apple baum scholarship paper analysis of socio-economic and

APPLE BAUM SCHOLARSHIP PAPER

Analysis of Socio-economic and Demographic Factors

Affecting Food Away From Home Consumption:

A Synopsis*

by

Rodolfo M. Nayga, Jr.Department of Agricultural Economics and Business

Massey University, New Zealand

Oral Capps, Jr.Department of Agricultural Economics, Texas A&M University

College Station, Texas

Introduction

U.S. consumers States spent nearly $550billion for both food at home (FAH) and foodaway from home (FAFH) in 1991, up 6.4 percentfrom 1989. This total includes spending at allretail outlets (e.g. food stores, restaurants), atservice establishments (e.g. meals at lodgingplaces, snacks at entertainment facilities), plusallowances for food served in institutions (e.g.schools, hospitals), in the travel industry (e.g.airlines), and for military spending. Expendituresfor FAH amounted to almost $300 billion, up 5.7percent from 1989. Spending for FAFH, on theother hand, came to roughly $250 billion, 7.2percent above the 1989 level (Food RetailingReview, 1991).

In recent years, U.S. consumers have eatenan increased number of meals outside the home.Very roughly, the change has been from aboutone meal in four to about one in three, an increaseof about 33 percent during the last 25 years(Manchester, 1990). As exhibited in Table 1, theshare of food spending for FAFH rose from 26.6percent in 1960 to 45.4 percent in 1990. In con-trast, the share of food spending allocated forFAH dropped from 73.4 percent in 1960 to 54.6percent in 1990.

Food expenditures continually take asmaller share of consumers’ disposable income.Food expenditures, as a percent of income,decreased from 16.3 percent in 1970 to 13.8percent in 1989. The percentage of disposableincome going to FAH has also declined from 10.8percent in 1970 to 7,6 percent in 1989. In con-

U

-Synopsisof a Ph.D. dissertation, Texas A&M University, December 1991. For full details, see Nayga, 1991.

Journal of Food Distribution Research February 93/page 69

—.

Table 1. Expenditures for Food at Home and Food Away from Home

Year Food at Home Food Away from Home

NominaP Perce& Nomin& Perce~.-----------—--------- -------- ------- ----------1960 54,121 73.4 19,607 26.6

1965 60,542 69.8 26,197 30.2

1970 77,527 66.2 39,583 33.8

1975 119,850 63.8 68,109 36.2

1980 185,638 60.6 120,530 39.4

1981 198,520 60.1 131,563 39.9

1982 206,184 59.4 140,722 40.6

1983 217,114 58.8 152,272 41.2

1984 228,447 58.3 163,093 41.7

1985 235,935 57,9 171,463 42.1

1986 244,897 57.0 184,957 43.0

1987 254,058 55.5 203,869 44.5

1988 266,163 54,8 219,625 45.2

1989 282,548 55.0 230,785 45.0

1990 306,600 54.6 254,500 45,4

“millions of dollarbAs a percentage of expenditures on all foods

Source: U.S. Department of Agriculture and Food RetailingReview, 1991Edition.

February 931page 70 Journal of Food Distribution Research

trast, the percentage going to FAFH has increasedfrom 5.5 percent in 1970 to 6.3 percent a decadelater.

The away from home market is composedof commercial foodservice establishments (i.e.restaurants, fast food places, cafeteria) and non-commercial outlets (i.e. school or military diningrooms, child care centers). Although noncommer-cial outlets serve more food to more people, theyaccount for only 30 percent of the total retailvalue of FAFH,

The majority of the FAFH sales from thepast came from conventional restaurants. How-ever, fast food eating establishments have morethan tripled in number since the early 1960s. Thefast food industry has started placing outlets inlocations not previously served such as schools,military bases, and international markets. Inaddition, menus are being enlarged to includeitems such as salad bars, soups, baked potatoes,burgers, and whole grain buns. Franchised res-taurants are now also facing growing competitionfrom within the industry, from supermarkets andother food stores that prepare take-out food andfrom hotels that offer dining room service andcatering.

Socbeconomic and Demographic Forces

The move toward eating out is prompted bychanges in consumer lifestyles. Some socio-eco-nomic and demographic factors that come intoplay are: a growing number of women, marriedand single, in the work force; increasing impor-tance of convenience in eating out; more familiesliving on two incomes; the impact of advertisingand promotion by large food service chains; andmore people in the age group of 25 to 44 who areinclined to eat out often (Putnam and Van Dress,1984). Only about seven percent of all house-holds now fit the old stereotype family of a work-ing husband, a wife who does not work forwages, and two children (Kinsey, 1990).

Americans are spending less of their lives infamily households. Married couples with childrenare declining as a share of all households (U.S.Statistical Abstracts, 1988 and 1989), Accordingto the 1990 Census, the percentage of U.S. house-

holds headed by married coupl~ declined from 60percent in 1980 to 55 percent in 1990. Thirtypercent of the households in 1990 were non-familyhouseholds or those in which no two individualsare related by blood, marriage, or adoption, upfrom 27 percent in 1980. In most cases, thesenon-family households consist of one person,often elderly and living alone (Noah, 1991).

The one-adult households are fastest grow-ing and are likely to exhibit non-conventional foodconsumption patterns (i.e. FAFH consumption).For instance, single and employed persons livingalone spend much more on FAFH than any othergroup (Manchester, 1990). The growth of one-adult households is fostered by later marriages,continued high divorce rates, decreased fertilityand increased longevity.

More women now are in the labor force,increasing the demands for convenience in homeprepared food, for home delivered food and forFAFH. In fact, over 70 percent of women age25-44 are in the labor force, and notably about 75percent of these working women work full time.Moreover, labor force participation rate of womenincreased from 52 percent in 1980 to 58 percentin 1990 (Waldrop and Exter, 1991).

