4. RICE DEMAND ANALYSIS

Introduction

The 1994 Marrakech Agreement of the General Agreement on Tariff and Trade (GATT) Uruguay Round started a process of agricultural market liberalization. The new round of World Trade Organization (WTO) negotiations, launched in 2000, is expected to bring this process further. The world rice market is a thin one; 90 percent of production and consumption occur in Asia. The GATT/WTO decisions will design a new structure for world agricultural markets, and it is important to understand how this will influence the rice market in the near future.

Japan reached a high level of economic growth earlier than other Asian nations. Recently, the Newly Industrialized Economies (NIEs) of the Republic of Korea, Taiwan Province of China and Singapore have been catching up rapidly and attaining higher per capita incomes. Many Asian nations may eventually reach the economic levels of Japan, Europe and the United States. Japanese consumption behaviour is a key indicator for forecasting the future consumption patterns of Asian nations. For example, the Republic of Korea accepted the same minimum access import requirements in the GATT negotiations as Japan did, and the agriculture sector of Taiwan Province of China is very similar to that of Japan. By investigating Japanese consumption behaviour as representative of high-income consumers, this report will shed some light on the future direction of Asian and world rice demand.

In addition to a general concern about Japanese consumption behaviour, it is also of great interest to ascertain whether rice is a normal or an inferior good, i.e. whether per capita rice consumption goes up or down as income increases. Since rice is a very important food staple in Asian countries, many domestic and international agricultural trade policies are centred on it. Such important agricultural policies would be misdirected if they were based on a belief that rice is an inferior good, without a rigorous and robust estimation of that characteristic.

When assessing food balances, the literature on the rice market is mainly concerned with supply-side factors (Oniki, 1996; Fujiki, 1993; 1998; 1999). Considering the uncertain environment of the rice market in the future, the demand side should not be neglected. In order to obtain an accurate forecast of the impact of liberalization of the Japanese rice market, it is necessary to estimate demand elasticities precisely.

Background

Japan reached high per capita income far earlier than other Asian nations did. As per capita income grows, food consumption patterns change. Many studies report on the Westernization of the Japanese diet: less calorie intake comes from rice and more from animal meat, and the fat content of food has increased. Because of geographical reasons and preferences, in a typical Japanese diet the calorie intake from fish has a larger share than that from meat.

This chapter gives an overview of the time trend of rice consumption patterns. Figures 4.1 to 4.3 show descriptive consumption patterns in Japan. All data are taken from the Food Balance Sheet published Japan's Ministry of Agriculture, Forestry, and Fisheries (MAFF, 1997). These are supply-side data: data on the quantities of food items delivered to consumers are added together.^[7]

Figure 4.1: Annual per capita rice consumption, 1960-1999

Source: MAFF, 2001.

Figure 4.2: Annual per capita meat, poultry and fish consumption, 1960-1999

Source: MAFF, 2001.

Figure 4.3: Average daily per capita calorie intake, 1960-1999

Source: MAFF, 2001by

Figure 4.1 shows the annual per capita rice consumption of Japan. It is well known that aggregate rice consumption in Japan has been declining over recent decades, as is common among high-income countries. Per capita rice consumption peaked in 1962 and had almost halved by the end of the 1990s. Figure 4.2 shows annual per capita meat (beef and pork), poultry and fish consumption. By weight, Japanese households consume more fish than meat and poultry, and this is one of the unique features of the Japanese food consumption pattern. Figure 4.3 shows the average daily per capita calorie intake from meat, poultry and fish and shellfish. Since 1980, meat and poultry have become a larger source of calories than fish. Although Japanese fish consumption by weight is higher than that of meat and poultry, in nutrition terms, meat and poultry provide a higher calorie intake than fish does.

Two considerations should be noted from these figures. First, they provide little information about the price and income elasticities of each commodity. Calorie intake is purely a behavioural variable, and it does not reveal any clear price information; many authors cite a multitude of reasons why calorie elasticity and price and income elasticity are different (Deaton, 1997). Second, the food balance sheet provides macrodata, which may not capture individual household consumption patterns accurately. That is, there may be an aggregation problem. For estimating income elasticities, household survey data should provide a better picture of individual household consumption patterns.

In this report, the income elasticity has been estimated from cross-sectional survey data to shed some light on important issues. Two important questions regarding Japanese rice consumption are investigated: whether rice is an inferior good in Japan; and whether rice is a complement or a substitute for meat, fish and other food items.

Is rice an inferior good in Japan?

