shares by Sector (%) Burkina Benin - Mali - Niger - Senegal - Togo - Country Faso - Cotonou Bamako Niamey Dakar Lomé Ouaga Food and Non-Alcoholic Beverages 29.0 29.6 35.0 36.2 36.4 33.4 Alcoholic Beverages and Tobacco 1.7 2.9 0.5 1.1 0.8 1.8 Clothing and Shoes 6.1 6.6 7.2 7.1 5.2 8.5 Housing 14.4 11.2 15.9 13.2 17.8 9.8 Gas, Electricity and Water 6.9 6.7 9.6 7.6 8.3 5.4 Furniture and Household Maintenance 3.8 4.6 5.3 5.7 6.0 4.4 Health and Education 8.5 9.4 5.3 5.7 6.4 7.9 Transport and Communication 14.7 16.3 13.5 12.4 9.8 10.8 Leisure and Culture 3.1 3.8 1.7 2.7 2.9 2.5 Hotels and Restaurants 7.0 4.9 1.7 4.6 2.2 9.3 Diverse Goods and Services 4.8 4.0 4.3 3.6 4.1 6.2 Source: Authors’ calculation based on 1-2-3 Surveys 32 Table 3: Average Annual Household Expenditure Shares by Sector and Quintile (%) Country Benin - Cotonou Burkina Faso - Ouaga Mali - Bamako Quintile (Total Expenditure) 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 Food and Non-Alcoholic Beverages 34.7 34.8 33.6 31.9 26.6 37.6 40.7 37.2 34.5 24.1 39.7 43.3 41.5 38.2 29.7 Alcoholic Beverages and Tobacco 4.5 3.2 3.3 3.7 4.4 7.6 4.6 5.2 5.5 4.7 4.6 3.8 2.2 1.4 1.6 Clothing and Shoes 6.9 7.2 7.3 7.1 6.5 6.9 6.6 7.0 7.0 7.7 7.6 6.9 8.5 8.2 8.5 Housing 19.5 13.8 12.0 13.0 13.7 15.9 12.6 11.7 11.5 11.1 27.1 20.3 18.0 17.6 12.7 Gas, Electricity and Water 6.7 7.0 7.3 7.4 6.2 6.5 6.7 7.5 7.1 6.8 6.0 6.3 7.7 9.4 12.0 Furniture and Household Maintenance 3.4 4.0 3.5 3.6 4.2 4.1 3.6 3.7 4.2 5.7 4.2 4.7 4.5 5.5 7.3 Health and Education 4.4 5.9 8.4 8.5 10.6 5.8 7.0 8.6 9.6 10.7 4.6 4.2 4.6 4.7 6.9 Transport and Communication 9.2 10.5 10.3 11.9 17.1 6.7 8.4 9.7 11.7 19.9 8.3 10.0 10.2 10.9 16.5 Leisure and Culture 3.3 2.8 3.1 3.8 4.0 3.1 2.8 3.8 3.9 4.4 1.4 1.6 1.7 1.5 2.3 Hotels and Restaurants 12.2 9.3 9.4 7.5 5.7 14.6 8.2 6.7 5.9 3.6 11.8 6.1 3.6 4.5 2.7 Divers Goods and Services 4.9 5.5 5.5 4.6 4.4 4.3 3.8 3.8 4.2 4.4 4.4 4.3 4.4 4.6 5.0 33 (Table 3 continued) Country Niger - Niamey Senegal - Dakar Togo - Lomé Quintile (Total Expenditure) 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 Food and Non-Alcoholic Beverages 41.3 48.9 44.9 40.6 29.9 40.1 43.3 42.3 40.1 30.0 28.7 33.2 34.7 33.1 29.9 Alcoholic Beverages and Tobacco 7.3 3.7 4.2 2.2 3.4 4.4 1.2 2.9 1.7 1.4 6.9 6.6 4.0 4.8 4.0 Clothing and Shoes 5.9 6.8 8.1 8.4 7.9 5.5 5.6 5.2 5.1 6.0 7.6 7.6 9.1 8.7 9.5 Housing 19.2 14.4 13.8 13.6 12.6 18.1 19.0 20.3 17.8 16.9 17.4 12.6 10.8 9.6 8.3 Gas, Electricity and Water 8.5 7.5 6.1 7.6 8.4 10.2 9.4 9.6 8.6 7.5 5.9 5.7 4.8 5.9 5.6 Furniture and Household Maintenance 3.9 4.6 6.1 5.6 7.2 5.3 4.4 5.4 5.9 7.2 4.5 4.3 3.8 4.2 5.0 Health and Education 3.5 3.4 4.9 5.4 7.1 3.6 4.6 4.7 5.7 8.8 5.3 5.9 7.3 7.0 10.2 Transport and Communication 6.1 5.8 5.8 