In addition to these changes, per capitaincome in the United States has been rising. U.S.per capita personal income grew by an average6,6 percent nationally in 1989, according to theCommerce Department. In 1982 dollars, percapita income increased over 43 percent between1970 and 1989. More families now are livingwith two incomes. Families with two or moreearners increased from 56 percent in 1980 to 58percent in 1990 of family households. There isalso a growing gap between the poor and the rich.Uneducated and poor households (mostly non-whites and single women) with incomes under$15,000 per year now account for about a third ofthe households. These households have a budgetshare for food of about 50 percent and are mostlyparticipants in food assistance programs. In con-trast, households with annual incomes over$35,000 spend only less than eight percent of theirincome for food. For this group, price will beless important than food safety, quality, taste andexperience (Kinsey, 1990).


The U.S. population is more diverse thanever. As in the 1980s, the growing ethnic diver-sity of America will create new challenges forfood marketers specially in the away from homesector. In early 1980s, less than 20 percent ofAmericans belonged to a racial or ethnic group.By 1990, nearly 25 percent of the U.S. populationbelonged to a racial or ethnic minority. Specifi-cally, Americans of “other races” (e.g. those thatare not white or black) grew seven times fasterthan the overall population in the 1980s (AmericanDemographics, March 1991). The white andblack populations only changed six and 13 per-cent, respectively, between 1980 and 1990 (13usi-ness Week, June 17, 1991). By 2005 there will bean equal number of blacks and hispanics in theUnited States which will make up over one quar-ter of the population, These trends will mean thatthe overall make-up of consumer needs and pref-erences will continue to change.

Rationale of the Study

Food consumption relationships are tradi-tionally specified between socioeconomic factorsand quantity or expenditure measures. Althoughnumerous studies exist on FAFH quantity orexpenditure models (see Table 2), these studieshave used earlier data sets and hence, do notnecessarilyy reflect recent market conditions.Moreover, only the McCracken and Brandt studyhas investigated FAFH consumption by type offood facility. There is, therefore, a need todevelop FAFH quantity or expenditure models bytype of facility using recent data sets (i.e. the1987-88 National Food Consumption Survey).The development of these models would not onlylead to improved market planning but would alsoallow comparison of the results with previousstudies.

These approaches, however, would notallow inferences regarding the nutritional status ofan individual’s diet. Large quantities of foodconsumed or large expenditures on food eitheraway from home or at home may not necessarilymean adequacy in terms of nutrient consumption(Adrian and Daniel, 1976). To quote McCrackenand Brandt (1987), “knowledge is the key torapid and eftlcient adjustments in the food systemto changing consumer demands. Along with the

impact of these changing food consumption pat-terns on the food distribution system itself, thenutritional intake of consumers is likely to be alsoaffected. ” Also, the increasing complexity of thefood industry as well as the heightened consumerinterest in health and nutrition have increased theneed for a complete understanding of nutrientconsumption patterns in the United States. It is,therefore, imperative that the effect of thesechanging food consumption patterns on the fooddistribution system and on the nutritional status ofthe consumer be examined. Although consider-able literature exists on demand models for nutri-ents (i.e. Price, et al., 1978; Akin, et al., 1983;Chavas and Keplinger, 1983; Scearce and Jensen,1979; Devaney and Fraker, 1989; Adrian andDaniel, 1976; Lane, 1978; Davis and Neenan,1979; Windham, et al., 1983), little attention ispaid to the analysis of the demand for nutrientsderived from either FAFH or FAH. Evaluationof the nutrients that are consumed from eitherFAFH or FAH would provide a means of assess-ing the nutritional quality of American diets aswell as a comprehensive description and under-standing of nutrient consumption patterns not onlyfrom total food consumed but also from FAFHand FAH.

Little is known too about the demographicand socio-economic characteristics of individualswho have either eaten away from home or individ-uals who have eaten a particular meat productaway from home or at home. With the exceptionof the Lee and Brown piece (see Table 2), nostudies as yet have analyzed the effect of socio-demographic and economic factors on the decisionto eat FAFH. Furthermore, Lee and Brown’sstudy used the 1977-78 National Food Consump-tion Survey and, therefore, may not reflect currentmarket conditions. The restaurant and fast foodindustries would benefit from a study that wouldprovide them some information regarding thedemographic and socio-economic profile of con-sumers who eat out. Likewise, informationderived from this study would be useful for pro-cessors and producers who want to anticipatefiture market changes and derived demands fortheir products.

This research attempts to fill these voids byusing the Individual Intake phase of the 1987-1988


Table 2. Selected Studies on Food Away from Home (FAl?H)Expenditure and Consumption

Socio-DemographicResearcher(s) Data Set? Factors Considered Focus of the Study

Derrick, Dardis, 1972-73 CES Income, Household size, Impact of demographic vari-Lehfeld Age, Education, Region, ablea on FAFH expenditure

Urbanization, Employment,Race, Marital Status

Prochaska and 1965-66 Urbanization, Income, Race, Effect of opportunity cost ofSchrimper NFCS (Spring Region, Employment, homemaker’s time on FAFH

Portion) Number of children consumption

Kinsey PSID Various sources of income, Effect of various sources ofRace, Employment, House- income on marginal propens-hold size ity to consume FAFH

Redman 1972-73 and, Income, Employment, Impact of socio-economic fac-1973-74 BLS, Family composition, Educa- tors and women’s time alloca-CES tion, Age, Region, Race, tion on FAFH

Urbanization

Sexauer 1960-61 and, Family size, Age, Urban- Effects of demographic shifts1972-73 BLS, ization, Education, Sex, and income distributionCES Employment, Income changes on FAFH expenditure

Table 2. cent

Lee, Brown 1977-78 Income, Urbanization, Factors affecting away fromNFCS Region, Employment, Educa- home and at home consump-

tion, Race, Household size tion and the decision to eatout

McCracken and 1977-78 Education, Age, Retirement, Factors affecting FAFHBrandt NFCS Region, Race, Urbanization, expenditure by type of facility

Income, Household composi-tion and size

Lippert, Love 1980 BLS, Income, Employment, Relationship between FAFHCES Education, Family size and expenditure and socio-

composition, Region, Race, economic characteristics ofUrbanization household between 1972-73

and 1980

‘ BLS, CES = Bureau of Labor Statistics, Consumer Expenditure Survey. NFCS = National FoodConsumption Survey. PSID = Panel Study of Income Dynamics,

Journal of Food Distribution Research February 931page73

National Food Consumption Survey. Informationon food consumption and nutrient intake of indi-viduals in the away from home and at home mar-kets are available from this particular data set.This research will attempt to assess the impact ofvarious socio-economic characteristics of theindividual not only on total nutrient consumptionbut also on nutrient consumption away from homeand at home. This specification would allow notonly the assessment and comparison of demandfor nutrients derived fkom FAFH and nutrientsderived from FAH but also would allow a com-parison of the results with previous studies.

