• unlimited access with print and download
    $ 37 00
  • read full document, no print or download, expires after 72 hours
    $ 4 99
More info
Unlimited access including download and printing, plus availability for reading and annotating in your in your Udini library.
  • Access to this article in your Udini library for 72 hours from purchase.
  • The article will not be available for download or print.
  • Upgrade to the full version of this document at a reduced price.
  • Your trial access payment is credited when purchasing the full version.
Buy
Continue searching

Three essays on demand for organic milk in the U.S., environment and economic growth in Japan, and life expectancy at birth and socio-economic factors in Japan

Dissertation
Author: Mitsuko Chikasada
Abstract:
This dissertation consists of three separate empirical essays on: 1) Censored demand system estimation and analysis for U.S. organic milk; 2) A multilevel modeling approach to examine the environmental Kuznets curve hypothesis using SO2 and NOx concentration data in Japan; and 3) A study on life expectancy at birth in Japan between 1955 and 2005 using a dynamic panel data approach. The abstract of each essay is described below. [Essay One] Organic dairy sales have been growing rapidly in the past decade. Within the organic dairy category, organic milk has had the largest share of sales. Against the backdrop of such a rapid growth, the first essay empirically examines consumer behavior toward organic milk in 2004 and 2005 by using consumer purchase data collected by ACNielsen. Two empirical problems caused by the data (missing price information for non-purchasing households, and extreme censoring) are overcome in this study. As for missing price data, I systematically match consumers by the stores at which they shop, and transfer price data from a purchasing household to a non-purchasing household. The second empirical problem (censoring) is addressed by using a censored demand system. The zero expenditure shares are accounted for by estimating a translog demand system with a Quasi-Maximum Likelihood method. The compensated own-price elasticities show that organic milk is more sensitive than non-organic milk to own-price changes. Cross-price elasticities indicate that organic milk purchases increase when non-organic milk price rises. Yet, the reverse is not true. With respect to the impacts of demographic factors, household income does not significantly affect organic milk purchases. Actually, lower income households are estimated to have larger expenditure shares on organic milk than higher income households. Households with a female head and non-white households have less expenditure shares on organic milk compared to their counterparts. [Essay Two] The hypothesis of an environmental Kuznets curve (EKC) has been studied and examined theoretically and empirically by many researchers since the 1990s. Fewer studies have looked at an EKC relationship within a country, and to my knowledge, no studies have looked at the impacts of different political units on environmental pressures in one country. Therefore, the purpose of the second essay is to investigate the empirical relationship between environmental pressure and income in Japan by looking at municipal level SO 2 and NOx concentration data and income data at both the prefecture and national levels using a multilevel modeling approach. By incorporating municipal and prefecturespecific effects into the model (the random intercept model), I find an inverted U-shaped (EKC) relationship at the prefecture level only for SO 2 concentrations. However, after allowing the slope of prefectural income to vary among prefectures (the random coefficient model), I find no evidence for an EKC. Similarly, with respect to NOx , I find an EKC at prefecture level when I adopt the random intercept model for both the whole dataset and non-roadside station data. Yet the random coefficient models do not show an EKC relationship for both datasets. The roadside station data do not show significant associations between prefecture income per capita and NO x concentrations in any models. As for the relationship between national GDP per capita and NOx concentration, the roadside station data show upward sloping curves while the non-roadside station data show downward sloping curves. [Essay Three] Japanese life expectancy at birth has increased dramatically, particularly since the end of the Second World War, and is now one of the highest levels in the world. Backed by such a rapid growth, the third essay empirically examines the major determinants of Japanese longevity using prefecture-level data at five-year intervals on life expectancy at birth for males and females, and health and socio-economic factors between 1955 and 2005. This study uses a dynamic panel data model, which allows us to take into account the dynamic nature of adjustment and also to control for serial correlation, which can otherwise cause inconsistency in the estimates. More specifically, I use both the Arellano and Bond estimator and the system GMM estimator. As far as I am aware, this paper is the first to employ a dynamic panel data model in this field. The results show that income is positively related to life expectancy at birth for males and females from 1955 to 1980, but the association between these two variables disappears for the period from 1980 to 2005. The percentage of the population living in rural areas adversely affects life expectancy for males in the period between 1955 and 1980, while it positively affects life expectancy for females in the latter period (1980 to 2005). College education is positively related to life expectancy for both males and females only during the higher life expectancy period (1980 to 2005), but has no effect during the lower life expectancy period (1955 to 1980).