7.8 16.8 5.3 4.7 4.7 7.6 13.8 9.4 7.9 8.7 10.0 13.4 Leisure and Culture 2.1 1.8 2.5 2.9 3.6 2.3 2.9 2.4 2.6 3.9 2.9 2.9 2.8 2.7 3.3 Hotels and Restaurants 13.6 7.0 7.3 6.8 4.0 10.8 6.1 2.7 2.4 2.2 13.2 13.1 11.0 10.2 7.7 Divers Goods and Services 3.8 3.4 4.4 4.0 4.2 4.0 4.2 3.8 4.8 4.3 7.5 6.2 6.9 7.2 5.8 Source: Authors’ calculation based on 1-2-3 Surveys 34 Table 4: National Annual Household Expenditure Shares by Sector and Distribution Channel (%) Burkina Faso - Country Benin - Cotonou Mali - Bamako Niger - Niamey Senegal - Dakar Togo - Lomé Ouaga Distribution Channel formal informal formal informal formal informal Formal informal formal informal formal informal Food and Non-Alcoholic Beverages 2.0 27.0 1.1 28.5 0.4 34.5 0.5 35.8 1.0 35.4 0.3 33.0 Alcoholic Beverages and Tobacco 0.2 1.5 0.1 2.7 0.0 0.5 0.0 1.1 0.0 0.7 0.1 1.8 Clothing and Shoes 1.4 4.7 0.8 5.8 1.0 6.2 1.1 6.1 0.9 4.3 0.2 8.3 Housing 0.5 13.9 0.7 10.5 0.9 15.0 0.4 12.8 1.7 16.0 0.1 9.7 Gas, Electricity and Water 4.0 2.9 4.3 2.4 6.0 3.6 4.3 3.3 5.0 3.3 3.0 2.4 Furniture and Household Maintenance 0.8 3.1 0.5 4.1 0.4 4.9 0.6 5.1 0.7 5.3 0.2 4.3 Health and Education 7.6 0.9 7.7 1.7 3.7 1.6 4.5 1.3 5.3 1.1 6.3 1.5 Transport and Communication 6.4 8.3 8.3 8.0 5.4 8.0 6.7 5.7 6.5 3.4 3.0 7.8 Leisure and Culture 1.4 1.7 1.3 2.5 0.4 1.3 0.8 1.9 1.3 1.6 1.0 1.5 Hotels and Restaurants 0.5 6.5 0.4 4.5 0.2 1.5 0.6 4.0 0.4 1.9 0.4 8.9 Divers Goods and Services 1.0 3.7 0.6 3.5 0.7 3.5 0.6 3.0 0.8 3.3 0.4 5.8 Source: Authors’ calculation based on 1-2-3 Surveys 35 Table 5: National Annual Household Expenditure Shares by Product Category (%) Burkina Benin - Mali - Niger – Senegal - Togo - Country Faso - Cotonou Bamako Niamey Dakar Lomé Ouaga Budget share formal goods Formal distribution channel 4.3 5.5 7.2 4.5 8.3 0.4 Informal distribution channel 9.5 11.9 15.1 9.3 22.2 7.2 Budget share informal goods Formal distribution channel 0.7 0.5 0.3 0.1 0.5 0.1 Informal distribution channel 17.1 17.3 25.5 15.0 17.7 21.7 Budget share imported goods Formal distribution channel 10.0 10.1 5.2 6.7 4.1 8.8 Informal distribution channel 18.3 21.9 16.5 31.6 11.3 24.5 Budget share services Formal distribution channel 10.7 9.5 6.4 8.6 10.7 6.1 Informal distribution channel 29.4 23.2 23.7 24.1 25.1 31.3 Note: For each country, budget shares sum up to 100 percent. Source: Authors’ calculation based on 1-2-3 Surveys 36 Table 6: National Shares of Informal Households in Overall Expenditures (%) Burkina Benin - Mali – Niger - Senegal - Togo - Country Faso - Cotonou Bamako Niamey Dakar Lomé Ouaga Goods Formal goods Formal distribution channel 36.8 24.1 31.0 48.1 26.2 35.5 Informal distribution channel 50.0 39.7 43.7 50.5 34.6 56.3 Informal goods Formal distribution channel 27.0 37.4 