This study also attempts to identify, indefinitive fashion, the demographic and socio-economic characteristics of individuals who haveeaten away from home and individuals who haveeaten a particular meat product either away hornhome or at home. Moreover, FAFH quantitymodels that examine the effects of sociodemo-graphic and economic factors on FAFH consump-tion by type of facility are also developed. Due tothe unavailability of the household expenditurephase of the 1987-88 National Food ConsumptionSurvey during the completion stage of this study,no theoretical and empirical models of householdexpenditures for FAFH are developed in thisstudy.

Monthly time series data are used in athree-commodity complete demand system frame-work to derive price and income elasticity esti-mates for food away from home, food at home,and non-food. Although the commoditiesinvolved are broad aggregates, the results of thisstudy could be used by the food distribution andretail industry as an aid in making importantpricing and policy decisions.

Objectives

The objectives of this research are: (1) toemploy a complete demand systems approachusing time series data for FAFH, FAH, and non-food to estimate own-price, cross-price, andincome elasticities; (2) to determine factors affect-ing individual intake of nutrients derived fromFAFH, nutrients derived from FAH, and nutrientsderived from total food consumption; (3) to iden-tify the demographic and socio-erxmomic charac-

teristics of consumers who have eaten away fromhome and of those who have eaten a particularmeat product either away from home or at home;and (4) to determine factors affecting FAFHconsumption by type of facility using quantitymodels. The model specifications are exldbited inthe Appendix.

Data Sources

Two sets of data are used in this study.The first data set consists of monthly time seriesdata from 1970 to 1989 gathered from variousgovernment statistical documents or publications.This data set is used in the estimation of thedynamic Almost Ideal Demand System. Thesecond data set is the 1987-88 National FoodConsumption Survey (NFCS) from the U.S.Department of Agriculture. This data set is usedis the estimation of the nutrient demand equations,the Logit models, and the quantity models forfood away from home.

The monthly time series data from 1970 to1989 consist of FAFH, FAH, and non-foodexpenditures, consumer price indexes, and con-sumption expenditures. The FAFH expendituredata are derived from monthly retail sales ofeating and drinking places in the United States.Eating and drinking places include restaurants,lunchrooms, cafeterias, and fast-food operationsor refreshment places. In 1989, commercialeating and drinking places accounted for two-thirds of the retail equivalent value of expendi-tures for FAFH (Food Retailing Review, 1991).The other third of the expenditures value forFAFH came from schools, hotels and motels,military facilities and other facilities. The FAHexpenditure data, on the other hand, are derivedfrom monthly retail sales of food stores. Foodstorm include grocery stores, meat and fish mar-kets, and bakeries. The source of both the eatingand drinking places, and the food stores sales datais the Bureau of Census. Non-food expendituresare derived by deducting both FAFH and FAHexpenditures from total consumption expenditures.Another variable used in the analysis is the laborforce participation rate of women. These data areobtained from the various issues of Employmentand Earnings publications.

February 93/page 74 Journal of Food Distribution Research

The data used in the nutrient demand, logitanalyses, and quantity models for FAFH comefrom the U.S. Department of Agriculture’s 1987-88 Nationwide Food Consumption Survey(NFCS). This data set is the most recent of thenational household food consumption surveysconducted by USDA. The 1987-88 survey con-tains two parts: (1) household food use and (2)individual intake. The household phase providesinformation on food used by the household for aone-week period and on the cost of that food.The individual intake phase, on the other hand,provides three days of information on food intakeof household members. The individual intakephase of the 1987-88 NFCS data set marks onlythe fifth time that nationwide information on thedietary intakes of individual household membershas been collected by USDA. Only the individualintake phase is used in the empirical analysis partof this study due to the unavailability of the hous-ehold food use phase during the completion stageof this dissertation.

The individual intake phase of the 1987-88NFCS data set provides data on three days of foodand nutrient intake by individuals of all agessurveyed in the 48 contiguous states. These indi-viduals were asked to provide three consecutivedays of dietary data. Respondents were alsoasked about the sources of each food eaten.Sources included food that was eaten at home,food brought into the home but later eaten awayfrom home, and food that was never brought intothe home. USDA considers food from the firsttwo sources to be from the home food supply.Thus, this study considers food from the first twosources to be food at home (FAH), and the thirdsource as food away from home (FAFH). Infor-mation is also available about the place where theFAFH was obtained (i.e. restaurants, school, fast-food eatablishments, or someone else’s home).

The individual intake data set also includesinformation of the individual on the followingvariables: urbanization, region, race, sex, employ-ment status, food stamp participation, WIC partic-ipation, National School Lunch and NationalSchool Breakfast Programs participation, specialdiet information, household size, age, householdincome, food sources, and foods consumed.

The response rate by households in thesurvey was low. In particular, participation byhouseholds drawn into the sample was below 35percent. This is lower than in previous NFCSdata sets. USDA indicated that a major reason forthis occurrence was “heavy respondent burden” interms of the amount of information asked fromeach respondent.

As in any cross-sectional study, severalissues arise in handling the data set. The originalnumber of respondents in the survey is 11,045.However, several individuals in the sample haveincomplete socio-economic and demographicinformation. Subsequently, observations withmissing individual relevant socio-economic anddemographic information were deleted.