vii TABLE OF CONTENTS

List of Figures……………………………………………………………………………....ix List of Tables……………………………………………………………………………...…x Acknowledgements………………………………………………………………………....xi

1. CENSORED DEMAND SYSTEM ESTIMATION AND ANALYSIS FOR U.S. ORGANIC MILK…………………………………………………................................ 1 1.1 Introduction…………………………………………………………………….1 1.2 Background…………………………………………………………………….4 1.3 Literature Review………………………………………………………………6 Past Studies on Consumer Behavior toward Organic Products…………..6 Past Studies on Consumer Demand for Milk in General……………......12 Past Studies on Estimation of Censored Demand Systems………….......14 1.4 Specific Objectives……………………………………………………………16 1.5 Data and Methodology……………………………………………………......17 Data……………………………………………………………………...17 Model……………………………………………………………………19 Estimation Method………………………………………………………21 1.6 Results and Discussion………………………………………………………. 25 1.7 Summary and Conclusions……………………………………………………44 1.8 References…………………………………………………………………….50

2. A MULTILEVEL MODELING APPROACH TO EXAMINE THE ENVIRONMENTAL KUZNETS CURVE HYPOTHESIS USING SO 2 AND NO x

CONCENTRATION DATA IN JAPAN…………………………………………….… 56 2.1 Introduction…………………………………………………………………56 2.2 Literature Review………………………………………………………………59 Literature Review on Why We Should Expect to Find an EKC….59 Literature Review on How an EKC has been empirically examined….63 2.3 Data and Methodology……………………………………………………......73 Data and Variables………………………………………………………73 Models…………………………………………………………………78 2.4 Results and Discussion………………………………………………………. 84 2.5 Conclusions………………………………………………………………………107 2.6 References…………………………………………………………………….109

3. A STUDY ON LIFE EXPECTANCY AT BIRTH IN JAPAN BETWEEN 1955 AND 2005 USING A DYNAMIC PANEL DATA APPROACH……………………………116

viii 3.1 Introduction………………….....………………………………………………116 3.2 Literature Review………………………………………………………………118 Literature Review on Determinants of Longevity…………………….118 Background and Literature Review on Japanese Longevity………….121 3.3 Data and Methodology……………………………………………………......134 Data and Variables…………………………………………………….134 Econometric Models…………………………………………………...137 3.4 Results and Discussion………………………………………………………. …..146 3.5 Conclusions………………………………………………………………………158 3.6 References…………………………………………………………………….161

ix LIST OF FIGURES

Figure 2.1 Multilevel structure of the data…………………………………………………81 Figure 2.2 SO 2 concentration and GDP per capita (Model 1)……………………………92 Figure 2.3 SO 2 concentration and GDP per capita (Model 2)……………………………92 Figure 2.4 SO 2 concentration and prefecture income per capita (Model 2)………………..93 Figure 2.5 SO 2 concentration and GDP per capita (Model 3)……………………………93 Figure 2.6 SO 2 concentration and prefecture income per capita (Model 3)………………..94 Figure 2.7 NO x concentration and GDP per capita (Model 1)……………………………..94 Figure 2.8 NO x concentration and GDP per capita (Model 2)……………………………..95 Figure 2.9 NO x concentration and prefecture income per capita (Model 2)……………….95 Figure 2.10 NO x concentration and GDP per capita (Model 3)………………………….96 Figure 2.11 NO x concentration and prefecture income per capita (Model 3)……………96 Figure 2.12 NO x concentration and GDP per capita: Roadside Monitoring Station (Model 1).................................................................................................102 Figure 2.13 NO x concentration and GDP per capita: Roadside Monitoring Station (Model 2)………………………………………………………………102 Figure 2.14 NO x concentration and prefecture income per capita: Roadside Monitoring Station (Model 2)………………………………………………………103 Figure 2.15 NO x concentration and GDP per capita: Roadside Monitoring Station (Model 3)………………………………………………………………103 Figure 2.16 NO x concentration and prefecture income per capita: Roadside Monitoring Station (Model 3)………………………………………………………104 Figure 2.17 NO x concentration and GDP per capita: Ambient Air Pollution Monitoring Station (Model 1)……………………………………………………….104 Figure 2.18 NO x concentration and GDP per capita: Ambient Air Pollution Monitoring Station (Model 2)………………………………………………………105 Figure 2.19 NO x concentration and prefecture income per capita: Ambient Air Pollution Monitoring Station (Model 2)……………………………………….105 Figure 2.20 NO x concentration and GDP per capita: Ambient Air Pollution Monitoring Station (Model 3)………………………………………………………106 Figure 2.21 NO x concentration and prefecture income per capita: Ambient Air Pollution Monitoring Station (Model 3)…………………………………………106