24.3 56.2 33.8 55.4 Informal distribution channel 49.6 42.1 46.1 49.1 35.2 57.5 Imported goods Formal distribution channel 34.6 24.3 31.5 35.2 24.5 48.0 Informal distribution channel 48.5 38.2 41.8 51.8 34.5 56.7 Services Formal distribution channel 36.0 24.8 34.3 43.2 20.8 48.6 Informal distribution channel 35.5 26.9 28.3 37.1 18.4 48.4 Note: Informal Households are defined as those for whom the informal sector is the primary income source. From each cell, expenditure shares of formal households can be calculated as 100 percent minus the expenditure share of informal households. The expenditure share of formal households on formal goods distributed through formal distribution channels in Benin, for example, is 100 - 36.8 = 63.2. Source: Authors’ calculation based on 1-2-3 Surveys 37 Table 7: Budget Elasticities for Sectors (OLS) Burkina Benin - Mali - Niger - Senegal Togo - Country Faso - Pooleda Pooled Cotonou Bamako Niamey - Dakar Lomé Obs Ouaga Elasticities by Sector Food and Non-Alcoholic Beverages 0.80*** 0.76*** 0.77*** 0.74*** 0.80*** 0.95*** 0.80*** 4160 Alcoholic Beverages and Tobacco 0.87*** 0.84*** 0.51*** 0.28 0.16 0.74*** 0.57*** 1601 Clothing and Shoes 1.07*** 1.14*** 1.19*** 1.17*** 1.24*** 1.27*** 1.14*** 3696 Housing 0.68*** 0.64*** 0.48*** 0.59*** 0.81*** 0.42*** 0.70*** 4159 Gas, Electricity and Water 0.87*** 0.86*** 1.18*** 0.90*** 0.79*** 0.86*** 0.97*** 4138 Furniture and Household Maintenance 1.23*** 1.19*** 1.24*** 1.27*** 1.36*** 1.15*** 1.21*** 3868 Health and Education 1.24*** 1.32*** 1.26*** 1.28*** 1.48*** 1.33*** 1.25*** 3948 Transport and Communication 1.36*** 1.75*** 1.68*** 1.71*** 1.85*** 1.38*** 1.57*** 3962 Leisure and Culture 1.11*** 1.18*** 1.27*** 1.32*** 1.39*** 1.22*** 1.15*** 3416 Hotels and Restaurants 0.55*** 0.39*** 0.42*** 0.50*** 0.41*** 0.58*** 0.42*** 3198 Diverse Goods and Services 0.84*** 1.07*** 1.12*** 1.01*** 1.16*** 0.86*** 1.01*** 3888 Elasticities by Sector (2SLS) Food and Non-Alcoholic Beverages 0.59*** 0.52*** 0.49*** 0.34*** 0.52*** 0.49*** 0.51*** 4160 Gas, Electricity and Water 1.04*** 1.29*** 1.55*** 1.37*** 1.04*** 1.56*** 1.30*** 4138 Dependent Variable is log of total household expenditure on a specific product group; Independent Variable is log of total household expenditure; Included Control Variables are district dummies, log number of household members (OECD-modified Adult Equivalent Scale), share of adult women in the household, elderly in the household, type of family structure, gender of household head, age of household head, completed primary education of household head, religion of household head (Muslim or Catholic Christian) and informal sector is source of household head's primary income; *** p<0.01, ** p<0.05, * p<0.1 based on