Summary of the Results and their Implications

Dynamic Almost Ideal Demand System Model

The first part of this study pertains to thedevelopment of a three commodity completedemand system using monthly time series data forFAFH, FAH, and non-food. Using the linearapproximation of the dynamic version of theAlmost Ideal Demand System, price and expendi-ture elasticities are derived that could aid in theimprovement of market planning and pricingdecisions for the food distribution sector. Othervariables included in the system are the laborforce participation rate of women and monthlyseasonal dummy variables. Consistent with priorexpectations, the results indicate that the share oftotal expenditure going to FAFH increases as thelabor force participation rate of women increases.However, the expenditure share for FAHdecreases as the labor force participation rate ofwomen increases. These results could imply thatwomen tend to rely more on FAFH than FAH asthey become more involved in the labor force andas their opportunity costs of time become higher.Differences in the seasonal patterns are also evi-dent in the FAFH and FAH equations.

In accord with economic theory, all theexpenditure elasticities are positive and all theown-price elasticities are negative. The demandfor FAFH is more price sensitive than the demandfor FAH based on the compensated own-price

Journal of Food DistributionResearch February 931page75

elasticities of 43.741 and -0.426 for FAFH andFAH, respectively. Based on the compensatedcross-price elasticities, there seems to be generalsubstitutability relationships among the three broadcommodities of FAFH, FAH, and non-food. Theprice of non-food, however, has a larger effect onthe demand of either FAFH or FAH compared tothe price effects of FAFH and FAH on thedemand for non-food.

This part of the study documents the use oftime series data in determining the demand forFAFH, FAH, and non-food. Although the com-modities examined are broad aggregates, theanalyses in this study could be used by the fooddistribution and retail industry as an aid in maldngimportant pricing and policy decisions.

NutrientDemand Models

Another objective of this research is todetermine the factors affecting individual intake ofnutrients derived from either all food, FAFH, orFAH, Although considerable literature exists ondemand models for nutrients, little attention ispaid to the analysis of the demand for nutrientsderived from either FAFH or FAH. Evaluationof the nutrients that are consumed from eitherFAFH or FAH provides a means of assessing thenutritional or dietary quality of American diets aswell as provide a comprehensive description andunderstanding of nutrient consumption patterns notonly from total food consumption but also fromFAFH and FAH. In addition, the informationobtained from this study on the effects of variousdemographic and socio-economic variables on theindividual’s nutrient intake ftom FAFH could beused by nutrition educators and policy makers intargeting their nutrition education programs anddollars.

The results in this study generally indicatethat average nutrient intakes from FAFH arelower than average nutrient intakes from FAHwith the exception of fats, saturated fatty acids,mono-unsaturated fatty acids, and polyunsaturatedfatty acids. Interestingly, the results also indicatethat roughly 30 percent of the food energy kilo-calories comes from FAFH while the remaining70 percent comes from FAH. These results haveimportant implications for the away from home

food industry. The fast food industry, forinstance, has been criticized for serving foodwhich are “unhealthy”due to its high fat content.An example would be a fast food meal of a ham-burger, tkench fries, and milk shake which con-tains approximately half the RDA of calories andprotein for the adult male but only gives aboutone-third of the RDA of vitamin C, thiamin, andniacin, and even lesser amounts of iron, calcium,vitamin A, and riboflavin (Putnam and Van Dress,1984).

Concern about health and fitness hasencouraged consumers to prefer foods perceivedas “fresh” and “light. ” In a 1988 Gallup pollconducted for the National Restaurant Association,almost 60 percent of adult consumers claimed tobe very interested in nutrition-conscious menuitems in the away from home industry. As moreconsumers demand menu items with improvednutritive value, FAFH operators will have toadapt to these nutrition-conscious patrons if theyare to stay competitive. Some of the fast foodchains have, however, started to respond posi-tively to the demand of the customers for healthierand low fat food. An example of this isMcDonald’s which recently unveiled their 91percent fat free “McLean Deluxe” burger. Like-wise, Hardee’s is rolling out a new low fat burgercalled the ‘Real Lean Deluxe” (Wall StreetJournal, July 15, 1991; p. Bl). In addition,restaurants and fast food chains are enlarging theirmenus to include salad bars and lighter dishes.

Individuals residing in suburban areas haveslightly higher intake of most of the nutrientsanalyzed from FAFH as a percentage of RDAthan individuals residing in central cities or non-metro areas. Likewise, individuals from theSouth have slightly higher intakes of most nutri-ents from FAFH as a percentage of RDA thanthose from other parts of the country. Also hav-ing higher intakes from FAFH as a percentage ofRDA are the following: employed individualscompared to unemployed individuals; non-foodstamp recipients compared to food stamp recipi-ents; and those who are not on a special dietcompared to those who are. Across all sex andage groups, the average intake of nutrients fromFAH is generally 50 percent or over of RDA.Except for the intake of calcium and magnesium


by certain population groups, the individuals inthe sample seem to have generally acquired atleast two-thirds of the recommended daily allow-ances from all foods which indicate that the sam-ple, which is representative of the population, isbasically of good health status.

The nutrient consumption regression modelshave also disclosed some interesting results. Forinstance, individuals who reside in central cities orsuburban areas consume lower amounts of fatsand the various fatty acids per 1000 kilocaloriesfrom FAFH than individuals who reside in non-metro areas. Significant regional differences arealso evident in the consumption of various nutri-ents from FAFH. For example, individuals fromthe South seem to generally consume moreamounts of various nutrients except alcohol fromFAFH but less amounts from FAH than individ-uals from other regions of the United States.Males, as expected, consume more nutrients flomeither FAFH or FAH than females with the excep-tion of fats and the various fatty acids, Evident inthe results is the positive relationship between theweight of an individual and the consumption ofvarious nutrients from FAFH.

Employed individuals consume moreamounts of various nutrients except alcohol anddietary fiber away from home than unemployedindividuals. In contrast, employed individualsgenerally consume less amounts of various nutri-ents at home compared to unemployed individuals.As expected, individuals who are on a special dietand individuals who receive food stamps signifi-cantly consume less amounts of numerous nutri-ents away from home than their counterparts.

An increase in household size is generallyassociated with an increase in nutrient consump-tion from FAH but a decrease in nutrient con-sumption from FAFH. This result is expectedconsidering that the cost of eating out is propor-tionate to the number of persons eating out.Moreover, age and income are generally signifi-cant factors affecting individual consumption ofmany nutrients either from FAFH or FAH. Inparticular, consumption of almost all the nutrientsfrom FAFH except for some energy yieldingnutrients decreases (increases) initially with suc-cessive increments of age (income) and then

increases (decreases). The consumption pattern ofmost nutrients from FAH is the reverse withinitial increases (decreases) and then followed withdecreases (increases) with successive incrementsof age (income).