Figure 3.1 Life Expectancy at Birth (Male and Female), Japan……………………….…124

x LIST OF TABLES

Table 1.1 Definition of Variables……………………………………………………..........30 Table 1.2 Summary Statistics of 2004 Data………………………………………………..30 Table 1.3 Summary Statistics of 2005 Data………………………………………………..31 Table 1.4 Unconditional Uncompensated Price Elasticities (2004)………………………..36 Table 1.5 Unconditional Total Expenditure Elasticities (2004)…………………………....36 Table 1.6 Unconditional Compensated Price Elasticities (2004)…………………………..36 Table 1.7 Unconditional Demographic Elasticities (2004)…………………………….......36 Table 1.8 Unconditional Uncompensated Price Elasticities (2005)………………………..37 Table 1.9 Unconditional Total Expenditure Elasticities (2005)…………………………....37 Table 1.10 Unconditional Compensated Price Elasticities (2004)………………………....37 Table 1.11 Unconditional Demographic Elasticities (2004)…………………………….....37 Table 1.12 Unconditional Expectation of Expenditure Shares…………………………….39 Table 1.13 Unconditional Expectation of Expenditure Shares on Organic Milk…………..42 Table 1.14 Unconditional Expectation of Expenditure Shares on Organic Milk…………..42 Table 1.15 Unconditional Expectation of Expenditure Shares on Organic Milk…………..43

Table 2.1 The Number of Municipalities Included in the Model…………………………85 Table 2.2 Summary Statistics for SO 2 Models and for NOx Models………………………86 Table 2.3 Estimation Results for SO 2 Mean Annual Concentrations………………………90 Table 2.4 Estimation Results for NO x Mean Annual Concentrations……………………91 Table 2.5 Summary Statistics for SO 2 Models and for NO x Models: Roadside Station and the Other Station……………………………………..……………………..98 Table 2.6 Estimation Results for NO x Mean Annual Concentrations: Roadside Monitoring Station Data…………………………………………………………………100 Table 2.7 Estimation Results for NO x Mean Annual Concentrations: Ambient Air Pollution Monitoring Station Data………………………………………………….…101

Table 3.1 OECD Country Ranking of Life Expectancy at Birth (years), 2001…..……….122 Table 3.2 Summary Statistics of the Variables……………………………………………147 Table 3.3 Dynamic Panel Data Estimation Results for Life Expectancy at Birth for Males in Japan…………………………………………...……………………………154 Table 3.4 Dynamic Panel Data Estimation Results for Life Expectancy at Birth for Females in Japan……………………………………………………………………...155 Table 3.5 Table 3.3 Dynamic Panel Data Estimation Results for Life Expectancy at Birth for Males between 1955 and 2005………………………………………….156 Table 3.6 Table 3.3 Dynamic Panel Data Estimation Results for Life Expectancy at Birth for Females between 1955 and 2005……………………….……………….157