Rice is a staple food in Japan, and its great importance in the Japanese diet is well known. In 1995, 10 748 000 tonnes of rice were produced domestically, and 10 485 000 tonnes were consumed. Rice is used by a variety of sectors, but mostly by the household. According to the Management and Coordination Agency (1999), in 1995, 93.21 percent of rough rice was purchased by the milling sector and 74.38 percent of milled rice was consumed directly by households (see Table 4.1).

It is important to understand whether rice is a normal or an inferior good. Japan has one of the highest per capita gross domestic products (GDPs) in the world. If rice were an inferior good, rice consumption would fall as per capita GDP grows. If that were the case, and if Japan were considered as the leading case for other Asian countries, it would be possible to project lower world rice demand in the future as the incomes of Asian nations increase.

Among researchers it has been accepted that income elasticities for rice and other food staples decline as per capita income increases. Researchers believe that in such developed countries as Japan rice became an inferior good several decades ago.

However, there is conflicting evidence regarding whether rice is an inferior good. One of the most influential studies on rice consumption in Asia is an empirical study conducted by Ito, Peterson and Grant (1989). Utilizing aggregate national-level data, the authors concluded that rice was an inferior good in high-income Asian countries. They estimated the income elasticity of rice in Japan to be -0.091 in 1964 and -0.708 in 1984. Kako, Gemma and Ito (1997) projected Japanese rice demand by applying a log-linear function estimated by OLS using time series data for the period 1970 to 1991. These authors supported their colleagues' results; they found evidence that rice was an inferior good and that meat products were substitutes for rice. Estimated own-price elasticity was -0.130, and expenditure elasticity was -0.308. In a recent study, Price and Gislason (2001) investigated the habit formation of Japanese consumption, utilizing time series survey data from 1963 to 1991. The authors found that the expenditure elasticity of cereal was -0.01 in the short term and -0.015 in the long term. This result indicates that cereal, including all kinds of rice and noodles, is indeed an inferior good.

TABLE 4.1
Industry output

Source: Management and Coordination Agency, 1999.

Sawada (1980) studied the Japanese food demand system using the Rotterdam model. Monthly survey data covered the period from 1956 to 1975. Food item categories included cereal, fish, meat, eggs, vegetables, fruits, FAFH, other food and non-food items. The expenditure elasticity for rice (-0.395) was negative, and rice was the only item to show a negative expenditure elasticity. The own-price elasticities for rice, fish, meat and FAFH were -0.17, -0.545, -0.799 and -0.353, respectively. Utilizing monthly time series survey data from 1964 to 1979, Sawada (1983) estimated Japanese price and expenditure elasticities with the AIDS model. Results indicated that the own-price elasticities for staple food, fish, meat and FAFH were -0.903, -1.125, -0.981 and -1.522, respectively. The own-price elasticity for staple food was particularly high, and was similar to that of meat. In another study, Sawada (1984) used a two-level food demand system to estimate own-price and cross-price elasticities for 17 products in Japan. Data were taken from the FIES and were monthly time series from 1963 to 1981. The own-price elasticity for rice was -0.26, and its expenditure elasticity was -0.73. It is noteworthy that, among the 17 products, rice was the only one to show a negative expenditure elasticity. The results indicated that the magnitude of price elasticity was similar to that of expenditure elasticity in absolute terms, i.e. if a good is price-elastic, it tends to be expenditure-elastic as well. It is interesting to note that the expenditure elasticity for beef (1.93) was the highest among the 17 goods, and the expenditure elasticity for FAFH was the second highest (1.82). Sawada (1986) later estimated the income elasticities of cereal and related goods using survey data from 1976 to 1984 by means of the generalized Houthakker's method. The author concluded that demand was largely influenced by income, the number of household members and the age of the household head. Price elasticities for food were small compared with income elasticities, and this tendency was particularly strong with respect to the number of household members.

Bouis (1991) objected to the Ito, Peterson and Grant study (1989) and claimed that time series estimates of grain consumption have a downward bias owing to the urban-rural migration pattern and the decreasing importance of rice production. From their estimated calorie-income elasticities, Bouis and Haddad (1992) and Bouis (1994) claimed that cross-sectional data estimates of income elasticity are upwardly biased owing to leakage from actual consumption, such as meals for guests and animal feeding in developing countries. As Chern (2000) and Huang and Bouis (1986) pointed out, plotting aggregate consumption against per capita income simply showed the correlation between the two variables, and did not necessarily reveal the true consumption behaviour. Accurate incomeelasticitycan be obtained from cross-sectional data and, in this report, income elasticities among various income classes have been estimated.