Logit Models

Little information is known about the demo-graphic and socio-economic characteristics ofindividuals who have either eaten away fromhome or individuals who have eaten a particularmeat product away from home or at home. Logitmodels are developed to investigate the decision toeat FAFH and the decision to eat a particular meatproduct either away from home or at home. Theresults of these logit models would be of signifi-cant interest to not only the restaurant and fastfood industry but also to the various meat indus-tries (i.e. beef, pork, lamb, poultry, fish).

The results indicate that the following indi-viduals are less likely to eat FAFH: individualsfrom the Northeast compared to those in theSouth; blacks and hispanics compared to whites;unemployed individuals compared to employedindividuals; food stamp recipients compared tonon-food stamp recipients; those on a special dietcompared to those not on a special diet; individ-uals who consumed their food mostly duringweekdays compared to individuals who consumedtheir food mostly during weekends; and largerhouseholds compared to smaller households.

The likelihood of eating FAFH decreaseswith age but increases with income, ceteris pari-bus. Also, the probability of eating FAFH issignificantly lower during the first and third quar-ters of the year compared to the second quarter ofthe year.

Generally, the demographic and socio-eco-nomic profiles of individuals eating the same meatproduct are different across the three differentsources: FAFH, FAH, and all food. Similarly,the demographic and socio-economic profiles ofindividuals who eat fkom the same source aredifferent across the various meat products.

Contrasting results are apparent between theFAFH and FAH logit models across the various


meat products, Employed individuals, forinstance, are more (less) likely to eat a particularmeat product away from home (at home) thanunemployed individuals. This result could berelated to the fact that employed individuals mighthave less time to prepare home-cooked meals thanunemployed individuals. In addition, resultsgenerally indicate that the probability of eating ameat product away from home (at home)decreases (increases) as the household sizeincreases. Once again, the cost of eating out ismore with larger households than smaller house-holds. The weekend variable is also a significantfactor in most of the meat logit models for FAFHbut not in meat logit models for FAH. This resultcould imply that during weekends, the likelihoodof eating meats away from home is greater thanthe likelihood of eating meat at home.

Blacks are generally more likely to eatpoultry, fish and shellfish but less likely to eatbeef than whites either away from home or athome, Individuals residing in central cities andsuburban areas are also more likely to eat poultry,fish and shellfish but less likely to eat beef andpork at home than those residing in non-metroareas, Males are more likely to eat beef andpork. Individuals who are on a special diet aregenerally more likely to eat poultry, fish andshellfish, lamb, veal, and game but less likely toeat beef and pork than those who are not on aspecial diet. This result might be related to theperception that beef and pork have higher fatcontent than other meat products. Thus, the beefand pork industries might have to emphasize the“leanness” of their products in their promotioncampaigns to recapture health and nutrition con-scious consumers. The poultry, fish, and lambindustries, on the other hand, should not onlycontinue their efforts on promoting the “healthful-ness” of their products but also focus on attractingminorities (i.e. blacks) as well as individuals wholive in non-metro areas.

The logit models in this study identified thetypes of individuals that are more likely to eatFAFH as well as the types of individuals who aremore likely to eat various meat products eitheraway from home or at home, The identificationof these types of consumers is essential in analyz-ing consumption behavior and developing specific

marketing programs. Consequently, these infor-mation should aid market analysts focus theirefforts on the group of consumers less likely toeat out or on the group of consumers less likely toeat a particular meat product either away fromhome or at home.

QuantityModels

The quantity models are also developedusing the Heckman procedure to determine thefactors affecting FAFH consumption not onlyfrom all food facilities but also from the differenttypes of FAFH facility. The FAFH consumptionmeasure used is number of meals purchased by anindividual. The findings from the quantity modelfor all types of FAFH facilities indicate that thefollowing variables significantly affect the numberof meals purchased: regional variables as a group;race variables as a group; ethnicity; sex; house-hold size; weekend variable; age; and income.Importantly, the results also indicate thatemployed individuals consume more meals awayfrom home than unemployed individuals. Thisresult supports the hypothesis that individuals withhigher opportunity cost of time, assuming thatemployed individuals have higher opportunity costof time than unemployed individuals, purchaseand consume more meals away from home.

The disaggregate regression estimates forthe number of meals consumed from restaurants,fast food facilities, and other away from homefood facilities reveal differing significance of thevarious sociodemographic factors by type ofFAFH establishment. Of most importance is theresult that suggests that employed individualssignificantly consume more meals from fast foodfacilities but only marginally consume more mealsfrom restaurants than unemployed individuals.Income is not statistically significant in any of thethree models on FAFH facilities. These findingsmay suggest that individuals eat at restaurants notonly to save time but also to acquire some recre-ational diversion. Moreover, these results maysuggest that consuming meals away from home inrestaurants and fast food facilities depends less onincome than on the employment status of theindividual.


These results from the quantity models maybe of considerable importance for the restaurant,fast food, and other away from home industries.For instance, the findings in this study suggestthat marketing efforts by the-se FAFH industriesshould be focused on individuals who purchaserelatively fewer meals away from home. Theseindividuals may include those with larger house-hold sizes, females, those who are unemployed,and even those who are on specials diets. Thefast food industry (includes cafeterias and self-service restaurants in this study) should also caterto the taste of older people if it wants to boost itssales.

Concluding Remarks andAreas for Further Research

The increasing complexity of the away fromhome food industry as well as the heightenedconsumer interest in health and nutrition, haveincreased the need for a complete understandingof away from home consumption patterns in theUnited States. To keep up with the recent trendsin consumer demand, the FAFH industry mustcontinually be in the midst of creating a vast arrayof concepts that appeal to specific consumer tastesand preferences. It is, therefore, imperative thatthe effects of these changing food consumptionpatterns on the food distribution system and on thenutritional status of the consumer be examined.

