Urban Demand for Food, Beverages, Betelnut and Tobacco. in Papua New Guinea. John Gibson 1

Transcription

1 Urban Demand for Food, Beverages, Betelnut and Tobacco n Papua New Gunea John Gbson 1 ABSTRACT Data from household surveys conducted between 1985 and 1990 were used n a regresson model to estmate the mpact that changes n household ncomes have on the demand for each of 36 major foods, beverages, betelnut and tobacco wthn urban areas of Papua New Gunea. Locally produced crops that should be n hgh demand n future, f urban ncomes rse, nclude betelnut, fresh vegetables, sweet potato, and fresh fruts. All else the same, the hgher value of future demand for these crops may justfy extra research spendng, as opposed to spendng on crops whose demand wll not respond strongly to rsng urban ncomes. KEY WORDS Demand, ncome elastcty, Papua New Gunea, research prortes, urban consumers. 1 Department of Economcs, Unversty of Wakato, Prvate Bag 3105, New Zealand.

2 INTRODUCTION In recent years Papua New Gunea (PNG) agrculturalsts have begun to pay more attenton to market opportuntes for supplyng food to the local market. Ths swtch n attenton has been caused both by low world prces for the tradtonal export crops and local market growth, especally wthn urban areas. Although precse data on the value of local food producton for the urban market are unavalable, nformed estmates suggest that ths ndustry rvals the man tree crop ndustres as an ncome source for rural producers (Shaw, 1985). For agrcultural planners to allocate scarce research and extenson funds amongst the many foods that can be grown for the urban market, nformaton about the demand prospects for those foods s requred. In partcular, t would be helpful to dentfy the crops that wll be n hgh local demand n the future, so that farmers can be encouraged to produce these crops. Examnng the demand patterns of urban households can help agrcultural planners dentfy market opportuntes by showng () whch foods have large exstng markets, and () whch foods wll be most heavly demanded n future as urban ncomes change. There are several examples of demand analyss beng used to nform agrcultural research prores n the nternatonal lterature. For example, Sarma and Gandh (1990) used demand analyss to forecast future consumpton patterns of food grans n Inda, under alternatve scenaros of economc growth and development. Pnstrup-Andersen, de Londono, and Hoover (1976) used data on demand patterns to see how reallocatng the agrcultural research budget would affect the nutrtonal status of the urban poor n Colomba. However, these types of analyses have not prevously been possble n PNG because of the lack of nformaton on consumer demand patterns for foods and agrcultural crops. 1

3 Varous concepts can be used to measure exstng market demand. Two of these are: 1. The average share of the household budget spent on an tem. 2. The proporton of households who consume the tem. Items wth a large average share n household budgets wll generate more revenue for producers because of the large expendtures that households make. However the average budget share can be nflated when a mnorty of the populaton have an ntense demand for an tem, whle the remanng households do not consume t at all. In ths case, demand may be vulnerable to what happens to ths small group of consumers. Foods consumed heavly by expatrate households would be a relevant example for PNG. Therefore, t s valuable to know what proporton of households consume a partcular food, so that the stablty of ts demand can be assessed. Two economc concepts for predctng future market demand under changng ncomes are: 1. The ncome elastcty of demand. 2. The margnal budget share. The ncome elastcty of demand measures the percentage by whch the quantty demanded of an tem ncreases followng a one percent ncrease n household ncome. When the demand ncreases by more than one percent, an tem s known as a "luxury good", when the demand ncreases by between zero and one percent, t s a "normal good", and when the demand goes down as ncomes rse, t s an "nferor good" (Sadoulet and de Janvry, 1995). The margnal budget share measures how a household allocates any addtons to ts budget, n contrast to the exstng dvson of the budget. For example, f a household's ncome rose by K100 per year, the margnal budget share for a partcular food shows how many Kna or toea from that K100 would be spent on the food. Hence, the margnal budget share s a good measure of the value of future demand. Ths paper provdes estmates for PNG urban areas of these four dfferent measures of market 2

