A Kamakawa
Associate Professional Officer
AGAD, FAO, Rome
The Livestock Development Planning System version 2 (LDPS2), developed by the Animal Production and Animal Health Division of FAO in 1997, is a computer-based planning and training tool designed for livestock development planners to assist them in decision-making.
LDPS2 assists planners to:
identify and quantify the herd/flock composition and size required to meet the given demand for meat and milk;
identify and quantify the feed and livestock constraints in meeting demand; and
assess various livestock development programs or interventions.
The module is an Excel workbook and has three main calculation routines; demand-driven routine, resource-driven routine and herd growth routine (for ruminants). LDPS2 consists of eight interrelated sheets (including two hidden sheets), each one assuming a particular function. There are 19 animal systems modelled in LDPS2. The 7 broad animal categories (dairy cattle, beef cattle, sheep, goats, buffaloes, pigs and poultry) are subdivided into animal systems, with 4 different systems for cattle, 3 systems for sheep, goats and buffaloes and 1 combined system for pigs (intensive and traditional system) and poultry (village production, commercial meat and egg production).
LDPS2 has been applied in three cases - comprehensively in China, small stocks for Iraq and cattle production in Swaziland.
The China case study
Livestock production in China was classified into 18 different production systems, according to agro-ecological zones, and analysed with LDPS2. The main findings were as follows:
Technical discrepancies in statistics over reporting, particularly in beef production and growth rates in recent years.
The calculated manure production from the modern intensive sector for pig and chicken point to the necessity of a recycling system in co-operation with crop farmers.
Scenarios for the year 2005 assuming continuous advances in productivity point to a huge potential for productivity gains.
Feed resources are the most important constraint for expanding livestock production.
Oil for food programme (Iraq)
Livestock production in Iraq in 1988 (pre-war) and 1997 was modelled and a strategy to increase animal-derived dietary protein availability in the short term was designed. Import of sufficient quantities of parent stock (fertile eggs or chicks) and feed concentrate for village and commercial poultry as well as for dairy cattle and sheep was calculated by LDPS2 to recover the production level equal to or more than that of 1988.
Cattle production in Swaziland
Cattle production in Swaziland was divided into four production systems by breed and land tenure (exotic dairy, communally grazed, commercial and government ranches) and analysed with LDPS2. For the dairy, economical projection by the government on the milk demand in the following decade was evaluated and the necessity to import milk and dairy breeds was pointed out, as were the feed resource constraints to expand herd growth. As for beef production, improvement of productivity was highlighted as a major policy target; how fertility rate affects herd growth and efficient resource use was technically simulated. Also some technical inconsistency in statistics was revealed. The analysis was documented as a training manual.
LDPS2 and the user's guide are available as a diskette and a hard copy as well as on the FAO
homepage, http://www.fao.org/waicent/faoinfo/agricult/aga/.
Paul Riethmullero & Ruth Stroppiana
The University of Queensland
Brisbane, Australia
Substantial economic and social development in Asia over the last twenty years or so have led to increased disposable incomes and improved standards of living for many people in the region. The purpose of this paper is to investigate the relationship between the consumption of food and income, prices and socio-economic factors over the period 1965 to 1993 in Indonesia, Thailand and the Philippines. Ten different food items will be analysed in each of these countries using a single equation model for each item. The way demand responds to changes in economic and socio-economic factors may vary for different food commodities and among consumers living in different countries. Since the same set of explanatory variables is used in each model, the results of this study will shed light on whether this is the case. Besides allowing this issue to be examined, the econometric results will provide estimates of price and income elasticity for the foods investigated. Elasticity estimates can be used in forecasting demand and as a guide in policy decisions for domestic governments. Knowledge of the extent to which these factors influence food consumption is important because if the decline in the consumption of a particular food has primarily resulted from the effect of prices, then those involved in the production and marketing of that food could focus on reducing production and marketing costs to maintain or expand its market. However, if the shifts in consumption result from changes in other factors, an alteration in price (due, for example, to a change in border protection) may not make much difference to consumption.
Trends in food consumption
Most of the food consumption data used in this paper come from the International Economic Databank.1 The data include direct consumption of the commodity and consumption of any processed products derived from that commodity. Per capita consumption of each food was calculated from the aggregate information for each of the years for which data were available. Although the types of foodstuffs included were selected to represent foods that are important in the diet, the level of de-aggregation of the food items was determined mainly by the accuracy and availability of the data. Beef, pork, chicken, fish, milk, rice, wheat, sugar, fruit and vegetables were the foods examined. These represent both traditional staples and more typical ‘western’ food and constitute a significant part of the diet in Indonesia, Thailand and the Philippines. Per capita consumption rates in the three countries are presented for 1965 to 1993 in Figures 1 to 6. It is clear from these figures that fish is by far the most important of the non-plant foods. Rice is by far the main staple.
As shown in Figure 1, per capita consumption of fish and chicken in Indonesia grew slowly between 1965 and 1993 and beef consumption declined. Food industry sources blame the lack of growth of meat products on the fact that excessive consumption of meat is viewed as unhealthy by many Indonesians (Evilistna, 7 Feb 1997). The large percentage of Muslims in the population restricts the demand for pork due to the religious prohibition on the consumption of pig meat. Since the mid 1980s, however, per capita consumption of pork has increased. The Chinese population and the large increase in the number of foreign visitors and expatriates living in Indonesia are responsible for the increased demand for pork. Per capita consumption of milk is low and has increased by less than two kilograms from 1965 to 1993. Figure 2 shows that rice consumption has been declining in recent years while wheat consumption has been increasing, albeit from a very small base.
Figures 3 and 4 show that rice, fish, fruit and vegetables are the main items in the Thai diet. Rice is the main staple, typically consumed three times a day. The tropical climate means that wheat production is negligible. Per capita consumption of wheat is low - less than 10 kilograms in 1993. However, due to the greater availability of bread and bakery products, wheat consumption in urban areas is growing strongly (USDA, 1993). In comparison to international averages, Thai people are not large consumers of meat. This may reflect the influence of the Buddhist religion. Over the last few decades, per capita consumption of beef, pork and chicken has increased, but slowly. For example, per capita consumption of beef only increased by about half a kilogram per decade between 1965 and 1993. Both beef and pork are expensive in comparison to chicken. Fish consumption has grown much faster than that of beef, pork or chicken, even though fish is a rather expensive food item in Thailand.2 Per capita consumption of milk is low and showed very little growth until the late 1980s. Fresh milk is a relatively inexpensive food in Thailand. However, many homes, especially in the rural areas, lack refrigerators in which to store perishable foods.
As shown in Figure 5, per capita consumption of fish is well above the demand for the different meats. High retail prices mean that beef remains of little importance in the Filipino diet for all but the high-income households, while per capita consumption of pork and of chicken has grown slowly over the last few decades. Figure 6 shows that vegetable products, especially cereals, play a significant role in the Filipino diet. Government intervention has kept the retail price of rice low because rice makes up a large portion of the food budget. One of the most notable changes in dietary patterns that have occurred in the Philippines is the rapid increase in the demand for wheat since the mid 1980s. Per capita consumption of wheat increased from 12 kilograms in 1985 to almost 27 kilograms in 1993.3 An increase in supply occurred when wheat importation and flour distribution were privatised in the mid 1980s, and declining prices for imported wheat have kept domestic flour prices low (Levin and Lin, 1993, p5).4 Throughout the 1970s and 1980s, the Philippines experienced very high rates of inflation and the government controlled staple food prices. Retail prices for rice and vegetables have not varied much from year to year. However, retail prices of meat have steadily increased, and chicken is relatively cheap in comparison to other types of meat. The retail prices of beef, pork and fish grew at much the same rate until the late 1980s, when the growth rate of beef prices outpaced that for fish and pork.