This study identifies several socio-economicand demographic characteristics affecting foodaway from home consumption. The objectives ofthis study are: (1) to estimate own-price, cross-price, and income elasticities for food away fromhome, food at home, and non-food using a com-plete demand systems approach and time seriesdata; (2) to determine factors affecting individualintake of nutrients derived from either food awayfrom home, food at home, or total food consump-tion; (3) to identify the sociodemographic charac-teristics of consumers who have consumed foodaway fkom home and those who have eaten aparticular meat product either away from home orat home; and (4) to determine factors affectingfood away from home consumption by type offacility. The individual intake phase of the 1987-1988 National Food Consumption Survey is usedto accomplish the last three objectives.

Consistent with prior expectations, resultsindicate that as the labor force participation rate ofwomen increases, the share of total expendituregoing to food away fkom home increases but theexpenditure share of food at home decreases. Aswell, results generally indicate that distinct differ-ences exist between the type of characteristics thatsignificantly affect the intake of nutrients fkomfood away from home and food at home. Varioussocio-economic and demographic characteristicsalso affect the following: (1) the likelihood ofeating food away from home; (2) the likelihood ofconsuming a particular meat product (i.e. beetpork; lamb, veal, and game; poultry; and fish andshellfish) either away from home or at home; andthe number of meals purchased away from homeby type of facility.

Due to the unavailability of the householdfood use or expenditure phase of the 1987-88NFCS data during the completion stage of thisdissertation, no theoretical and empirical modelson household expenditures for FAFH are devel-oped in this study. Although quantity models aredeveloped using number of meals as measure forFAFH consumption, it is still essential to developtheoretical and empirical models of FAFH expen-ditures considering the significant alterations inexpenditure patterns and the apparent substitutionof expenditures on FAFH for FAH. The resultsof this type of study using the household food usephase of the 1987-88 NFCS data set can then becompared to the results from the McCracken andBrandt study in 1987 which used the 1977-78NFCS data.

Future research should also focus on theeffect of increased availability of convenience andprepared foods, which reduce the opportunity costcomponent of eating at home, on FAFH consump-tion. For instance, little is known about the cross-price elasticities of eating away from home and ofeating a home cooked meal or prepared/conveni-ence food at home. As well, limited informationis available regarding the factors that influencerelative market shares of these %ommoditiea”over time.

So far, no studies have dealt with FAFHexpenditures on a commodity basis (i.e. beef,pork, poultry, fish, etc.). Scant information is,


therefore, available on demand parameters forFAFH expenditures by type of commodity, Thistype of research could be handled with the use ofthe Consumer Reports on Eating Share Trends(CREST) data by the NPD Group. The CRESTdata series, collected by the NPD Group since1976, is gathered via a comprehensive anddetailed diary in which 12,800 U.S. householdsrecord their restaurant visits and purchases ofmeals, snacks, and beverages. The households,which are dispersed among the 48 contiguousUnited States, are recruited by mail using a strati-fied random quota sampling system. The CRESTdata series tracks over 140 different food andbeverage items. This series is the most compre-hensive data set available on household purchasepatterns of food in the away from home market.

The role of inventory demand and habits onconsumer expenditure patterns on FAFH and FAHshould also be examined. This research would,however, need the availability of a comprehensivetime series data with variable price and quantitydata on food from the away from home and athome markets, The analysis in this type of studycan be centered on the use of the Houthakker-Taylor state adjustment model. Generally, inven-tory demand tends to dominate habits in the shortterm. Likewise, short-run consumer behavior, asopposed to longer-run consumer behavior, isusually influenced more by consumer inventoriesthan habits, particularly for food. It would beinteresting to know the role of inventory demandand habits not only on aggregate FAFH and FAHbut also on disaggregate commodities from awayfrom home and at home markets.

Related to the topic of habit formation isstructural change. In demand analysis, shifts inthe utility fimction may result from outside infor-mation and other external influences on the con-sumer or from variables related to past decisions(i.e. habit formation). It would be worthwhile toinvestigate structural change in consumer behaviorwithin the FAFH market. Ingco andManderscheid (1988) discussed various ways totest for the presence of structural change indemand models. Although difllcult, it would alsobe important to assess the magnitude of structuralchange, if there are any, in consumer behaviorwithin the away fkom home food industry. The

results that can be obtainedhorn this type of studycould be vital in policy analysis and forecastingfor the food industry.

References

Adrian, J. and R. Daniel, “Impact of Socio-economic Factors on Consumption ofSelected Food Nutrients in the UnitedStates,” American Journal of AgriculturalEconomics, 58(1976):31-38.

Akin, J. D., D. K. Guilkey, and B. M. Popkin,“The School Lunch Program and NutrientIntake: A Switching Regression Analysis,”American Journal of AgriculturalEconomics, 65(1983):477-85.

American Demographics, March 1991,

Business Week, “Suddenly, Asian Americans area Marketer’s Dream,” June 17, 1991, pp.54-55.

Chavas, J. P. and K. O. Keplinger, “Impact ofDomestic Food Programs on NutrientIntake of Low Income Persons in theUnited States,” Southern Journal of Agri-culturalEconomics, 15(1983):155-63.

Davis, C. G. and P. H. Neenan, ‘The Impact ofFood Stamps and Nutrition Education Pro-grams on Food Groups Expenditure andNutrient Intake of Low Income House-holds,” Southern Journal of AgriculturalEconomics, 11(1979): 121-29.

Derrick, F., R. Dardis, and A. Lehfeld, “TheImpact of Demographic Variables onExpenditures for Food Away from Home,”Journal of the Northeastern AgriculturalEconomics Council, 11(1982): 1-10.

Devaney, B. and T. Fraker, “The DietaryImpacts of the School Breakfast Program,”American Journal of AgriculturalEconomics, 71(1989):932-48.

Food Retailing Review, 1991 Edition, The FoodInstitute, January 1991.

Februsry 93/page 80 Journal of Food Distribution Research

Heckman, J. J., “Sample Sekction Bias as aSpecification Error, ” Econometrku,47(1979):153-161.

Houthakker, H. and L. Taylor. ConsumerDemand in the UnitedStates: Analysis andProjections. Cambridge, Mass., HarvardUniversity Press, 1970.