xi ACKNOWLEGEMENTS

I would like to express my deepest gratitude to my academic advisors, Dr. Edward C. Jaenicke and Dr. David G. Abler for their continuous encouragement, valuable advice, and enduring support. I would also like to thank my dissertation committee members, Dr. James S. Shortle and Dr. Robert Schoen for their constructive guidance and encouragement. I would like to thank the U.S. Department of Agriculture’s Economic Research Service for partially funding my research on demand for organic milk in the U.S. (Essay One). I particularly thank Dr. Carolyn Dimitri for her support. I am also grateful to Dr. Shannon C. Stokes and Dr. Setsuko Ohta for their valuable advice for my third essay on life expectancy at birth and socio-economic factors in Japan. I am thankful to my friends I met in State College, including Yuen Leng Chow, Atsuko Nonoguchi, Yuki Yano, Jeong Hwan Bae, Sundar Shreslta, Xiaohua Yu, Bangya Ma, Hwansoo Sung, Spence Ford, Carmen Cooper, Etsuko Tussey and Pat Morrissey. Without their encouragement, I would have never come to this stage. I deeply thank my old best friends, Haruko Migita, Namiko Satake, Miyako Iida, Kazuko Fukushima, and Miki Sakai for their support, which gave me energy to go through every difficulty. I wish to express my deep appreciation to Dr. Daisaku Ikeda, my mentor in life and founder of the Soka schools. I would not have dreamed of studying abroad without his

xii encouragement. My appreciation also goes to Ms. Sachiko Wakai, who has been encouraging me since I learned from her in elementary school. Last, I would like to extend my great appreciation to my parents, Junichi and Eiko Chikasada, and my brother, Koji Chikasada. They have kept encouraging and helping me. I am also grateful to my deceased grandparents, Osamu and Tamiyo Maruyama, and Shuso and Fukue Chikasada. They have prayed for my health, success, and happiness.

1 1. CENSORED DEMAND SYSTEM ESTIMATION AND ANALYSIS FOR U.S. ORGANIC MILK 1.1 INTRODUCTION During the past decade, consumer demand for organic products in the U.S. has been growing drastically, at rates of about 15 to 20 percent annually according to Nutrition Business Journal. Compared to other broad categories of organic foods, organic dairy has consistently been among the top five in sales, and had the highest annual growth rate in 2004 and 2005. Within the broad category of organic dairy, organic milk gained the largest share of sales. With this rapid growth as a backdrop, this study’s overall objective is to empirically investigate consumer behavior toward organic milk in 2004 and 2005. The study period comes two years after the U.S. Department of Agriculture (USDA) implemented national organic standards on October 21, 2002. Before that date, there had been no nationally unified standards for organically produced food. It is likely that the USDA organic labeling requirements that accompanied the national organic standards caused food manufacturers, retailers and consumer to modify their marketing and consumption behaviors. Rather than focus on the direct impact of this change in policy, this study investigates consumer behavior after firms and consumers grew more accustomed to the new regulations. Therefore, the study focuses on consumer behavior in

2 2004 and 2005, a period sufficiently after the change in policy to avoid transitory behaviors. This research uses consumer purchase data collected at the household level for as many as 40,000 households by ACNielsen for its “HomeScan” panel. After returning home from a supermarket or some other food retailer, HomeScan panel participants are commissioned to re-scan their purchased products using a home scanning device that transmits data back to ACNielsen. Coupled with demographic data on each participating household, these data provide a rare opportunity to characterize consumer demand for organic products. More specifically, because they represent a generally complete set of purchases by each participating household, these data can support the estimation of a consumer demand system that is consistent with microeconomic theory. The data cause two empirical problems, however, when it comes to investigating demand for organic products: (i) extreme censoring and (ii) missing data from non-purchasing households. Extreme censoring occurs because approximately 90 percent of households in the data do not purchase organic milk. These households, therefore, will have zeros for their quantities of organic milk purchases. In this study, the censoring issue is taken into consideration in order to obtain unbiased and consistent estimates by using the method introduced by Yen et al. (2003), where a censored translog demand system is

3 estimated by a Quasi-Maximum Likelihood (QML) method. The high percentage of households not purchasing organic milk also causes the second major empirical obstacle: Because the HomeScan data describe only those products that are actually purchased, there are many households with missing price data for organic milk. Many previous studies (e.g., Cox and Wohlgenant 1986; Dong and Kaiser 2005; Yen et al. 2003) overcome missing price data by extrapolating or estimating prices based on geographic market areas. An alternative and potentially more accurate method involves transferring price data from a purchasing household matched to a non-purchasing household that shopped at the same store. For example, if households A and B purchase any dairy product at store X, and household A buys organic milk but household B does not, then one can add the organic milk price data from household A to data for household B by presuming household B sees the same prices as household A. This sort of matching algorithm was used by Keane (1997). Another empirical issue involves the size and aggregation level of the estimated demand system. ACNielsen data aggregates dairy purchases into three broad categories: milk, yogurt/butter, and cheese. It would be preferable, therefore, to estimate a six good demand system – one non-organic and one organic good for each specific product group aggregated by ACNielsen. However, this grouping exacerbates the extreme