3 A study conducted by the Philippine Association of Flour Millers indicated that 85 percent of the population eats bread instead
of rice for breakfast, even though the domestic price of bread is generally higher than equivalent amounts of rice (Levin and Lin,
1993, p5).
4 Due to the tropical climate, wheat production is negligible in the Philippines.
The general model
Economic theory provides very little information as to which is the most appropriate functional form for modelling demand. From a preliminary examination of the data, it was suspected that the double-log model provided the best fit for the data in this study. Box-Cox transformation tests were conducted, and these confirmed the view that the double-log constant elasticity functional from provided the best fit for the data.
Using In to denote the natural logarithm, the general model estimated is (1)
lnQijt=β0ij+β1jlnYjt+β2ijlnPijt+β3ijlnSijt+β4jlnOjt+β5jlnDjt+β6jlnAjt+β7jlnUjt+β8jlnFjt+β9Tt+μijt (1)
in which
Qijt | = per capita consumption of food item i in country j in year t |
Yjt | = real per capita personal consumption expenditure in country j in year t |
Pijt | = real retail price of the food item i in country j in year t |
Sijt | = real retail price of a substitute food item i in country j in year t |
Ojt | = openness index in country j in year t |
Djt | = population density in country j in year t |
Ajt | = percentage of the population aged 65 years and over in country j in year t |
Ujt | = percentage of the population who live in urban areas in country j in year t |
Fjt | = female participation in full-time employment in country j in year t |
Tt | = time |
μijt | = error term |
i | = food items included in the study |
j | = countries included in the study |
t | = year, 1965 to 1993. |
Real personal consumption expenditure (PCE) in the three countries was used as a proxy for income. PCE was derived by deflating local currency personal consumption expenditure taken from IMF (1995) by the consumer price index.
Data on average annual retail prices for each food item and for a close substitute in food retail establishments in Thailand were obtained from the National Statistical Office (1995). Each retail price series was converted to the per kilogram price in baht and deflated by the consumer price index to obtain real prices. Data on average annual retail prices for Indonesia and the Philippines were obtained from statistical yearbooks and other government and non-government sources. Table 1 lists these sources for retail price data in Indonesia and the Philippines. Each retail price series was converted to the per kilogram price in rupiah or pesos, depending on the country. These prices were deflated by the consumer price index for the country to obtain real prices.
As can be seen in Table 1, the retail prices for a number of food items in Indonesia and the Philippines were not available for all of the years analysed. For those years where price data were not available, an instrumental variable approach was used. This involved regressing known values of the retail price series on one or more exogenous variables, the instruments. The estimates from this model were used to obtain estimates of the missing value.5 The general from of the model used to obtain the instrumental variable estimates is (2)
in which
Pijt | = real retail price of the food item i in country j in year t |
Zijkt | = instrumental variable (s) for the real retail price of food item i in country j in yeart |
εijt | = error term |
α0ij and αijk | = parameters to be estimated |
i | = food items |
j | = Indonesia or the Philippines |
t | = year. |
The instruments used depended on the availability of ‘good’ variables. According to Pindyck and Rubinfeld (1991, p220), a ‘good’ instrumental variable is one that is highly correlated with the dependent variable but not correlated with the error term. Once an instrumental variable was selected, it was subjected to the Hausman specification test.
The age structure of the population is likely to be important in determining aggregate food demand. The reason for this is that elderly people and young people often have distinct dietary patterns. In developed countries, for example, elderly people spend less at fast-food restaurants and are less likely to try new foods, while young persons are less committed to custom and tradition and lean more toward new food products (Stigler and Becker, 1977, p83; Riethmuller and Smith, 1993, p60). The situation in developing countries may be that older people are more likely to adhere to traditional foods. Over the period of this study, Indonesia, Thailand and the Philippines have experienced some ageing of the population. Reasons for this include declining birth rates, lower infant mortality, increased access to medical attention and better education and nutrition. 6 To account for the influence of age on consumption, the percentage of the population aged 65 and over in each of the countries is included in the general model. The data came from the World Bank (1995) and the United Nations (1994).
The substitution of goods for time has been considered by many economists since Becker (1965) first wrote on the role of time in influencing choice. When a woman enters the labour force, much of her time is taken by her job, so she is likely to spend less time preparing and cooking food. Because women are generally the primary buyers and cookers of food in Asia, their participation in paid employment may be an important determinant of food consumption. It is expected that, as female participation in the labour force increases, other foods are substituted for traditional staples. T Urbanisation - or the movement of people from rural to urban areas - influences food consumption in a number of ways. It can lead to an increase in the range of foods available for consumption and in the supply of technology to a greater percentage of the population. This may change the type of food consumed. For example, increases in the supply of electricity, microwave ovens and refrigerators to family homes widen the range of foods that can be stored and prepared. Also, because physical activity levels in rural areas tend to be higher than urban activity levels, the diet of rural consumers tends to have a higher proportion of carbohydrate-rich staples. This is particularly the case for rice (Bouis, 1990). Previous studies have shown that increasing levels of urbanisation are associated with changing dietary patterns, away from traditional staples and toward wheat and animal products, such as meat, fresh milk and other dairy products in Asia.7 The percentage of the population living in urban areas in each of the countries included in this study was obtained from IEDB and included in the general model.
7 See Bouis (1991), Huang and David (1993), Rae (1998) and Widjajanti and Li (1995).
For most people, diet varies according to the availability of food resources. In rural areas, people usually attempt to produce some of their own food, even if they are not strictly farmers. Their ability to do so will be reduced with the encroachment of cities and towns on rural areas. Urban encroachment has placed increasing pressure on arable land. Hence, population density per square hectare of arable land is used in this study as a proxy for the ability of households to produce their own food. Arable land data taken from statistics complied by FAO (1996) and population data taken from the World Bank (1995) were used to calculate the population density per square hectare of arable land for the countries over the period 1965 to 1993.
The reason for this is that staples such as rice and vegetables are usually time-consuming to prepare, compared to foods such as bread, meat and dairy products. Data on the percentage of women in full-time employment between 1965 and 1993 in Indonesia, Thailand and the Philippines was taken from IEDB.8
8 IEDB extracted the data from the World Bank World Tables and Social Indicators of Development.
Table 1. Average annual retail prices data sources for Indonesia and the Philippines
Price data sources and food items | Year(s) |
---|---|
Indonesia | |
Indonesia Bureau of Statistics (1995) | |
Yearly average retail prices in Jakarta | |
Beef | 1965–1993 |
Fish (tuna) | 1965–1993 |
Milk (condensed) | 1965–1993 |
Indonesia Central Bureau of Statistics (1994) | |
Average retail prices in Jakarta | |
Rice | 1965–1993 |
Sugar | 1965–1993 |
ABARE (1995) | |
International commodity prices | |
Wheat | 1965–1993 |
FAO Database (1997) | |
Producer prices | |
Oranges | 1965–1993 |
Potatoes | 1965–1993 |
Riethmuller (1997) | |
Retail prices | |
Pork | 1975–1993 |
Chicken | 1975–1993 |
The Philippines | |
Bureau of Agricultural Statistics (1997) | |
Yearly average retail prices country-wide | |
Rice | 1970–1993 |
Beef (rump) | 1970–1993 |
Pork (ham) | 1970–1993 |
Chicken (broiler) | 1970–1993 |
Milk | 1980–1993 |
Philippine National Statistics Office (1995) | |
Average retail prices in Metro Manila | |
Fish (milk fish) | 1980–1993 |
Sugar (refined) | 1980–1993 |
ABARE (1995) | |
International commodity prices | |
Wheat | 1965–1993 |
FAOSTAT Database (1997) | |
Producer price• | |
Bananas | 1965–1993 |
Cabbages | 1965–1993 |
Tariffs and non-tariff barriers restrict consumers' access to a range of goods and services, including food products. To take this into account, an openness index was included in the model. The openness index was calculated by using the formula (X+M)/GDP, in which X is the export of goods and non-financial services, M is the import of goods and non-financial services, and GDP is the gross domestic product at market prices. These data were obtained for each country from the World Bank (1995).