Ingco, M. and L. Manderscheid, ‘ModellingParameter Variation in EconometricModels,” Agri, Econ, Report No. 506,Dept. of Agri. Econ., Michigan StateUniversity, March 1988.

Kinsey, J., “Working Wives and the MarginalPropensity to Consume Food Away FromHome,” American Journal of AgriculturalEconomics, 65(February 1983):10-19.

Kinsey, J., “Diverse Demographics Drive theFood Industry,” Choices, Second Quarter1990, p. 23.

Lane, S., “Food Distribution and Food StampProgram Effects on Nutritional Achieve-ment of Low Income Persons in KernCounty, California,” American Journal ofAgricultural Economics, 60(1978):108-16.

Lee, J. and M, Brown, “Food Expenditures atHome and Away from Home in the UnitedStates- A Switching Regression Analysis,”lhe Review of Economics and Statistics,68(1986):142-47.

Lippert, A. and D. Love, “Family Expendituresfor Food Away from Home and PreparedFoods,” Family Economics Review,3(1986):9-14.

Manchester, A. C., “Food Expenditures at aGlance,” National Food Review, ERS-USDA, 13(1990):25.

McCracken, V. A. and J. A. Brandt, “HouseholdConsumption of Food Away From Home:Total Expenditure and by Type of FoodFacility,” AmericanJournal of AgriculturalEconomics, 69(May 1987):274-284.

Nayga Jr,, R. M., Anulysis of Socio-economicand Demographic Factors A#ecting FoodAway j?om Home Consumption, Ph.D.dissertation, Texas A&M University,December 1991.

Noah, T., “Married Couples Are Found to HeadFewer Households,” WaZlStreet JournaZ,June 11, 1991, p. A6.

Pindyck, R, S. and D. L. Rubinfeld, EconometricModels and Economic Forecasts, ThirdEdition, McGraw-Hill, 1991,

Price, D. W,, D. A. West, G. E. Schier, and D.Z. Price, “Food Delivery Programs andOther Factors Affecting Nutrient Intake ofChildren,” American Journal of Agricul-tural Economics 60(1978):609-18.

Prochaska, F. J. and R. A. Schrimper,“opportunity Cost of Time and OtherSocioeconomic Effects on Away fromHome Food Consumption,” AmericanJour-nal of Agricultural Economics,55(November 1973):595-603.

Putnam, J. J. and M. G. Van Dress, “ChangesAhead for Eating Out,” National FoodReview. 26(1984):15-17.

Redman, B., “The Impact of Women’s TimeAllocation on Expenditure for Meals Awayfrom Home and Prepared Foods,”American Journal of AgriculturalEconomics, 62,2(May 1980):234-37.

Scearce, W. K., and R. B. Jensen, “Food StampProgram Effects on Availability of FoodNutrients for Low Income Families in theSouthern Region of the United States,”Southern Journal of AgriculturalEconomics, 11(1979):113-20.

Sexauer, B., “The Effects of Demographic Shiftaand Changes in the Income Distributions ofFood Away From Home Expenditure,”American Journal of AgriculturalEconomics, 61(December 1979):1046-57.

Journal of Food Distribution Research Febmary 931page 81

U.S. Dept. of Agriculture, National Food Situa-tion, Economic Research Service,Washington D. C., various issues.

U.S. Dept. of Commerce, Survey of CurrentBusiness, Bureau of Economic Analysis,Washington D. C., various issues.

U.S. Dept. of Commerce, Business Statt”stics,Bureau of Economic Analysis, WashingtonD. C., various issum.

U.S, Dept. of Commerce, Current PopulationRepotis, Bureau of Census, WashingtonD. C., various issues,

U.S, Dept. of Labor, Employment and Earnings,Bureau of Labor Statistics, WashingtonD. C., various issues.

U.S. Statistical Abstracts, 1988-89.

Waldrop J. and T. Exter, “The Legacy of the1980s,’’AmericanDemographics,13(March1991):32-38.

Wall Street Journal, “Lean and Mean: Hardee’sJoins Low-Fat Fray,” July 15, 1991, p. B1.

Windham, C., B. Wyse, and R. Hansen,“Nutrient Density of Diets in the USDANationwide Food Consumption Survey,1977-1978: II. Adequacy of Nutrient Den-sity Consumption Practices,” JournaZof theAmerican Dietetic Associm”on, 82(January1983):34-43.

Dynamic Almost Ideal Demand System Model

The dynamic linear approximation almostideal demand system (LA/AIDS) specificationused in this study involves a three commoditydemand system. The three commodities analyzedare FAFH, FAH, and non-food. The demandsystem is also augmented with the inclusion of alagged budget share, eleven monthly seasonaldummy variables and a variable representing thelabor force participation rate of women. Monthlydummy variables are used to capture the effects ofseasonality. The coefficients associated with thesevariables may be positive or negative. Moreover,the labor force participation rate of women isincluded in the analysis since it is hypothesizedthat as the labor force participation rate of womenincreases, the demand for FAFH also increasesbecause of increasing opportunity cost of time ofwomen. On the other hand, the demand for FAHas the labor force participation rate of womenincreases could either increase or decreasedepending on the relative strength of the substitut-ability between home-cooked food and conve-nience or prepared foods.

Februsry 93/page 82

The consumer price indexes for FAFH,FAH, and CPI-less food are used as price vari-ables in the analysis. In addition, personal con-sumption expenditures data are used as totalexpenditure instead of disposable personal incometo limit the influence of savings in the analysis.

When estimating demand systems, oneequation must be omitted. This will avoid singu-larity in the variance-covariance matrix of theresiduals across equations. The commodity that isarbitrarily omitted in the model is non-food. Themodel is estimated using iterative Zellner’s seem-ingly umelated regression technique (IZEF) withhomogeneity and symmetry restrictions imposedand with a first order serial correlation correction.

Nutrient Demand Models

The twenty-eight nutrients selected for theanalysis, along with the units of measurement, areshown in Table A, 1 below. Based on previousstudies and conditioned on the data available inthe 1987-88NFCS, the independentvariablesusedinclude urbanization, region, race, sex, employ-ment, household size, age, height, weight, and

Journsl of Food Distribution Research

income. Dummy variablea pertaining to whetherthe individual receives food stamps or not;whether the individual is on special diet or not;and whether the intake of nutrient occurred mostlyduring a weekend or a weekday are also included.