4 censoring problem discussed above and, so far, makes estimation impossible. However a four good system (non-organic milk, non-organic yogurt/butter, non-organic cheese, and organic milk) can be estimated successful. Five demographic variables (Household income, household size, college education, female head, and non-white) can also be successfully included in the system. Therefore, my study represents a successful preliminary effort to include organic product categories in a demand system. It is perhaps the first successful attempt to estimate a censored demand system with QML method in the context of organic dairy products. In the following sections, I first present a brief introduction and background information on the market for organic products. Second, I review three specific topics in the economics literature: consumer behavior toward organic products; consumer demand for milk in general, and estimation of censored demand systems. Subsequently, I outline specific research objectives, describe the HomeScan data in more detail, and explain how a censored demand system is estimated. Finally I discuss the result, and conclude in the last section.

1.2 BACKGROUND According to Nutrition Business Journal (NBJ) data (2006), consumer demand for

5 organic products in the U.S. has been growing rapidly during the past decade, at an annual rate of approximately 15 to 20 percent. More narrowly, the category of organic dairy products ranked second in organic food sales for 2004 and 2005, behind the fruits and vegetables, and ahead of nondairy beverages, packaged or prepared foods, and breads and grains (NBJ 2006). Among these categories, organic dairy grew most rapidly in 2004 and 2005, when sales increased by approximately 25.1 percent in 2004 and 23.5 percent in 2005 (NBJ 2006). Organic milk was first sold in conventional supermarkets in 1993, and by the end of 1996, eight conventional supermarkets sold it (Glaser and Thompson 2000; Dimitri and Greene 2002). Milk is considered to be one of the first organic goods that consumers purchase (Kiesel and Villas-Boas 2007). During this decade, organic milk sales have the largest share in the organic dairy category. In addition to the largest share, organic milk sales have been growing rapidly: The annual growth rate of organic milk sales was 27.3 percent in 2004 and 24.6 percent in 2005 (NBJ 2006). Moreover, organic milk has had the second largest sales of all organic food subcategories for those two years, behind the fresh produce category (NBJ 2006.). The USDA implemented national organic standards on October 21, 2002. Prior to that date, private organizations and some states established their own certification standards,

6 some as early as the 1970s (Dimitri et al. 2002; Oberholtzer et al. 2005). However, there had been no nationally unified standards for organically produced food before the national organic standards were established. Among other requirements, the national standards list approved production and handling methods, labeling procedures, and recordkeeping and auditing procedures for both producers and certifiers. Under the standards, all the raw, fresh and processed products can be labeled as “organic” only if they are produced and handled by USDA’s regulations. Three kinds of labels are developed and defined by USDA (National Organic Program, USDA): (i) 100 percent organic, (ii) organic, and (iii) made with organic ingredients (Dimitri et al. 2002). Kiesel and Villas-Boas (2007) found that the probability of buying organic milk was increased by the USDA organic seal.

1.3 LITERATURE REVIEW Past Studies on Consumer Behavior toward Organic Products Consumer behavior toward organic foods in the U.S. has been examined in both industry and academic fields. Industry studies have mainly used consumer surveys, while academics have used various methodological approaches to examine the characteristics affecting consumers’ purchases (Dimitri and Greene 2002; Oberholtzer et al. 2005).