4 demand for 36 major foods, beverages, betelnut and tobacco. The mplcatons of these demand estmates for agrcultural research prortes are then dscussed. METHODS AND DATA SOURCES There are two mportant requrements for the functonal forms that are used to estmate ncome elastctes of food demand. They should be flexble, by allowng ncome elastctes to dffer between rch and poor households, because the usual pattern s for ncome elastctes of food demand to fall as ncome rses (Tmmer, Falcon and Pearson, 1983). The functonal form also should be able to be estmated even when a household has zero consumpton of partcular foods; otherwse those households have to be dropped from the sample, whch can cause sample selecton bas (Deaton, 1989). The "share-log" functonal form, where the budget share of the th food (w ) s regressed on the logarthm of household ncome (x) meets these requrements. In addton to ncome, other varables affectng the budget share of partcular foods need to be ncluded n the model. These varables nclude the sze of the household, ts age, gender and ethnc composton, and dummy varables for the town the household s located n. Ths gves the followng regresson model: w = α + β ln x + Θj z j + u, (1) where z j s the vector of control varables, α, β and Θj are parameters to be estmated, and u s a random error term. The coeffcent β gves the rate at whch the budget share for the th food changes as the logarthm of ncome changes, w ln x, and ths can be transformed nto the elastcty of the budget share wth respect to ncome, ln w ln x, by dvdng by w (because ln w = w w ). The fact that the budget share s the product of prce, p, and quantty, q, dvded by ncome, x (and hence ln w = ln p + ln q ln x ) allows the ncome elastcty of the 3

5 quantty demanded of the th food, ε to be derved from the formula: ε = β w + 1. (2) Because budget shares vary by household, the ncome elastcty calculated wth equaton (2) also vares. For example, the estmated rce demand of rch households can be less ncome elastc than the rce demand of poor households, because rce has a bgger budget share for poor households. Ths s consstent wth the emprcal pattern usually found (Tmmer, Falcon and Pearson, 1983). The margnal budget share was estmated by multplyng the ncome elastcty of demand for the th food by the average share that food has n household budgets. These margnal budget shares must obey the "addng-up" condton that they sum to one. In other words, the value of all extra demands, followng a rse n ncome, must exactly equal the value of the extra ncome (Deaton and Muellbauer, 1980). Ths condton provdes a cross-check on the plausblty of the estmates. Data Household ncome and expendture data were obtaned from urban household surveys conducted by the Natonal Statstcal Offce between 1985 and These surveys were carred out n nne PNG urban areas: Port Moresby, Goroka, Wewak, Ambunt, Keta, Arawa, Panguna, Lae, and Rabaul. In all, over 1400 households were surveyed, and complete data were avalable for 1095 households (Gbson, 1995). Ths s the largest household expendture survey ever carred out n PNG. The survey collected nformaton on purchases, gfts, producton, and sales of food wthn a two week perod, as well as changes n household food stocks. From these components, the value 2 I am grateful to the Natonal Statstcal Offce for allowng access to these data. 4

6 of each household's consumpton of 36 food tems was estmated. Although the survey collected nformaton on household wage and busness ncome, the actual ncome varable used n ths study s household total expendture. Total expendture s less affected by transtory components (e.g. extra overtme pay) and s more closely related to the concept of permanent ncome, whch s what economsts beleve nfluences demand (Deaton and Muellbauer, 1980). The total expendture varable ncludes the value of consumpton of all goods and servces, ncludng mputed rent. Households were surveyed at dfferent ponts n tme so the effects of nflaton were accounted for by usng a prce ndex calculated for each town. Full detals of the constructon of varables and the qualty control procedures that the data were subjected to are contaned n Gbson (1995). RESULTS AND DISCUSSION The four dfferent measures of market demand are reported n Table 1 for 36 foods, beverages, betelnut and tobacco. Measures for the aggregate non-food group (contanng all other goods and servces) are provded for comparson. The tems wth the largest average share of household budgets are rce, alcohol, tnned meat, tobacco products, tnned fsh, poultry, and bread and bscuts. These tems make up over one quarter of the total budget for the average household. The tems that are consumed most commonly by households are rce, bread and bscuts, tnned fsh, soft drnks, and fresh vegetables. At least 90 percent of households consumed these wthn the two week perod. Hence t s lkely that these tems have very stable demand. In contrast, alcohol has a hgh average share of the budget but only 40 percent of the households consumed alcohol. The food tem consumed by the smallest proporton of households was pork, wth less than 10 percent of households consumng wthn the two week perod. Pork s sometmes regarded as the natonal meat so ths low consumpton seems surprsng but t may reflect the hgh prce of pork n urban areas and the fact that pork s consumed durng celebratons and ceremonal events rather 5