Most of the time-series data collected for this study showed either a positive or a negative trend, indicating that the mean was not consistent over the time series. A common procedure to de-trend time-series data is to first-difference the series until it becomes stationary. This procedure was not used in this analysis because of the limited degrees of freedom and the difficulties involved in the interpretation of the coefficients which have been differenced more than once. As an alternative, a time variable was added to the model to de-trend the data, in the sense that the coefficients of the other variables in the equation do not explain changes in the level of the dependent variable. Instead they explain deviations of the dependent variable from its trend value (Johnson, Johnson and Buse, 1987, p357). Yet another advantage of including a trend variable is that it picks up the effects of any omitted variables - for instance difficult-to-quantity factors such as tastes and preferences. This is useful because it reduces the potential bias in the coefficients of the other variables in the equation.
In estimating this model, the homogeneity condition was imposed and tested at the five-percent level of significance, using an F test9. A high degree of multiple collinearity among the explanatory variables was found in all the estimated equations10. Therefore, ridge regression was used in order to disentangle the separate influences of the explanatory variables. Ridge regression involves adding a constant k (O<k<1) to the variances of the explanatory variables. The value of k for each model was chosen using a ridge trace. The guidelines suggested by Hoerl and Kennard (1970) were used in order to select the ‘best’ minimum value of k.11 Simply stated, these guidelines are when: the system stabilises and has the general character of an orthogonal system; the coefficients have reasonable absolute values; the coefficients change to the proper signs; and the residual sum of squares is not inflated to an ‘unreasonable’ value.
9 The homogeneity condition imposes the restriction that the sum of the price and income elasticity be zero.
10 In a number of instances the pair-wise correlation coefficients between any two regressors was in excess of 0.8.
11 The ridge trace is a two-dimensional plot of values of ridge coefficients for a number of values of k.
Each equation was tested for first-order auto-correlation. This was done by the calculation of the Durban-Watson statistic as well as examining the residual plot of each estimated equation to determine the size and pattern of the residuals. In a number of models, positive auto-correlation was found to be severe. The presence of auto-correlated errors is often a symptom of some error or oversight in the specification of the model - for example, the omission of important variables. In addition, the imposition of the homogeneity restriction generates positive serial correlation in the errors of those equations that reject the restrictions most strongly. This suggests that the rejection of homogeneity in demand analysis may be due to insufficient attention to the dynamic aspects of consumer behaviour (Deaton and Muellbauer, 1980, p312).
While a static approach was employed in this analysis, corrections for auto-correlation introduce some dynamic elements into the model12. Dynamic models are generally employed when the effects of past consumption decisions are allowed for. The advantage of dynamic models is that auto-correlation is less frequent in these models (Houthakker and Taylor, 1970, p35). Therefore, if auto-correlation was present in the general model (1), then the inclusion of a lagged dependent variable was tested for significance and added whenever it was significant. In auto-regressive models, the Durban-Watson statistic does not measure first-order auto-correlation directly, so the Durban h statistic was calculated to test for first-order auto-correlation13.
12 The standardised corrections for first-order auto-correlation are not available in SHAZAM when the homogeneity condition is
imposed.
13 Because of the limited number of observations and because there seemed to be no reason to expect higher order auto-correlation,
the estimated equations were not corrected for such a correlation.
The results
Indonesia
Food is the largest expenditure item in the household budget in Indonesia, constituting more than half of consumer expenditure. In rural areas, where almost two thirds of the population live, food accounts for more than 60 percent of the budget (Widjajanti and Li, 1995, p93). The income estimates in Tables 2 and 3 are significantly different from zero for nine of the foods studied. The exception is pork. The income elasticity of demand for pork is not significantly different from zero, even at the five-percent level of significance. A reason for this is that pork is mainly consumed by the Chinese population, which tends to be more affluent than the other Indonesians. The small coefficients in income elasticity of demand for most foods suggests that income is only moderately responsible for changes in per capita consumption of these foods. The inelastic income coefficients for beef and milk were expected, because per capita consumption of these foods experienced very little change over the period studied14. The income coefficient for rice suggests that future increases in income should not generate much increase in per capita consumption of rice in Indonesia. This result supports the findings of Ito, Peterson and Grant (1989), Huang, David and Duff (1991) and Oka and Rachman (1991). Per capita consumption of chicken was the most responsive to income changes. This suggests that future income growth may result in larger increases in the demand for chicken than for the other foods studied.
The own-price elasticity of demand coefficients for the ten foods in Tables 2 and 3 are the correct signs, statistically significant and close to zero. A possible reason for the small price elasticity of demand for the traditional foods may be the high level of government intervention in controlling food prices, particularly for those foods that are a key part of the diet. Rice is by far the main staple and is typically served at every meal. Indonesians in the civil service are given a rice ration. Even though the quality of the rice may not be particularly high, it is nonetheless available for consumption. Rice self-sufficiency was achieved in the early 1980s, but domestic agricultural policy is still directed toward rice. Prices of imported competing products, such as wheat, have been kept high by tariff and non-tariff barriers to encourage Indonesian rice farmers to keep on producing rice. This has resulted in higher retail prices for these competing products and has limited their consumption. Per capita consumption of wheat - production of wheat is almost insignificant due to climatic conditions - is relatively small and has increased very slowly.
Table 2. Estimated coefficients for traditional foods in Indonesia
Rice | Fish | Vegetables15 | Fruit | Sugar | |
---|---|---|---|---|---|
Income | 0.06*** | 0.04** | 0.08** | 0.10*** | 0.07*** |
Rice price | -0.08*** | - | - | - | -0.03** |
Wheat price | 0.02 | - | - | - | - |
Fish price | - | -0.05** | - | - | - |
Beef price | - | 0.01 | - | - | - |
Vegetable price | - | - | -0.13*** | 0.07** | - |
Fruit price | - | - | 0.05 | -0.18* | - |
Sugar price | - | - | - | - | -0.04** |
Openness index | 0.19*** | - | - | - | 0.15** |
Population density | 0.31*** | 0.26** | - | 0.14*** | - |
Aged 65 and over | 0.51*** | 0.44*** | 0.24*** | 0.09** | 0.13* |
Urbanisation | 0.17*** | 0.29*** | 0.08** | 0.11*** | 0.18*** |
Female participation | 0.27*** | - | - | - | 2.01*** |
Time | -0.01** | 0.01*** | 0.02*** | 0.00* | 0.01*** |
Lagged consumption | 0.19*** | - | - | - | - |
R2 | 0.99 | 0.98 | 0.96 | 0.89 | 0.81 |
Value of k | 0.05 | 0.05 | 0.10 | 0.08 | 0.05 |
DW statistic | - | 1.39 | 1.79 | 1.58 | 1.72 |
Durbin h statistic | 1.14 | - | - | - | - |
Notes: Three, two and one asterisk indicate significance of estimated coefficients using a two-tailed test at the 0.1 percent, 1 percent and 5 percent levels of significance, respectively.