7 Industry studies have examined the motivation for consumers’ purchasing organic foods. In the 1980s and 1990s, environmental concern was the main motivation, but recently their motivations have diversified (Dimitri and Greene 2002; Oberholtzer et al. 2005; Dimitri and Lohr 2007). Environmental concern is still one of the biggest reasons for consumers to purchase organic foods (Hartman Group 2000; Whole Foods 2004), and other motivations have been found by various surveys, including health and nutrition (The Food Marketing Institute 2001; Hartman Group 2000; Whole Foods 2004), taste or food quality (Hartman Group 2000; Whole Foods 2004), availability (Hartman Group 2000), and support for small and local farmers (Whole Foods 2004). The Fresh Trends survey (1996, 1998, 2000, 2002) suggested that purchasing behaviors hardly differed between men and women. The Fresh Trends survey (2001) also found that for 12 percent of respondents, whether a product is organic is a prime reason for their purchasing decision. With respect to the reasons for not purchasing organic foods, price (Walnut Acres 2001, 2002; Whole Foods 2004; Hartman Group 2002) and availability (Hartman Group 2002) were the two biggest reasons. Recent research shows that organic consumers have become much more diverse, which makes it difficult to predict the organic purchases by people’s income and ethnicity (Dimitri and Lohr 2007; Howie 2004; Barry 2004). In other words, high income and white

8 people are still organic purchasers, but people in other categories, such as lower income and non-white, also account for a large share of organic purchasers (Dimitri and Lohr 2007; Howie 2004; Barry 2004). Academic studies applied various methodological approaches to examine organic consumers’ behavior (Dimitri and Greene 2002). Various researchers have investigated what kind of factors would affect consumers’ buying decision of organic produce. Such factors are summarized as follows: price, size and packaging, whether the product is on sale, (Estes and Smith 1996), household size and income (Loureiro and Hine 2002; Govindasamy and Italia 1999; Akgüngör et al. 2007), education level (Loureiro and Hine 2002; Thompson and Kidwell 1998; Akgüngör et al. 2007), age and gender (Govindasamy and Italia 1999), the presence of children in a household (Thompson and Kidwell 1998), appearance of fresh produce (Estes and Smith 1996; Thompson and Kidwell 1998), concern about environment (Loureiro and Hine 2002; Cicia et al. 2006), concern about food safety and health (Loureiro and Hine 2002; Govindasamy and Italia 1999; Cicia et al. 2006; Akgüngör et al. 2007), and familiarity with alternative agriculture (Govindasamy and Italia 1999). With respect to the effect of higher education on purchasing organic foods, Loureiro and Hine (2002) and Thompson and Kidwell (1998) found different results: The former mentioned that a person with higher level of education was willing to pay more for organic

9 produce, while the latter said that consumers with higher education were less likely to purchase organic products. Empirical studies estimating demand systems of organic produce have become possible since scanner data became available. Glaser and Thompson (1999) estimated demand systems for conventional and organic frozen vegetables using monthly U.S. national-level supermarket scanner data from ACNielsen between September 1990 and December 1996, and estimated elasticities by employing the nonlinear AIDS model. Glaser and Thompson (2000) examined demand for conventional and organic beverage milk by using ACNielsen supermarket scanner data from April 1988 to December 1996 and Information Resources, Inc. (IRI) data from January 1993 to December 1999 and from November 1996 to December 1999. Both data sets were scanner data collected at supermarket level. While they did not account for censored data, they did use the nonlinear AIDS model to estimate four separate demand systems for the following categories: whole milk, 2 % milk, 1 % milk, and nonfat/skim milk. Each system had three goods: conventional branded milk, conventional private label milk, and organic milk. They found that organic milk is the most price elastic of all the three goods, private-label milk is the most price inelastic, and branded milk is slightly more elastic than the private-label one. For the whole, 2 % and nonfat/skim milk systems, their results showed that organic and

10 branded milk are substitutes. The cross-price elasticities between organic and branded milk, and between organic and private-label milk, showed asymmetric relationships: the change in price of organic milk causes small changes in branded and private-label milk purchases, whereas the change in price of branded and private-label milk results in quite large changes in organic milk purchases. With respect to expenditure elasticities, branded and private-label milk showed reasonable results: most of them were close to one. However expenditure elasticity of organic milk turned out to be negative and quite large in absolute values. They explained that such results were likely to come from the shape of the expenditure elasticity equation in the AIDS model, i i it w e β +=1 . When β (parameter estimate) is negative and w (expenditure share of organic milk) is small, the expenditure elasticity for organic milk becomes quite large negative number. Dhar and Foltz (2005a) estimated a demand system of three kinds of milk products, such as organic, and recombinant bovine somatotropin (rBST) free, and unlabeled milk. In their study, a quadratic almost ideal demand system was estimated by controlling for price and expenditure endogeneity. According to their results, uncompensated and compensated own-price elasticities of organic milk are both -1.37, while those of unlabeled milk are -1.04 and -1.08, respectively. Asymmetric relationship between organic and unlabeled