Is rice a complement or a substitute for meat, fish and other food items?

Many time series studies show that people consume more meat and poultry as their per capita income increases. Japan is no exception: the consumption of meat and poultry has been increasing, while that of rice has decreased since the 1960s. At the same time, consumers have started to have more varied diets. Regarding the history of calorie intake in Japan, Morishima, Aita and Nakagawa (1993) provide a succinct historical view from the early 1920s. The authors compare the budget shares of 12 food items in 1970, 1980 and 1989. The budget share of grains decreased (from 16.9 percent in 1970 to 12.1 percent in 1989), while that of FAFH increased substantially (from 9.3 percent in 1970 to 15.6 percent in 1989). This observation indicates that consumption patterns are shifting from starchy food to non-starchy food and from consumption at home to FAFH.

Tokoyama and Egaitsu (1994) investigated trends in Japanese food consumption patterns in the period 1963 to 1991. They classified per capita consumption patterns in various ways to show the trends of goods. They concluded that saturation and stability characterize Japanese dietary patterns: the demand trends of foods have been stable for a while. Recently, a preference for higher-quality food has been replaced by diversification and health concerns in food consumption. In addition, as a result of the increased opportunity cost of labour for household work, the demand for convenient food has risen.

As these studies show, the variety and nature of Japanese dietary patterns have shifted substantially over the years, and it is important to understand the demand relationship of rice among other major food items. This report estimates demand relationships among rice, meats, poultry, fish and other products in order to investigate the substitution and complementary patterns.

Cross-sectional analysis

One of main objectives of this report is to analyse food consumption patterns, as well as to conduct an econometric analysis of the food demand structure in Japan. Cross-sectional household data from the 1997 FIES are used. The FIES provides household data on a monthly basis. Participants are asked to keep a Household Schedule, a Family Account Book and a Yearly Income Schedule. They join the survey for six months, and every month one-sixth of participating households are replaced by new ones.

Estimations are based on a total of 95 223 observations. The food items surveyed are non-glutinous rice, bread, noodles, fresh fish and shellfish, fresh meat, milk, eggs, fresh vegetables, fresh fruits, fats and oil, and FAFH. The present report is unusual in that the income elasticities of rice and other related foods are estimated with large degrees of freedom. Such a cross-sectional study of Japanese consumption patterns is virtually unique, and the authors have no knowledge of other instances in which survey data have been used to estimate a food demand system. The results produced for this report are therefore of great potential interest to demand analysts and policy-makers. In order to incorporate household-level microdata, the authors applied the following single equation models: the Working-Leser model estimated by OLS; Heckman's sample selection model; and the Tobit model. Most coefficients have correct signs and are statistically significant. For a complete demand system analysis, the linearly approximated almost ideal demand system (LA/AIDS) and the almost ideal demand system (AIDS) were applied. The concept of a flexible complete demand system yields consumption behaviour estimates that have many desirable properties: the aggregation, homogeneity and symmetry conditions can be tested, as was rarely the case for previous demand studies on this topic. The LA/AIDS poses a unit of measurement problem. In order to obtain more accurate estimations, LA/AIDS models with two price indices were compared: the Stone price index and the Laspeyres price index. In order to correct a censored dependent variable problem, a censored regression approach was used, as mentioned in Chapter 3.

Empirical results

To investigate the differences in demand structure among income groups, the original sample of 95 223 households was divided according to annual income levels (see Table 4.2).

TABLE 4.2
Classification of households by income level

Note: The exchange rate of US$1 = 126 Yen (as of 23 July 2001) is used.
Source: FIES, 1997.

When the Working-Leser model is estimated with the price data described in the previous paragraph, zero consumption households with a zero budget share are assumed to face the mean prices for the particular geographic location, month and income level concerned. In the censored regression, data are corrected by Heckman's two-step procedure. Econometric models and estimation models are described in Chapter 3. In addition to the usual expenditure and price variables, many demographic variables are also considered. Specifically, these include age of household head, household size, number of wage earners, and numbers of children aged five years or less, between six and twelve years, and between 13 and 18 years.

The estimates of expenditure and income elasticities from the whole sample working Leser model (OLS) are shown in Table 4.3. The results indicate that rice is not an inferior good according to this estimation. The expenditure elasticity of rice exceeds one. Other commodities are relatively expenditure-inelastic, with the exception of FAFH, which has the highest expenditure elasticity. It is noteworthy that the own-price elasticity for rice is very elastic. This indicates that Japanese consumers are sensitive to price changes in rice. If this estimate represents Japanese consumer behaviour correctly, rice imports - which should lead to a reduction in price - might benefit not only consumers but also rice farmers. The high expenditure and own-price elasticities for rice seem to be problematic and their credibility needs further validation.