- indicates that the variable was excluded from the estimated equation or not applicable.
15 Lack of accurate retail price data meant that the producer price of potatoes was used to represent change in vegetable prices and the producer price of oranges was used to represent change in fruit prices in Indonesia over the period 1965 to 1993.
Table 3. Estimated coefficients for non-traditional foods in Indonesia
Beef | Pork | Chicken | Milk | Wheat | |
---|---|---|---|---|---|
Income | 0.01** | 0.12 | 0.37*** | 0.02*** | 0.02*** |
Beef price | -0.08* | - | 0.13** | - | - |
Pork price | - | -0.13*** | - | - | - |
Chicken price | - | 0.01*** | -0.50*** | - | - |
Fish price | 0.07 | - | - | - | - |
Milk price | - | - | - | -0.02*** | - |
Wheat price | - | - | - | - | -0.08*** |
Rice price | - | - | - | 0.01*** | 0.06*** |
Openness index | -0.01*** | - | - | 0.13*** | 0.13*** |
Population density | 0.01*** | - | - | 0.56*** | - |
Aged 65 and over | - | 0.40*** | 1.14*** | 0.29** | 0.19** |
Urbanisation | 0.44*** | 0.26*** | 0.69*** | 0.21*** | 0.13*** |
Female participation | - | - | - | 0.35*** | 0.06*** |
Time | -0.00*** | 0.01*** | 0.02*** | 0.00 | 0.02*** |
Lagged consumption | - | - | - | - | 0.30*** |
R2 | 0.59 | 0.80 | 0.99 | 0.93 | 0.95 |
F statistic | |||||
Value of k | 0.04 | 0.20 | 0.05 | 0.20 | 0.30 |
DW statistic | 1.49 | 1.87 | 1.65 | 1.40 | - |
Durbin h statistic | - | - | - | - | 1.15 |
The government also intervenes in the meat market. Domestically produced chicken meat is protected against imported beef and dairy products. The unresponsiveness of the non-traditional foods to price changes may also be related to the low consumption of these foods. High retail prices of fresh milk and lack of refrigeration in Indonesian homes restrict the consumption of milk. A more commonly consumed product, especially in rural areas, is sweetened condensed milk dissolved in boiled water (Riethmuller, 1997, p51).16
For most of the period under study, Indonesia maintained high tariff rates on many agricultural and food products. In addition, a range of non-tariff barriers was being used and there were outright prohibitions on a number of foods (AgExporter, 1993, p7). The openness coefficients are significant for five foods - rice, sugar, beef, milk and wheat. The positive coefficients for rice, sugar, milk and wheat suggest that as tariff barriers in general decline, per capita consumption of these foods may increase. On the other hand, the negative coefficient for beef suggests that consumption of beef will decrease as tariff barriers are lowered. These findings are important because under the WTO accords and as part of the IMF restructuring package for the Indonesian economy negotiated in 1997 and 1998, tariffs in Indonesia will decline.
Indonesia is predominantly an agricultural country. The agricultural sector employs just under 50 per cent of the labour force and accounts for about one-fifth of GDP (World Bank, 1996). In comparison to many other Asian countries, Indonesia has abundant arable land resources and a relatively low population density per square hectare of arable land - at least on some islands. However, strong population growth, industrialisation and urban sprawl associated with economic growth mean that available agricultural land is being reduced. The coefficients on population density are significant for rice, fish, fruit, beef and milk. The positive relationship between population density and consumption of these foods implies that further declines of agricultural resources may be associated with increased consumption of these foods. This could be due to this variable picking up the effect of economic growth on food consumption. As well, it could be due to supply-side developments in which government policy has shifted the production focus away from many traditional crops toward the production of higher-value commodities such as fruit, meat and dairy products.
With the exception of beef, the coefficients on age structure were positive and significant for all of the foods studied. This indicates that as the number of people over the age of 65 years has increased, so too has consumption. If the elderly do revert back to a traditional diet, then the age coefficient - or the age elasticity - for the demand of traditional foods should be more responsive than in the case of the non-traditional foods. With the exception of chicken and possibly milk, this was found. Interestingly, the coefficient on the age variable in the rice model was the second highest of those estimated. Its value of 0.51 indicates that for a ten-percent increase in the proportion of the population aged 65 years or above, per capita rice consumption would be expected to increase by 5.1 percent, all else remaining the same. The highest value for the coefficient on age was found in the model estimated for chicken. Its value (1.14) indicates that a ten-percent increase in the proportion of elderly people in the population could increase consumption by 11.4 percent, again all else remaining the same. It may be that not all of this is attributable to the influence of age, or that the age variable is picking up factors associated with overall improvements in living standards, including income. The positive age coefficient for milk could reflect the recent implementation of nutrition programmes by the Indonesian government aimed at increasing consumption of milk among the elderly to combat calcium deficiency.
The industrialisation of Indonesia over the last twenty years or so has seen large numbers of people moving from farming areas and fishing villages to urban areas in search of higher wages and better opportunities. The percentage of the population living in urban areas in Indonesia more than doubled from 16 percent in 1965 to 33 percent in 1993. Therefore, it is not altogether surprising that the coefficients on urbanisation are statistically significant for all of the foods studied. The relatively large urbanisation coefficients for the perishable foods - chicken, beef, pork, fish and milk - could be due in part to urban areas generally having superior retailing and distribution systems compared to rural areas (Young, Twyford-Jones, Logie and Franks, 1995). The growing number of foreign visitors and expatriates living in the cities may be an additional factor responsible for the positive relationship between demand for these foods and urbanisation. The prices of beef and milk are high relative to other foods such as chicken. Therefore, the responsiveness of demand for these foods to urbanisation may reflect the income differential between people living in urban and rural areas.
Rising real incomes increase the opportunity cost of time and the supply of labour. The coefficients on female participation in the labour force were positive and statistically significant for rice, sugar, milk and wheat. The female participation coefficient for sugar is relatively large and suggests that sugar may be a proxy for convenience foods, as many of them have a high sugar content. The responsiveness of the demand for sugar may reflect the growth in the snack food industry over recent years.
Thailand
Until mid 1997, Thailand experienced over three decades of strong economic growth. Per capita GDP grew at an average of 7 to 8 percent per year in real terms during the 1960s and 1970s (National Statistical Office, 1995). As shown in Tables 4 and 5, the income coefficients for all of the foods studied were significant at the five-percent level or better. Per capita consumption of sugar, fish and pork were the most responsive to income changes. This suggests that future income growth may result in larger increases in the demand for these foods than for the other foods included in this study. The negative income elasticity of demand for rice (-0.01) is in agreement with the findings by Ito, Peterson and Grant (1989), Huang, David and Duff (1991) and Huang and David (1993).
Even though all of the own-price elasticity of demand coefficients for the ten foods are statistically significant, with the exception of fish, sugar and pork, the demand for traditional and non-traditional foods is not very responsive to price changes. The relatively high own-price elasticity of demand for fish probably reflects the high retail price of fish and the availability of substitutes for fish in Thailand. The highly significant cross-price elasticity of demand for fish with respect to beef suggests that beef is a substitute for fish there.