robust standard errors; a includes country dummies. Source: Authors’ calculation based on 1-2-3 Surveys 38 Table 8: Budget Elasticities for Sectors, Quadratic Specification (OLS) Burkina Faso - Country Benin - Cotonou Mali - Bamako Niger - Niamey Senegal - Dakar Togo - Lomé Ouaga Elasticities by Sector EXP EXP2 EXP EXP2 EXP EXP2 EXP EXP2 EXP EXP2 EXP EXP2 Food and Non-Alcoholic Beverages 4.47*** -0.13*** 7.11*** -0.23*** 8.03*** -0.25*** 7.67*** -0.25*** 6.06*** -0.18** 5.098*** -0.19*** Marginal effects 0.84 0.81 0.80 0.81 0.79 0.98 Standard errors 0.04 0.02 0.03 0.05 0.04 0.05 Observations 568 933 953 571 567 568 R² 0.69 0.68 0.66 0.54 0.65 0.60 Transport and Communication -2.89* 0.15** 0.47 0.05 -4.88** 0.23*** -9.57*** 0.40*** -6.44*** 0.29*** -3.37** 0.18*** Marginal effects 1.31 1.74 1.65 1.62 1.86 1.35 Standard errors 0.07 0.05 0.08 0.08 0.09 0.07 Observations 560 880 868 531 563 560 R² 0.66 0.64 0.46 0.56 0.57 0.50 Dependent Variable is log of total household expenditure on a specific product group; Independent Variable is log and squared log of total household expenditure; Included Control Variables are district dummies, log number of household members (OECD-modified Adult Equivalent Scale), share of adult women in the household, elderly in the household, type of family structure, gender of household head, age of household head, completed primary education of household head, religion of household head (Muslim or Catholic Christian) and informal sector is source of household head's primary income; *** p<0.01, ** p<0.05, * p<0.1 based on robust standard errors; Source: Authors’ calculation based on 1-2-3 Surveys 39 Table 9: Budget Elasticities for Expenditure Categories (2SLS) Burkina Benin - Mali - Niger - Senegal - Togo - Pooled Country Faso - Pooleda Cotonou Bamako Niamey Dakar Lomé Obs Ouaga Elasticities by Expenditure Category and Distribution Channel Elasticities of formal goods 0.85*** 1.38*** 1.33*** 1.22*** 0.95*** 0.58*** 1.22*** 4165 Formal distribution channel 1.68*** 2.26*** 2.09*** 1.87*** 2.11*** 1.79*** 2.02*** 2719 Informal distribution channel 0.18** 0.63*** 0.48*** 0.29*** 0.38*** 0.45*** 0.57*** 4156 0.63*** 4131 Elasticities of informal goods 0.66*** 0.56*** 0.45*** 0.62*** 0.92*** 0.33*** Formal distribution channel 1.28*** 0.84*** 1.19* -0.12 1.41*** 1.78*** 1.02*** 634 Informal distribution channel 0.40*** 0.42*** 0.33*** 0.61*** 0.63*** 0.28** 0.46*** 4131 1.12*** 4153 Elasticities of imported goods 1.36*** 1.13*** 1.49*** 0.98*** 0.91*** 1.55*** Formal distribution channel 1.95*** 1.70*** 2.04*** 2.31*** 1.65*** 3.02*** 1.96*** 3031 Informal distribution channel 0.73*** 0.66*** 1.06*** 0.42*** 0.27* 0.91*** 0.59*** 4148 Elasticities of services 1.26*** 1.42*** 1.32*** 1.54*** 