Table 4.4 shows expenditure elasticities by income bracket. Most estimates are relatively invariant with income level. Fresh fish and meat show that lower-income consumer demand tends to be more expenditure-elastic, while higher-income consumers are less elastic. Table 4.5 shows own-price elasticities by income bracket. There are no significant variations of elasticity estimates by income level. The own-price and expenditure elasticities for rice remain very high in these income submodels.

The parameters of the LA/AIDS and AIDS models with demographic and seasonal dummy variables are estimated by dropping the equation for FAFH. Homogeneity and symmetry conditions are imposed on the estimation. The iterative seemingly unrelated regression procedure (ITSUR) is applied in SAS for estimation. ITSUR runs less than 15 iterations to meet the convergence criteria of 0.0001 for all models.

Tables 4.6 and 4.7 show the elasticity estimates from the AIDS model with the inverse Mills ratios of rice, fats and oil, and FAFH, for which substantial numbers of households had zero consumption. Table 4.6 shows the results of uncompensated price elasticities and expenditure elasticity. The expenditure elasticity of rice exceeds one, which indicates that rice is a normal good. Rice is mildly complementary to all commodities except for FAFH. The results clearly indicate the substitutability between rice and FAFH. In fact, most estimated cross-price elasticities are numerically small, except for those associated with the price of FAFH. The results suggest a strong substitutability between FAFH and at-home consumption of bread, noodles, fresh meat, eggs, and fats and oil. In Table 4.7, the compensated price elasticities show mixed results. Rice is a substitute for fresh fish, while it is a complement for fresh meat consumed at home. In addition, a moderate substitution (with a cross-price elasticity larger than 0.1) appears to occur between fresh fish and fresh meat, fresh fish and vegetables, and fats and oil and vegetables. A moderate complementarity is estimated between rice and fats and oil.

In order to validate the robustness of these estimation results, the estimation results from OLS, Heckman's two-step, Tobit, the LA/AIDS and the AIDS models are compared. Table 4.8 compares the own-price elasticity estimates from all models. It is surprising that the uncompensated own-price elasticity for rice exceeds 1.7 in absolute term in the LA/AIDS and AIDS models. Furthermore, the high own-price elasticity of rice is robust across all models. The lowest estimate of own-price elasticity for rice is -1.2 in the conditional estimates of Heckman's two-step and Tobit estimators.

TABLE 4.3
Whole sample elasticities for 11 major food products (OLS)

Notes: The numbers in parentheses following the elasticity estimates are standard errors. All estimates are statistically significant at the 5 percent level.

Fresh fish has a higher own-price elasticity than fresh meat in all models. There are some possible reasons behind this finding. First of these is the importance of fish in the Japanese diet. Japan's fish eating culture has induced a wide range of variations in fish prices and consumption levels. In addition, there are many varieties and uses of fish. These varieties offer many substitution opportunities, which tend to result in higher own-price elasticity.

The low elasticities for milk and eggs are expected: there is little variation in price for these products. Although dairy products are becoming more popular, there are still only few varieties of them.

Table 4.9 presents a comparison of expenditure elasticity estimates across all models. The results are robust across the models. The largest expenditure elasticity for rice is 1.472 in the LA/AIDS model with the Stone price index, while the smallest is the 1.065 estimated from the AIDS model. This result is somewhat surprising; if rice is a staple food, an expenditure elasticity of 1.0 to 1.4 is fairly high. In every model estimated, rice and FAFH are the only two food items to have expenditure elasticities exceeding one. With respect to rice, this result is contrary to expectations. Combining this finding with the results from Table 4.8, there is a possibility that rice is no longer a staple food.

It is interesting to note that fresh fish has a higher expenditure elasticity than fresh meat, and this pattern is robust across the estimated models. Again, this result shows the importance of fish in the Japanese diet. Recently, meat consumption has been increasing, while the high price and expenditure elasticities show that consumers are sensitive to fish prices and their own incomes when they determine consumption patterns. Another possibility is that processed meats, such as ham, sausage and bacon, are not included in the fresh meat category. The inclusion of processed meats might change the result.

TABLE 4.4 - Expenditure elasticities by income bracket (OLS)

Notes: The numbers in parentheses underneath the elasticity estimates are standard errors. All estimates are statistically significant at the 5 percent level.