Table 4. Estimated coefficients for traditional foods in Thailand
Rice | Fish | Vegetables17 | Fruit | Sugar | |
---|---|---|---|---|---|
Income | -0.01*** | 0.39*** | 0.21*** | 0.07*** | 0.48** |
Rice price | -0.04*** | - | - | - | 0.01 |
Wheat price | 0.05** | - | - | - | - |
Fish price | - | -0.73*** | - | - | - |
Beef price | - | 0.33*** | - | - | - |
Vegetable price | - | - | -0.37* | 0.05 | - |
Fruit price | - | - | 0.16* | -0.12** | - |
Sugar price | - | - | - | - | -0.49** |
Openness index | 0.02** | -0.30** | - | 0.13*** | - |
Population density | - | 0.69** | - | - | - |
Aged 65 and over | 0.09*** | 0.48*** | 0.13** | 0.35** | 0.39** |
Urbanisation | -0.03*** | 0.33*** | - | 0.25*** | - |
Female participation | -0.46*** | 1.66*** | 0.44** | 0.20*** | - |
Time | -0.01 | 0.02*** | 0.01*** | 0.02 | 0.01*** |
Lagged consumption | - | - | - | 0.21*** | - |
R2 | 0.73 | 0.85 | 0.91 | 0.93 | 0.80 |
Value of k | 0.05 | 0.05 | 0.10 | 0.20 | 0.10 |
DW statistic | 1.56 | 1.48 | 1.69 | - | 2.39 |
Durbin h statistic | - | - | - | 1.70 | - |
In comparison to many other Asian countries, Thailand has abundant arable land resources and a low population density. The coefficients on population density are only significant for beef, chicken, milk and fish. The positive relationship between population density and consumption of these foods implies that further declines of agricultural resources may be associated with increased per capita consumption of these foods. These results make sense because the domestic agricultural policy has shifted the production focus away from traditional crops toward high value-added products such as meat, dairy and marine products (Young, Twyford-Jones, Logie and Franks, 1995, p54).
Thailand has a relatively young population, though increased life expectancy and improved standards of living are slowly changing the age structure of the population. The age coefficients are significant for all foods, with the exception of pork. This suggests that the age structure of the population may affect food consumption patterns in Thailand. The coefficients are positive for chicken, milk and all of the traditional foods. The positive relationship between the greying of the population and the demand for traditional food staples imply that the elderly in Thailand may prefer to consume increasing amounts of these foods over the other foods included in this analysis. This provides some evidence that as Thai people age, they go for the traditional diet. This seems to be a feature of Asian economies. The positive coefficient for milk is probably explained by the existence of government programmes to increase milk consumption among schoolchildren and the elderly. The positive coefficient for chicken may be a result of the income effect on the consumption of the relatively less expensive chicken meat. The negative coefficients for wheat and beef suggest that the older population may consume less of these foods than the younger population.
Table 5. Estimated coefficients for non-traditional foods in Thailand
Beef | Pork | Chicken | Milk | Wheat | |
---|---|---|---|---|---|
Income | 0.15** | 0.29*** | 0.02*** | 0.09** | 0.01*** |
Beef price | -0.15** | - | - | - | - |
Pork price | - | -0.37** | 0.01* | - | - |
Chicken price | - | 0.09* | -0.03*** | - | - |
Fish price | 0.01** | - | - | - | - |
Milk price | - | - | - | -0.17** | - |
Wheat price | - | - | - | - | -0.06* |
Rice price | - | - | - | 0.08 | 0.05* |
Openness index | - | 0.45*** | 0.06** | - | 0.09*** |
Population density | 1.55** | - | 0.31** | 0.41** | - |
Aged 65 and over | -0.27** | - | 0.27** | 0.09*** | -0.3*** |
Urbanisation | 0.88*** | - | 0.10*** | 0.34*** | 0.50* |
Female participation | - | 0.13*** | - | - | -0.04** |
Time | 0.00** | 0.02*** | 0.01*** | 0.01** | 0.00 |
Lagged consumption | - | - | 0.16** | - | 0.24*** |
R2 | 0.83 | 0.97 | 0.94 | 0.89 | 0.79 |
Value of k | 0.08 | 0.08 | 0.08 | 0.06 | 0.04 |
DW statistic | 1.79 | 1.33 | - | 1.68 | - |
Durbin h statistic | - | - | 1.26 | - | 1.68 |
From 1965 to 1993, the percentage of the population living in urban areas increased from less than 13 to more than 24. The results in Tables 4 and 5 suggest that urbanisation has influenced the consumption of food in Thailand. The negative coefficient for rice consumption in urban areas and the positive coefficient for wheat are in agreement with Huang and David (1993). The demand for beef and wheat was found to be the most responsive to changes in the urbanisation variable. This means that relatively large increases in the consumption of beef (the most expensive meat product) and wheat may be associated with future increases in the rate of urbanisation in Thailand.18 Western-style fast-food restaurants have mushroomed in the last few years. Therefore, increased consumption of wheat and meat products in urban areas could reflect increasing consumption of fast-food fare. Government estimates suggest that 23 percent of the food budget is spent on food away from home. In Bangkok, this figure jumps to 46 percent (Kurtz, 1997, p.12). These results may also be picking up an income effect, since incomes in urban areas tend to be higher than in rural areas. The income disparity between urban and rural areas in Thailand is one of the most pronounced in Asia and has sharpened with economic growth (USDA, 1993).
The Philippines
After the unsuccessful attempt to industrialise in the 1960s, the Philippine economy experienced a series of declines and has been characterised by political unrest and high levels of unemployment. In the mid-1980s, the newly elected Aquino government committed itself to the promotion of capital-intensive industries rather than labour-intensive goods (East-West Centre, 1989, p28). However, in comparison to that of many neighbouring countries, real per capita GDP growth in the Philippines has been low-less than two percent per year during the 1980s (World Bank, 1996). Food accounts for more than half of household expenditure, and very little of it is food away from home (Philippine National Statistics Office, 1995). As shown in Tables 6 and 7, the income elasticity of demand coefficients for all of the foods studied are significantly different from zero. Yet, with the exception of pork, chicken and possibly fish, those for the other foods are close to zero. The income elasticity coefficients for pork and chicken suggest that if per capita real income increases in the Philippines, consumers may prefer pork and chicken over the other foods included in this analysis. The negative income elasticity of demand for fish is in agreement with the results of previous research and suggests that fish is an inferior produce in the Philippines.19
With the exception of pork and chicken, the own-price elasticity of demand coefficients for the ten foods are statistically significant. The coefficients for the traditional foods are rather small in comparison to the non-traditional foods. This is as would be expected because government controls have been used to regulate the price of food staples to control inflation. The own-price elasticity for beef (-1.04) is very large. Holding all other factors constant, this implies that a ten-percent increase in price would result in a 10.4-percent decrease in per capita consumption of beef. The responsiveness of beef demand to prices in the Philippines was seen in early 1997 when an increase in the price of beef, by only a few pesos, led to a decrease in the consumption of beef and a dramatic decrease in live cattle imports. The large cross-price elasticity of demand for beef with respect to fish suggests that fish is a close substitute for beef in the Philippines. The estimated own-price elasticity coefficients for pork and chicken are not statistically different from zero, even at the five-percent level of significance. These results differ somewhat from those of a number of other studies.20