7 than as an everyday meal. The ncome elastctes of demand, and assocated standard errors, are reported n the thrd and fourth column of Table 1. These were estmated usng Ordnary Least Squares, wth heteroscedastcally consstent standard errors beng calculated (Whte, 1980). There were only three foods whose ncome elastctes were not statstcally sgnfcant (cassava, dred and other fsh, and other roots and tubers). The luxury goods, wth ncome elastctes greater than one, were alcohol, eggs, Englsh potatoes, poultry, beef and veal, and fresh frut (except bananas). For example, f the ncome of an urban household rose by one percent, ts demand for fresh frut should rse by 1.04 percent (±0.096). No nferor goods were dentfed but some mportant foods had ncome elastctes that were farly low, e.g., tnned fsh (0.15±0.055); rce (0.30±0.048); and sugar (0.33±0.076). Rsng urban ncomes wll not lead to bg percentage ncreases n the demand for these three basc foods. In contrast, most of the foods produced by rural households and sold n urban areas had ncome elastctes n the 0.50 to 0.75 range, thus sales of these tems should beneft strongly from rsng urban ncomes. The fnal column of Table 1 gves the margnal budget share for each tem. Ths s the amount of extra spendng on an tem f the household had an extra K100 ncome avalable. The "addng-up" condton that the sum of extra expendtures exactly equals K100 holds, and confrms the plausblty of the results. The margnal budget share gves the best overall estmate of the future value of a crop because t combnes the ncome elastcty wth a measure of the tem's monetary mportance. For example, 6

8 Englsh potatoes are a luxury food (ε=1.26±0.148), whch farmers should grow because ther demand wll rse wth ncreases n urban ncomes. However, Englsh potato s such a mnor budget tem, a household wth an extra K100 to spend would allocate only eght toea to extra purchases of ths tem. Even though sweet potato has a lower ncome elastcty, the same household would allocate K1.11 to extra sweet potato purchases. Hence, the research payoff would be greater, ceters parbus, from sweet potato than from Englsh potato. The tems n Table 1 are ranked by the value of ther margnal budget share (the fnal column). The tems that would gan the bggest share of ncreased spendng by urban consumers nclude alcohol, poultry, takeaway meals, tnned meat, soft drnks, and bread and bscuts. Amongst the tems produced by rural households and sold n urban markets, betelnut, fresh vegetables, and sweet potato would see the bggest ncrease n demand f urban household ncomes ncreased. The margnal budget share for alcohol s twce as large as for any other tem and ths may be a source of socal concern, because of the problems that excessve alcohol consumpton can create. Several of the control varables n the vector z j of equaton (1) were mportant nfluences on the demand for alcohol, especally the proporton of the household that was male and aged between 15 and 50 years, and whether the household head was from the Hghlands. All else the same, the budget share for alcohol would be two percentage ponts hgher n a household where men between the age of 15 and 50 comprsed one-half of the dwellers, compared wth a household where they were only one-ffth of the dwellers. Dscusson Ths orderng of food crops may be helpful n gudng research prortes, because t has been suggested that some crops have ganed more research attenton than others of greater economc 7

9 mportance (Shaw, 1985). The publcaton of these elastcty and budget share estmates provdes one economc crtera for rankng food crop research, although other factors, such as compatblty wth exstng farmng systems and envronmental sutablty of the crop play a role as well. One food crop that has receved consderable research attenton over many years n PNG s rce (Shaw 1985; Fereday, 1993). The results n Table 1 show that the margnal budget share for rce s lower than the margnal budget share for foods that are produced from wheat, whch s the other major cereal mport. Hence, agrcultural research that helps local farmers to ncrease ther share of the nputs used n the producton of bread and bscuts may gve a hgher payoff than research nto the local producton of rce. One example of research that mght help local farmers ncrease ther share of the nputs used n bread and bscut manufacture s the queston of blendng sweet potato flour wth wheat flour (Berros, 1992). The results also help to dsprove the hypothess that mported cereals are seen by local households as superor foods to the tradtonal starchy staples (Kannapran, 1993). If ths hypothess were true, the ncome elastcty of demand for cereals would be greater than for the local root crops, mplyng that consumers swtch to cereals as urban ncomes rse. Although bread and bscuts have a greater ncome elastcty of demand (0.75±0.044) than the major root crops, ths s not the case for rce (0.30±0.048), whch s the cereal wth the largest budget share. To further compare the ncome elastctes of cereals versus local starchy staples, foods were grouped nto () cereals, () meats and fsh, () fruts, vegetables and nuts, and (v) root crops and the shareweghted ncome elastctes of demand for each group were calculated (Table 2). The ncome elastcty of demand for the cereals group s 0.48, ndcatng a ten percent ncrease n urban household ncome would cause cereals demand to rse by 4.8 percent. The ncome 8