1.59*** 1.25*** 1.40*** 4174 Formal distribution channel 2.00*** 1.89*** 1.81*** 2.25*** 2.11*** 2.17*** 1.98*** 3066 Informal distribution channel 0.98*** 1.00*** 1.09*** 1.20*** 1.25*** 0.86*** 1.06*** 4170 Dependent Variable is log of total household expenditure on a specific product group; Independent Variable is log of total household expenditure which is instrumented by a wealth index; Included Control Variables are district dummies, log number of household members (OECD-modified Adult Equivalent Scale), share of adult women in the household, elderly in the household, type of family structure, gender of household head, age of household head, completed primary education of household head, religion of household head (Muslim or Catholic Christian) and informal sector is source of household head's primary income; *** p<0.01, ** p<0.05, * p<0.1 based on robust standard errors; a includes country dummies. Source: Authors’ calculation based on 1-2-3 Surveys 40 Figure 1: Elasticities by Product Category and Distribution Channel in Benin 41 Appendix: Demand System Estimations (SURE) Burkina Benin - Mali - Niger - Senegal - Togo - Country Faso - Cotonou Bamako Niamey Dakar Lomé Ouaga Elasticities by Sector Food and Non-Alcoholic Beverages 1.05*** 0.88*** 0.87*** 0.9*** 0.73*** 0.97*** Alcoholic Beverages and Tobacco 1.64*** 1.18*** 0.40 0.01 1.10*** 1.73*** Clothing and Shoes 1.56*** 1.88*** 2.62*** 2.25*** 2.44*** 1.86*** Housing 0.72*** 0.66*** 0.91*** 0.61*** 0.85*** 0.48*** Gas, Electricity and Water 1.09*** 0.96*** 1.42*** 1.04*** 0.78*** 0.88*** Furniture and Household Maintenance 1.50*** 1.43*** 2.01*** 2.16*** 1.52*** 1.28*** Health and Education 1.48*** 1.80*** 2.26*** 1.86*** 2.11*** 1.53*** Transport and Communication 2.01*** 2.81*** 2.64*** 2.68*** 2.12*** 1.74*** Leisure and Culture 2.30*** 2.32*** 2.7*** 2.45*** 3.03*** 2.04*** Hotels and Restaurants 0.53*** 0.55*** 0.59** 1.00*** 1.44*** 0.63*** Elasticities by Product Category and Distribution Channel Elasticities of formal goods Formal distribution channel 1.17*** 1.24*** 1.81*** 1.20*** 1.61*** 1.22*** Informal distribution channel 0.34** 1.17*** 0.22 0.09 0.29** 0.96*** Elasticities of informal goods Formal distribution channel 0.81*** 0.67*** 0.17 0.19 1.66*** - Informal distribution channel 0.51*** 0.59*** 0.58*** 0.28*** 0.42** 0.64** Elasticities of imported goods Formal distribution channel 1.93*** 1.72*** 2.36*** 2.05*** 1.99*** 1.85*** Informal distribution channel 0.61*** 0.90*** 0.11 0.56*** 0.08 1.01*** Elasticities of services Formal distribution channel 2.03*** 1.75*** 2.55*** 2.26*** 2.88*** 1.89*** Informal distribution channel 0.71*** 0.71*** 0.55** 0.91*** 0.62*** 0.67*** Source: Authors’ calculation based on 1-2-3 Surveys 42