20 For example, see Costales (1990).
Table 6. Estimated coefficients for traditional foods in the Philippines
Rice | Fish | Vegetables | Fruit | Sugar | |
---|---|---|---|---|---|
Income | 0.12*** | -0.25*** | 0.02*** | 0.14** | 0.16* |
Rice price | -0.21*** | - | - | - | 0.02 |
Wheat price | 0.09*** | - | - | - | - |
Fish price | - | -0.37** | - | - | - |
Beef price | - | 0.62* | - | - | - |
Vegetable price | - | - | -0.03** | 0.01*** | - |
Fruit price | - | - | 0.01*** | -0.16*** | - |
Sugar price | - | v | - | - | -0.18* |
Openness index | - | -0.20*** | 0.22*** | - | - |
Population density | - | -0.16*** | 0.15*** | 0.11*** | 0.01** |
Aged 65 and over | 0.27** | - | 0.23*** | v | 0.12*** |
Urbanisation | 0.26*** | - | 0.29*** | 0.27*** | - |
Female participation | 0.33*** | -0.42*** | - | v | - |
Time | -0.01** | 0.00** | 0.01*** | 0.00** | 0.00*** |
Lagged consumption | 0.25*** | - | - | - | - |
R2 | 0.99 | 0.63 | 0.90 | 0.86 | 0.74 |
Value of k | 0.03 | 0.08 | 0.10 | 0.20 | 0.10 |
DW statistic | - | 1.59 | 1.53 | 2.09 | 1.68 |
Durbin h statistic | 0.29 | - | - | - | - |
Although liberalisation measures are in place, the Philippines mainly followed import-substitution policies during much of its development. As a result, tariff rates on many agricultural products have been high and there is strict import licensing on some foods, such as pork and beef products (USDA, 1993). The openness coefficients are highly significant for six of the foods examined. The positive coefficient for beef, wheat and vegetables suggests that as tariff barriers decline, per capita consumption of these foods, which are largely imported (with the exception of vegetables), may increase. On the other hand, the negative coefficients for chicken, pork and fish suggest that consumption of these foods, in which the Philippines is almost self-sufficient, will decrease as tariff barriers are lowered. These findings are important because under the WTO accords, tariffs for a number of foods in the Philippines, especially for pork and chicken, will decline. Already in 1996, the government eliminated all non-tariff barriers for food imports. The only exception is rice, for which a quota remains in effect. The new regulations will allow the entry of previously banned food imports such as fresh onions and potatoes (Wade and Canono, 1997, p18).
Table 7. Estimated coefficients for non-traditional foods in the Philippines
Beef | Pork | Chicken | Milk | Wheat | |
---|---|---|---|---|---|
Income | 0.08*** | 0.36*** | 0.39*** | 0.12*** | 0.19** |
Beef price | -1.04*** | - | - | - | - |
Pork price | - | -0.74 | 0.20 | - | - |
Chicken price | - | 0.38*** | -0.59 | - | - |
Fish price | 0.96*** | - | - | - | - |
Milk price | - | - | - | -0.18*** | - |
Wheat price | - | - | - | - | -0.23** |
Rice price | - | - | - | 0.06 | 0.04** |
Openness index | 0.98*** | -0.53*** | -0.70*** | 0.43*** | - |
Population density | 0.23*** | -0.50*** | 0.32*** | 0.12*** | - |
Aged 65 and over | -0.81** | 0.72*** | 0.41*** | 0.63*** | 0.28*** |
Urbanisation | 0.61*** | - | 0.22*** | 0.11*** | 0.33** |
Female participation | 1.49*** | - | - | - | - |
Time | 0.01*** | 0.00*** | 0.01*** | 0.01*** | 0.00** |
Lagged consumption | - | - | - | - | - |
R2 | 0.58 | 0.66 | 0.87 | 0.94 | 0.88 |
Value of k | 0.08 | 0.06 | 0.05 | 0.08 | 0.10 |
DW statistic | 1.53 | 1.49 | 1.43 | 1.66 | 1.59 |
Durbin h statistic | - | - | - | - | - |
With the exception of fish and fruit, all of the coefficients on age structure are significant at the five-percent level or better. The most responsive foods to the changes in the age structure of the population are beef, pork, milk and chicken. The negative coefficient for beef suggests that a decrease in per capita consumption of beef may accompany further ageing of the population. On the other hand, the demand for pork, milk and chicken is expected to increase with the greying of the population.21 These results may be partially due to the price effect, because pork and chicken are relatively inexpensive meat products in comparison to beef.
Future consumption of animal products in Indonesia, Thailand and the Philippines
The demand models estimated in this analysis provided elasticity estimates that may be used to forecast changes in food demand. This section will illustrate how the elasticity estimates could be used to forecast changes in demand using the @RISK spreadsheet program (Palisade, 1996). The demand for animal products will be the focus of this section because these are the commodities likely to be most sensitive to future development in the economy.
Consumption of animal products in the three countries is low in comparison to western countries and showed little signs of growth between 1965 and 1993. Using the general model (1), per capita consumption of animal products in Indonesia, Thailand and the Philippines was found to be of the form of (3):
Qj = f(Yj,Pj,Sj,Oj,Dj,Aj,Uj,Fj) (3).
From (3) it is possible to obtain (4), which relates the percentage change in per capita consumption to the percentage change in each of the explanatory variables, and the elasticity associated with each of these variables.
Q*j = εYjY*j + εPjP*j + εsjS*j + εOjO*j + εDjD*j + εAjA*j + εUjU*j + εFjF*j (4)
in which
εkj | = the elasticity of each of the explanatory variables k with respect to per capita beef consumption in country j |
Q*j | = percentage change in per capita consumption of beef in country j |
Y*j | = percentage change in the real income in country j |
P*j | = percentage change in the real retail price of beef in country j |
S*j | = percentage change in the real retail price of a fish in country j |
O*j | = percentage change in the openness index in country j |
D*j | = percentage change in the population density in country j |
A*j | = percentage change of the population aged 65 years and over in country j |
U*j | = percentage change of the population who live in urban areas in country j |
F*j | = percentage change of females in full-time employment in country j |
j | = countries included in the study |
k | = explanatory variables |
Using expression (4), the change in per capita consumption of animal products could be estimated by combining a point elasticity estimate with a forecast of the change in the exogenous variables. This type of analysis could also include different income scenarios, such as low, medium and high-income growth. The approach taken here is more sophisticated than this, however, in that the distribution for each of the elasticity coefficients will be specified, and combined with the distribution for the exogenous variables. This will provide a distribution of changes in consumption for each of the foods, rather than a simple number corresponding to each of the income scenarios. The elasticity coefficients and their standard errors estimated in the previous section are used to provide the distribution of elasticity values in the simulations. The elasticity estimates which were found not to be significant at the five-percent level in the previous section are excluded from the current analysis. For those that were significant at the five-percent level or better, the elasticity estimates were specified in the @RISK spreadsheet program as normal distributions. In the distributions, the mean elasticity was the estimated coefficient, and the standard error of the estimate was the standard deviation of the distribution. The low, medium and high income growth estimates for Indonesia, Thailand and the Philippines in the medium term used in the analysis are 0,2 and 5 percent respectively. The estimates selected for use in this analysis are based on judgement influenced by forecasts of real GDP growth in government and non-government reports.
The percentage change in the exogenous variables were specified as triangular probability distributions in the @RISK spreadsheet program.22 This means that the distribution has a minimum value, a most likely value and a maximum value. The values for each of the explanatory variables are as follows:
In the aftermath of the Asian currency crisis there has been much uncertainty surrounding expected changes in food prices in Asia. Therefore, a wide range of changes for food prices were used in this analysis. The minimum, mean and maximum percentage changes in real retail food prices used in this analysis are -10, 0 and 10 percent respectively.
Past trends were used to estimate future changes in the openness index. Between 1965 and 1993, the openness index in the three countries grew between two and three percent per year. With the commitments made under the WTO accord by the governments in these countries, these growth rates are expected to continue in the medium term. Therefore the minimum, mean and maximum changes in the openness index used in this analysis are 1,2 and 4 percent respectively.
Declining agricultural land resources accompanied by high population growth rates in Indonesia mean that the percentage change in population density per square hectare of arable land may be relatively high over the next ten years or so. Therefore, the estimated values for population density in Indonesia used in this study are 1, 2 and 4 per cent. On the other hand, Thailand and the Philippines have an abundance of agricultural land resources accompanied by relatively low population growth rates. This means that the percentage change in population density per square hectare of arable land may be relatively low over the next ten years or so. Therefore, the estimated values used in this study are 1, 2 and 3 percent.