11 cross-price elasticities are also found in their research. Thompson and Glaser (2001) estimated demand systems of conventional and organic baby food by using both ACNielsen and Information Resources, Inc. (IRI) data. They used the quadratic almost ideal demand system (QUAIDS) in this study. All of those four studies (Glaser and Thompson 1999, 2000; Dhar and Foltz 2005a; Thompson and Glaser 2001) showed high own-price elasticity of demand for organic items compared to conventional items. As for cross-price elasticities, some frozen vegetables (broccoli, green peas and green beans) exhibited statistically insignificant cross-price elasticities, which suggested that between conventional and organic products the change in prices of their counter products did not affect the quantity demanded significantly (Glaser and Thompson 1999). As for the other three studies on organic milk and organic baby food, the conventional and organic products turned out to be substitutes as expected (Glaser and Thompson 2000; Thompson and Glaser 2001; Dhar and Foltz 2005a). None of these studies accounted for the censoring in the data. Zhang et al. (2006) used ACNielsen HomeScan data in 2003 in order to examine consumers’ socio-economic characteristics which contribute to the growth of organic produce market. They employed a generalized double hurdle model, and suggested that higher income and higher educated consumers should be targeted in marketing strategies.

12 Lin et al. (2008) analyzed a censored demand system for organic and conventional fresh fruits, including 6 conventional and 6 organic fruit products. Using the ACNielsen 2006 HomeScan data, the translog demand system was estimated by a two-step procedure. Some of their findings were similar to those of Glaser and Thompson (2000): organic products showed much larger own-price elasticities than conventional products, and cross price elasticities showed asymmetric results, where the price change in conventional goods elicit a change in organic purchases, while the price change in organic goods are less likely to elicit a change in conventional purchases.

Past Studies on Consumer Demand for Milk in General Many researchers have analyzed milk demand using household level data. They found significant impacts of various kinds of demographic variables on milk products, such as household income (Cornick et al. 1994; Haines et al. 1988; Heien and Wessells 1988; Huang and Rauniker 1983; Popkin et al. 1989; Rauniker and Huang 1984; Blaylock and Smallwood 1993; Schmit et al. 2002), ethnicity (Cornick et al. 1994; Haines et al. 1988; Heien and Wessells 1988, 1990; Huang and Rauniker 1983; Rauniker and Huang 1984), food stamp program participation or value of food stamps (Heien and Wessells 1988, 1990), composition or age structure of household members (Heien and Wessells 1988, 1990;

13 Huang and Rauniker 1983; Schmit et al. 2002), region of residence (Cornick et al. 1994; Heien and Wessells 1988, 1990; Huang and Rauniker 1983; Popkin et al. 1989), seasonality (Heien and Wessells 1988, 1990), percentage of meals at home (Heien and Wessells 1988, 1990), household size (Cornick et al. 1994; Huang and Rauniker 1983; Popkin et al. 1989; Rauniker and Huang 1984; Reynolds 1991; Schmit et al. 2002), education (Cornick et al. 1994; Haines et al. 1988; Huang and Rauniker 1983; Popkin et al. 1989; Rauniker and Huang 1984; Reynolds 1991), presence or number of children (Cornick et al. 1994; Haines et al. 1988; Popkin et al. 1989), occupation (Heien and Wessells 1988, 1990). Some of them estimated a complete food demand system including milk as one category of goods (Heien and Wessells 1988, 1990), and some of them analyzed milk within a milk, dairy, or beverage demand system including disaggregated milk products that are usually categorized by fat content (Cornick et al. 1994; Haines et al. 1988; Huang and Rauniker 1983; Popkin et al. 1989; Gould 1996; Schmit et al. 2002; Dhar and Foltz 2005b; and many others). Some of them considered censoring in their models, by using various methods, such as tobit analysis, Heckman-type sample selection models (Heckman 1979). Among them, Gould (1996) used the censored demand system approach developed by Lee and Pitt (1986), which accounted for the relationships between each goods. He considered a milk

14 demand system with three disaggregated milk categories, such as whole, skim/1%, and 2% milk. According to his research, all the own-price elasticities were negative and greater then -1 (inelastic), and all the types of milk were substitutes.