10 elastcty of demand for the root crops s slghtly hgher, at 0.49, so demand for the root crops benefts slghtly more from rsng urban ncomes than does demand for the cereals. Other locally produced fruts, vegetables, and nuts also have demands that rse by more than those of cereals when urban ncomes rse. Thus, rsng urban ncomes need not sgnal a detary transton away from locally produced foods. Instead, the demand profle for food tems assocated wth rsng urban ncomes ndcates an mportant market opportunty for rural producers to supply food nto the local urban market. CONCLUSIONS Income elastctes of demand and margnal budget shares were estmated for 36 food tems n urban Papua New Gunea markets. The rankng of tems accordng to ncome elastctes and margnal budget shares gves one economc crtera for the settng of research and extenson prortes. Locally produced crops that should be n hgh demand n future, f urban ncomes rse, nclude betelnut, fresh vegetables, sweet potato, and fresh fruts. All else the same, the hgher future demand for these crops may justfy ncreased research spendng, as opposed to research spendng on crops whose demand wll not respond strongly to rsng urban ncomes. The results also suggest that urban demand for wheat products wll eventually exceed demand for rce, f urban ncomes rse, so research nto the replacement of mported cereals wth local crops may gve a hgher payoff f t s drected at fndng substtutes for wheat flour. 9

11 REFERENCES BERRIOS, D. (1992). Sweet potato bread n Burund: An update. CIP Crcular. 19(1): 9. DEATON, A. (1989). Household survey data and prcng polces n developng countres. The World Bank Economc Revew. 3(2): DEATON, A. and MUELLBAUER, J. (1980). Economcs and Consumer Behavor. Cambrdge Unversty Press, New York. FEREDAY, N. (1993). Rce producton n Papua New Gunea: The experences of the Berena and Maprk rce projects. Polcy Workng Paper no. 4, Programmng, Plannng and Budgetng Dvson, Department of Agrculture and Lvestock, Konedobu, Papua New Gunea. GIBSON, J. (1995). Food Consumpton and Food Polcy n Papua New Gunea. Insttute of Natonal Affars, Port Moresby. KANNAPIRAN, C. (1993). Food securty: Rce producton polcy n Papua New Gunea. Polcy Dscusson Paper Department of Agrculture and Lvestock, Papua New Gunea. PINSTRUP-ANDERSEN, P., DE LONDONO, N. and HOOVER, E. (1976). The mpact of ncreasng food supply on human nutrton: Implcatons for commodty prortes n agrcultural research and polcy. Amercan Journal of Agrcultural Economcs. 58(1):

12 SADOULET, E. and DE JANVRY, A. (1995). Quanttatve Development Polcy Analyss. Johns Hopkns Unversty Press, Baltmore. SARMA, J. and GANDHI, V. (1990). Producton and Consumpton of Foodgrans n Inda: Implcatons of Accelerated Economc Growth and Poverty Allevaton. Internatonal Food Polcy Research Insttute, Washngton, D.C. SHAW, B. (1985). Agrculture n the Papua New Gunea Economy. Insttute of Natonal Affars, Port Moresby. TIMMER, P., FALCON, W. AND PEARSON, S. (1983). Food Polcy Analyss. Johns Hopkns Unversty Press, Baltmore. WHITE, H. (1980). A heteroscedastcally-consstent covarance matrx and a drect test for heteroscedastcty. Econometrca 48(4):

13 12

14 Table 2 Income elastctes of demand for major food groups * Food group Share-weghted ncome elastcty of demand Cereals 0.48 Meats, poultry and fsh 0.77 Frut, vegetables and nuts 0.67 Root crops 0.49 * Derved from data n Table 1. 13