Projected changes in the age structure of the population in each country were used to estimate a minimum, most likely and maximum change in the percentage of the population aged 65 years and over in Indonesia, Thailand and the Philippines. The age structure of their population is estimated to change slowly in the medium term (World Research Institute, 1998). The minimum, mean and maximum estimated changes in the percentage of the population aged 65 and over used in this analysis are 0,1 and 2 percent respectively.
As mentioned earlier, Indonesia, Thailand and the Philippines are predominantly agricultural countries, with over half of the population living in rural areas. Forecasts of urbanisation rates predict only moderate increases in the urbanisation of Indonesia and Thailand (World Research Institute, 1998). Therefore, the minimum, mean and maximum changes in the urbanisation rate of these countries used in this analysis are 1, 2 and 3 percent respectively. However, forecasts for the Philippines predict larger increases in the rate of urbanisation in the medium term. Thus the minimum, mean and maximum changes in urbanisation for the Philippines are 2, 3 and 4 percent respectively.
Female participation in full-time employment in Indonesia has increased slowly over the last few decades. High unemployment and a declining average annual growth rate of the labour force mean that the percentage change in female participation in full-time employment may be limited in the medium term. The minimum, mean and maximum changes in female participation in full-time employment for Indonesia used in this analysis are 0, 1 and 2 percent respectively. Female participation rates in full-time employment in Thailand and the Philippines have not changed much over the last few decades, though there has been a slight decline in female participation rates in these countries during the last ten years or so. The minimum, mean and maximum changes in female participation in full-time employment for Thailand and the Philippines used in this analysis are -1, 0 and 1 percent respectively.
Per capita consumption of beef, pork, chicken and milk in Indonesia, Thailand and the Philippines under three growth scenarios were derived by running three simulations for each food item. Each simulation had 500 iterations and the results were used to calculate their minimum, maximum and mean values, as shown in Table 8.
Table 8. Percentage change in per capita consumption of animal products in Indonesia, Thailand and the Philippines
Indonesia 0% 2% 5% | Thailand 0% 2% 5% | Philippines 0% 2% 5% | |
---|---|---|---|
Beef | |||
Minimum | -0.15 -0.12 -0.07 | -2.17 -1.79 -1.21 | -15.10 -14.93 -14.69 |
Mean | 0.89 0.91 0.94 | 4.60 4.90 5.35 | 5.26 5.42 5.66 |
Maximum | 2.46 2.49 2.53 | 10.99 11.41 12.07 | 26.97 27.11 27.32 |
Pork | |||
Minimum | -2.82 -2.15 -1.30 | -12.82 -11.60 -10.40 | |
Mean | 1.62 2.20 3.07 | -1.83 -1.07 -0.70 | |
Maximum | 7.17 7.89 8.97 | 9.82 10.68 11.96 | |
Chicken | |||
Minimum | -4.33 -3.94 -3.35 | -0.53 -0.48 -0.39 | -7.90 -6.99 -5.63 |
Mean | 2.52 3.26 4.37 | 1.40 1.44 1.51 | 0.79 1.57 2.74 |
Maximum | 8.56 9.45 10.80 | 3.67 3.71 3.76 | 8.49 9.17 10.19 |
Milk | |||
Minimum | 1.10 1.14 1.20 | -1.20 -1.01 -0.72 | -1.56 -1.32 -0.97 |
Mean | 2.30 2.35 2.40 | 2.01 2.19 2.47 | 1.52 1.77 2.13 |
Maximum | 3.58 3.62 3.70 | 4.88 5.05 5.29 | 4.58 4.80 5.14 |
Between 1965 and 1993, per capita consumption of beef in Indonesia, Thailand and the Philippines experienced very little change. As shown in Table 8, the results of the three scenarios for beef in each country do not differ greatly, because in the general model (1) beef was found to be relatively unresponsive to changes in income in these countries. For example, per capita consumption of beef in Indonesia is estimated to increase on average by just over 0.89 percent under the low-income growth scenario to about 0.94 percent under the high-income growth scenario. The most optimistic outcome for the different rates of growth was for consumption to increase by more than 27 percent in the Philippines. If such a rate of growth could be sustained for four to five years, per capita beef consumption in the Philippines could increase by 110 to 136 percent. At the other extreme, over the next four to five years beef consumption could fall by over 75 percent. The average increases in per capita consumption of beef under the three income growth scenarios in Thailand and the Philippines are much higher than for the other animal products included in this analysis. These results are not altogether surprising, given that per capita consumption of beef in these countries is well below the demand for most other meat products.
Relatively large increases in per capita consumption of pork in Thailand and the Philippines may also occur under the most optimistic income growth scenario. An interesting result is that the average changes in per capita consumption of pork in the Philippines under all three income growth scenarios are negative. During the last few decades, per capita consumption of pork in the Philippines has been well above the demand for beef and chicken. Therefore, this result may be due to consumption of pork approaching saturation levels.
The results of the three simulations for chicken and milk in Thailand vary less than those for beef and pork because these foods are less responsive to changes in income. These results suggest that consumption of these products in Thailand is expected to be only moderately affected by income changes. An interesting result in Table 8 is that the minimum expected growth rate for milk in Indonesia is greater than those under all three growth scenarios. The minimum growth rates for milk in Thailand and the Philippines are negative under all three scenarios.
By comparing the results of the three simulations for each of the animal products, it can be seen that the change in demand for each of the animal products under low, medium and high income growth scenarios is small. This is because the estimated income elasticity of demand coefficients for these products are relatively close to zero. Therefore, per capita consumption of these products is expected to be only moderately affected by future income changes. However, there is substantial variation in the minimum, mean and maximum responses of demand for each of the animal products - especially in the case of beef and pork. This indicates that future demand for these foods may be largely affected by the social development of these countries.
These results add to the findings of the previous analysis in that changes in consumption of animal products in Indonesia, Thailand and the Philippines are expected to be only moderately affected by income changes. However, the small response of demand to income changes found in this study may be due to the substitution of domestic products for imported products or higher-quality products. In addition, other factors not included in this analysis may also affect food consumption in the future, such as increased foreign travel, changing household arrangements and increased demand for healthy and safe food products.
Concluding comments
This paper has been concerned with analysing the relationship between economic and socio-economic factors and the demand for traditional and non-traditional foods in three Asian countries. The main findings of the study are:
The income elasticity estimates for the foods studied are relatively close to zero. Small income elasticity of demand for a number of these foods is not altogether surprising and was also found by previous research. Holding all other factors constant, these results suggest that per capita consumption of most of these foods is expected to be only moderately affected by any change in income levels.
The price elasticity of demand for the foods studied generally had the correct signs, but was not significantly different from zero. This suggests that prices are only moderately responsible for changes in per capita consumption of these foods. A possible reason for this is that food prices, especially staple food prices, are largely controlled by the government-this is especially the case in Indonesia and in the Philippines. Another reason for the small own-price elasticity of demand for some of these foods may be due to the retail price data used in this study. While the consumption data are rather aggregated, the price series are averages of monthly prices for de-aggregated food products or individual brand prices. In addition, there are large price differentials between similar products of the same type of food. This is especially the case for domestic and imported beef. These limitations of the price data apply to a number of foods included in this study and should be taken into consideration when interpreting the price elasticity estimates presented here.