Full document contains 183 pages
Abstract: This dissertation consists of three separate empirical essays on: 1) Censored demand system estimation and analysis for U.S. organic milk; 2) A multilevel modeling approach to examine the environmental Kuznets curve hypothesis using SO2 and NOx concentration data in Japan; and 3) A study on life expectancy at birth in Japan between 1955 and 2005 using a dynamic panel data approach. The abstract of each essay is described below. [Essay One] Organic dairy sales have been growing rapidly in the past decade. Within the organic dairy category, organic milk has had the largest share of sales. Against the backdrop of such a rapid growth, the first essay empirically examines consumer behavior toward organic milk in 2004 and 2005 by using consumer purchase data collected by ACNielsen. Two empirical problems caused by the data (missing price information for non-purchasing households, and extreme censoring) are overcome in this study. As for missing price data, I systematically match consumers by the stores at which they shop, and transfer price data from a purchasing household to a non-purchasing household. The second empirical problem (censoring) is addressed by using a censored demand system. The zero expenditure shares are accounted for by estimating a translog demand system with a Quasi-Maximum Likelihood method. The compensated own-price elasticities show that organic milk is more sensitive than non-organic milk to own-price changes. Cross-price elasticities indicate that organic milk purchases increase when non-organic milk price rises. Yet, the reverse is not true. With respect to the impacts of demographic factors, household income does not significantly affect organic milk purchases. Actually, lower income households are estimated to have larger expenditure shares on organic milk than higher income households. Households with a female head and non-white households have less expenditure shares on organic milk compared to their counterparts. [Essay Two] The hypothesis of an environmental Kuznets curve (EKC) has been studied and examined theoretically and empirically by many researchers since the 1990s. Fewer studies have looked at an EKC relationship within a country, and to my knowledge, no studies have looked at the impacts of different political units on environmental pressures in one country. Therefore, the purpose of the second essay is to investigate the empirical relationship between environmental pressure and income in Japan by looking at municipal level SO 2 and NOx concentration data and income data at both the prefecture and national levels using a multilevel modeling approach. By incorporating municipal and prefecturespecific effects into the model (the random intercept model), I find an inverted U-shaped (EKC) relationship at the prefecture level only for SO 2 concentrations. However, after allowing the slope of prefectural income to vary among prefectures (the random coefficient model), I find no evidence for an EKC. Similarly, with respect to NOx , I find an EKC at prefecture level when I adopt the random intercept model for both the whole dataset and non-roadside station data. Yet the random coefficient models do not show an EKC relationship for both datasets. The roadside station data do not show significant associations between prefecture income per capita and NO x concentrations in any models. As for the relationship between national GDP per capita and NOx concentration, the roadside station data show upward sloping curves while the non-roadside station data show downward sloping curves. [Essay Three] Japanese life expectancy at birth has increased dramatically, particularly since the end of the Second World War, and is now one of the highest levels in the world. Backed by such a rapid growth, the third essay empirically examines the major determinants of Japanese longevity using prefecture-level data at five-year intervals on life expectancy at birth for males and females, and health and socio-economic factors between 1955 and 2005. This study uses a dynamic panel data model, which allows us to take into account the dynamic nature of adjustment and also to control for serial correlation, which can otherwise cause inconsistency in the estimates. More specifically, I use both the Arellano and Bond estimator and the system GMM estimator. As far as I am aware, this paper is the first to employ a dynamic panel data model in this field. The results show that income is positively related to life expectancy at birth for males and females from 1955 to 1980, but the association between these two variables disappears for the period from 1980 to 2005. The percentage of the population living in rural areas adversely affects life expectancy for males in the period between 1955 and 1980, while it positively affects life expectancy for females in the latter period (1980 to 2005). College education is positively related to life expectancy for both males and females only during the higher life expectancy period (1980 to 2005), but has no effect during the lower life expectancy period (1955 to 1980).