The expenditure elasticity for fresh fruits is relatively high. This might be a reflection of the fact that fruits are relatively expensive among the goods in the survey. In the marketplace, there are always more expensive fruits than the ones that a consumer purchases. Furthermore, some fruits might be purchased as gifts. Because of this, a high expenditure elasticity might not reflect the true behaviour of at-home consumers.

Table 4.10 shows the income elasticities for all models. All expenditure elasticity estimates are linearly transformed to the income elasticity using Equation 3.6 from Chapter 3. In all estimations, the results show that Japanese rice is a normal good, and its income elasticity is in the range of 0.30 to 0.42. The results also show that only FAFH has a higher income elasticity than rice, with estimates ranging from 0.418 to 0.530. This result is robust across models. The estimated income elasticities indicate that none of the food items are luxury goods in Japan. When rice is expected to be a staple food, it may seem surprising that its income elasticity is greater than those of fish, meat, fruits or vegetables. However, the robustness of the econometric results, rather than intuition, should be used for reaching conclusions on the rice demand pattern in Japan.

TABLE 4.5 - Own-price elasticities by income bracket (OLS)

Notes: The numbers in parentheses underneath the elasticity estimates are standard errors. All estimates are statistically significant at the 5 percent level.

Table 4.11 reports the coefficients for demographic variables from the AIDS model. For rice, households with elderly heads or more family members tend to have higher budget shares for rice. On the other hand, households with more wage earners or children tend to have lower budget shares for rice. Households with young heads tend to have higher budget shares for bread, noodles and meat than households with older heads. Households with children of any age tend to consume more bread, noodles and milk, but less fish, vegetables and fruits than those without any children. Table 4.12 shows the coefficients for regional dummy variables. The Okinawa dummy is omitted from the model. It can be noted that all regional dummy variables for rice have a negative coefficient, implying that the budget share of rice is smaller in all other regions than in Okinawa. Tyugoku and Tohoku regions have the smallest budget shares for rice.

TABLE 4.6 - Uncompensated price and expenditure elasticities: AIDS model with inverse Mills ratio

TABLE 4.7 - Compensated price elasticities: AIDS model with inverse Mills ratio

TABLE 4.8 - Comparison of own-price elasticities

TABLE 4.8 (continued)

Notes: The numbers in parentheses underneath the elasticity estimates are standard errors. All estimates are statistically significant at the 5 percent level.

TABLE 4.9 - Comparison of expenditure elasticities

Notes: The numbers in parentheses underneath the elasticity estimates are standard errors. All estimates are statistically significant at the 5 percent level.

TABLE 4.10 - Comparison of income elasticities

TABLE 4.11 - Coefficients for demographic variables from AIDS model

Notes: The numbers in parentheses underneath the coefficient estimates are standard errors. All estimates are statistically significant at the 5 percent level.

TABLE 4.12 - Coefficients for regional dummy variables from AIDS model

Notes: The numbers in parentheses underneath the coefficient estimates are standard errors. All estimates are statistically significant at the 5 percent level. Okinawa is an omitted variable.

Concluding remarks

This report estimates the Japanese demand for 11 major products: non-glutinous rice, bread, noodles, fresh fish and shellfish, fresh meat, milk, eggs, fresh vegetables, fresh fruits, fats and oil, and FAFH. The own-price, cross-price and expenditure elasticities of the 11 products are estimated, and some selected demographic variables are analysed. Contrary to previous studies on rice consumption patterns, the empirical results reported here show that own-price and expenditure elasticities for rice are high in absolute terms. In addition, these high elasticities are robust across different estimators. These results imply that Japanese rice is no longer a staple food, and rice has become more of a luxury food than many other foods. The econometric results also show that rice and FAFH are strong substitutes. The estimation results for other goods confirm that FAFH is a substitute for bread, noodles, fresh meat, eggs and fats and oil. These results are not surprising because this report utilizes purchase data, but they do imply the importance of FAFH in Japanese dietary patterns. Results from demographic variables show that Japanese food consumption patterns are different across household age groups: households with older heads tend to have more traditional dietary patterns, including a higher budget share for rice. The general dietary patterns of younger households can be characterized as Westernized: these households tend to consume more bread, noodles and meat, which are not traditional Japanese food items.

Using 1997 household survey data, econometric results indicate that traditional Japanese dietary patterns have changed. Rice is no longer a staple food, and FAFH plays an important role in food consumption. There is no evidence that rice is an inferior good. It may even be appropriate to change a priori expectations for grain consumption in high-income countries.