Generally, the results of this study suggest that food consumption is affected by social development as well as by changes in prices and income. The extent to which the socio-economic factors influence demand vary widely among the foods under survey. Therefore, identifying the trends in the socio-economic characteristics of the population may become increasingly important to understand what drives food markets in each country.
The future growth in per capita consumption of animal products were estimated using sensitivity analysis. Bearing in mind the limitations of such an analysis, the results provide some indication of the expected changes in demand for animal products in Indonesia, Thailand and the Philippines. These results illustrate the importance of elasticity estimates in arriving at judgements about future market developments.
There are several ways in which this research could be extended. For example, a more de-aggregated
approach could be taken. It is well known that there is a large disparity between incomes in these countries.
This disparity is greatest between urban and rural areas. Had data been available, food consumption
patterns in rural and urban areas would have been analysed separately. The lack of data also precluded
making any allowance for rural producers who grow and consume their own produce, or including
information on price variations among regions or cities within a country. Nor was it possible to include other
variables which may also affect food consumption patterns, such as household size, average education
attainment levels, infrastructure availability or even religious preferences.
References
ABARE. 1995. Commodity Statistical Bulletin. Canberra: AGPS
AgExporter. 1993. Diet revolution in Pacific Rim cereals: new customer for US foods, in AgExporter, Vol5, No6, pp4–7
Becker, G. 1965. A theory of the allocation of time, in The Economic Journal. Vol75, pp493–517
Bouis, H. 1965. Evaluating demand for calories for urban and rural populations in the Philippines: consequences for nutrition policy under economic recovery, in World Development. Vol18, pp281–299
Bouis, H. 1991. Rice in Asia: Is it becoming a commercial good? Comment, in American Journal of Agricultural Economics. Vol73, No2, pp523–527
Bureau of Agricultural Statistics. 1997. Philippines (personal communication, Feb 1997)
Costales, A. 1990. Livestock in Philippine agriculture, demand projections for meat, feed and corn and prospects for backyard holding of carabao and cattle. Consultant's Report Submitted to the Department of Agriculture, Philippines, and the Asian Development Bank, Manila
Deaton, A & Muellbauer, J. 1980. An almost ideal demand system, in The American Economic Review. Vol70, No3, pp312–326
De Vega, M & Fisher, B. 1983. An Empirical Study of Food Consumption in Rural and Urban Households of the Philippines. Research Report No9, Sydney: Department of Agricultural Economics, University of Sydney, Australia
Directorate General of Livestock Services. 1992. Livestock development in Indonesia. Jakarta: Ministry of Agriculture
East-West Centre. 1989. Asia Pacific Report. Focus: China in the reform era. Honolulu
Economist, The. 1999. http://www.economist.com. 4 Feb 1999
FAO. 1996. FAO Production Yearbook. Rome: United Nations (including previous years)
FAOSTAT Database. 1997. http://apps.fao.org. 22 Aug 1997
Hoerl, A & Kennard, W. 1970. Ridge regression biased estimation for non-orthogonal problems and Ridge regression: applications to non-orthogonal problems, in Technometrics. Vol12, No1, pp55–67 and 69–82
Houthakker, H & Taylor, L. 1970. Consumer Demand in the US: 1929–1970. Cambridge: Harvard University Press
Huang, J & David, C. 1993. Demand for cereal grains in Asia: the effect of urbanisation, in Agricultural Economics. Vol18, No2, pp107–124
Huang, J, David, C & Duff, B. 1991. Rice in Asia: is it becoming an inferior good? Comment, in American Journal of Agricultural Economics. Vol73, No2, pp515–521
IMF. 1995. International Financial Statistics. Washington (including previous years)
Indonesia Central Bureau of Statistics. 1994. Statistical Yearbook of Indonesia. Jakarta (including previous years)
Indonesian Bureau of Statistics. 1995. Monthly Statistical Bulletin. Jakarta (including previous years)
Ito, S, Peterson, W & Grant, W. 1989. Rice in Asia: is it becoming an inferior good?, in American Journal of Agricultural Economics. Vol 71, No1, pp32–42
Johnson, A, Johnson, M & Buse, R. 1987. Econometrics Basic and Applied. New York: Macmillian Publishing Co
Kurtz, P. 1997. Thailand: marketplace and gateway to Indochina, in AgExporter. Vol19, No5, pp11–14
Levin, C & Lin, C. 1993. Determinants of Wheat Demand: Philippines. Washington: USDA Agricultural and Trade Analysis Division
McKinna, D. 1992. Food Processing Industry Strategy: Volume One. Brisbane: Queensland DPI
National Statistical Office. 1995. Statistical Yearbook of Thailand. Bangkok: Office of the Prime Minister (including previous years)
Oka, M. & Rachman, B. 1991. Food expansion policy and the nutritional status in Indonesia, in Indonesian Food Journal. Vol11, No4, pp51–68
Palisade. 1996. @RISK Advanced Risk Analysis for Spreadsheets. Newfield: Palisade Corporation
Philippines National Statistics Office. 1995. Monthly Bulletin of Statistics. Manila (including previous years)
Pindyck, R & Rubinfeld, D. 1991. Econometric Models and Economic Forecasts. Third edition, New York: McGraw-Hill Inc
Podbury, T, Ladlow, S, Roberts, I, Felton-Taylor, L, & Chaimun, S. 1995. East and Southeast Asian dairy markets, in Australian Commodities. Vol2, No3, pp340–357
Rae, A. 1998. The effects of expenditure growth and urbanisation on food consumption in East Asia: a note on animal products, in Agricultural Economics. Vol18, pp291–299
Rae, A. 1997. Changing food consumption patterns in East Asia: implications of the trend toward livestock products, in Agribusiness. Vol13, No1, pp33–44
Riethmuller, P. 1997. The Indonesian feed and livestock sector, mimeo
Riethmuller, P & Smith, D. 1993. The Asian Pacific live cattle trade: sacred cow or a new opportunity?, in Government policies in the Philippine beef industry discussion apers. St Lucia: University of Queensland, Discussion Paper No6
Rutherford, A. 1996. Locating Value Adding Activities in Northern Australia's Beef Industry: An Interregional Trade Modelling Approach. St Lucia: Department of Economics, University of Queensland.
Smith, D, Franco, M & Riethmuller, P. 1993. Recent Philippine policy developments and their effects on Australian live cattle exports, in Government Policies in the Philippine Beef Industry. St. Lucia: University of Queensland, Discussion Paper No 8
Stigler, G & Becker, G. 1977. De gustibus non est disputandum, in The American Economic Review. Vol67, No2, pp76–89
United Nations. 1994. Compendium of Social Development Indicators in the ESCAP Region. New York: Economic and Social Commission for Asia and the Pacific
USDA. 1993. International Agricultural and Trade Reports: Asia and the Pacific Rim. Washington: Economic Research Service RS-93-6.
Wade, J & Canono, J. 1997. Philippines: familiar market for US agricultural products, in AgExporter. Vol9, No5, pp18–19
White, K. 1993. SHAZAM the Econometrics Computer Program: Users Reference Manual Version 7.0. New York: McGraw-Hill Book Co
Widjajanti, E & Li, E. 1995. Food expenditure patterns in urban and rural Indonesia, 1981 to 1993, in Australasian Agribusiness. Vol3, No2, pp93–113
World Bank. 1995. World Tables. London: The John Hopkins University Press (including previous years)
World Bank. 1996. World Development Report. New York: Oxford University Press (including previous years)
World Resources Institute. 1998. World Resources: A Guide to the Global Environment. New York
Young, C, Twyford-Jones, P, Logie, K & Franks, R. 1995. The East Asian Agribusiness Market: Export Potential for North Queensland. Brisbane: Queensland DPI