Table A.1

List of Nutrients Used in the AnalysesAnd Their Unit of Measurement

Nutrient Unit of Measurement

1.2.3.4.

5.

6.

7.8.9.

10,11.12.

13.

14.15.16.17.18.19.20.21.22.23.24.25.26.27.28.

Food EnergyProteinTotal FatSaturated FattyAcidsMonounsaturatedFatty AcidsPolyunsaturatedFatty AcidsCholesterolCarbohydrateTotal DietaryFiberAlcoholVitamin ACarotenes

Vhamin E

Vhamin CThiaminRiboflavinNiacinVitamin B6FolateVitamin B12CalciumPhosphorusMagnesiumIronZincCopperSodiumPotassium

kilocaloriesgrams per 1000 kilocaloriesgrams per 1000 kilocaloriesgrams per 1000 kilocalories

grams per 1000 kilocaloriescalories

grams per 1000 kilocalories

milligramsgrams per 1000 kilocaloriesgrams per 1000 kilocalories

grams per 1000 kilocaloriesinternational unitsmicrograms retinol

equivalentsmilligrams alpha-tocopherol

equivalentsmilligramsmilligramsmilligramsmilligramsmilligramsmicrogramsmicrogramsmilligramsmilligramsmilligramsmilligramsmilligramsmilligramsmilligramsmilligrams

The general model specificationused is therefore:

Nh = b, + blurbanl + bzurban2 +b,regionl + b,region2 + b5region4 +b#ace2 + b7race3 + b8racwl + b~ispl+ blOsexl + bllemployl + blzfstampl +bl,dietl + blqhsize + bl~weight +bl~eight + bl,age + blsagesq +blgweekend + bmincome + bzlincomesq

where:

NM = average daily intake of nutrient k by indi-vidual i. The units of measurement are dis-played in Table A. 1.

urbanl = 1 if individual resides in a central city;O otherwise

urban2 = 1 if individual resides in a suburbanarea; O otherwise

regionl = 1 if individual is in the Northeast; Ootherwise

region2 = 1 if individual is in the Midwest; Ootherwise

region4 = 1 if individual is in the West; O other-wise

race2 = 1 if individual is black; O otherwiserace3 = 1 if individual is Asian or Pacific

Islander; Ootherwiserace4 = 1 if individual is of some other race; O

otherwisehispl = 1 if individual is hispanic; O otherwisesexl = 1 if individual is male; O otherwiseemployl = 1 if individual is employed; O other-

wisefstampl = 1 if individual is receiving food

stamps; O otherwisedietl = 1 if individual is on a special diet; O

otherwisehsize = household sizeweight = weight of the individual in poundsheight = height of the individual in inchesage = age of the individual in yearsagesq = square of the age of the individualweekend = 1 if the threeday intake of the indi-

vidual occurred mostly during a weekend;Ootherwise

income = household incomeincomesq = square of household income

One classification is eliminated from eachgroup of variables for estimation purposes. Thebase group are individuals who satisfy the follow-ing description: reside in a nonmetro area

Journal of Food Distribution Research

. .

February 93/page 83

(urban3); in the South (region3); white (raceI);non-hispanic (hisp2); female (sex2); not employed(employ2); not participating in the food stampprogram (fstamp2); not on a special diet (diet2);and the threeday intake occurred mostly during aweekday (weekday). Household income is usedinstead of individual income because the NFCSdata set only provides income information for thehousehold and not for an individual. The analysesare separated into three different food sources:FAFH, FAH, and all foods eaten, to determinenutrient consumption pattern differences and thefactors that affect nutrient consumption acrossthese three food sources. Since excess consump-tion of one nutrient does not compensate fordeficiencies in another, twenty-eight separatenutrient consumption models are specified foreach of the three food sources to explain nutrientintake. Also, the same set of independentvariables is used for each nutrient consumptionmodel because nutrients are constituent parts offood and therefore, may affect the consumption ofeach nutrient analyzed. Each nutrient may beaffected differently by the various independentvariables included, but there are no a priorireasons to include or exclude any of these factorsin any of the nutrient equations. The energyyielding nutrients: protein, total fat, saturated fattyacids, monounsaturated fatty acids,polyunsaturated fatty acids, carbohydrate, andalcohol are all expressed as nutrient densities or ingrams per 1000 kilocalories to allow propercomparison between individuals. Theanthropomorphic measurements of the individual -age, sex, height, and weight - are included as

independent variables to account for physicaldifferences between individuals. For instance,male individuals might eat more than femaleindividuals and taller and heavier individualsmight eat more than shorter and lighterindividuals. Thus, squared terms are included forincome and age in order to investigate possiblenordinearities in the Engel relationships forFAFH, FAH and all food consumption.

Depending on the proportion of zero obser-vations on a dependent variable, either the OLS orthe Heckman Sample Selection Procedure is usedin the analysis. When the proportion of zeroobservations on a dependent variable is high,omitting the zero observations in the OLS runs

will result in the estimates characterized by sam-ple selection bias. Subsequently, Heckman (1979)proposed a technique that amounts to estimatingthe omitted variable using probit analysis and thenemploying either OLS or generalized least squaresto the model with the inclusion of the estimatedomitted variable.

Logit Models

Logit analysis is used in the estimation ofmodels that would investigate the decision to eatthe following meat products:

(1) beet(2) pork;(3) lamb, veal, and game;(4) poultry; and(5) fish and shellfish

either away from home, at home, or both.

Logit models are employed in the analysesto circumvent the inadequacies of the linear proba-bility model and because of the dichotomousnature of the dependent variable that is used.These models are based on the cumulative logisticprobability function and are specified as (Pindyckand Rubinfeld, 1991):

P = F(Z) = F~i’@) = 1/(1 +e-z) = 1/(1 +e-m~)

where Z is a theoretical index determined by a setof explanatory variables X; F(Z) is the cumulativelogistic function; e represents the base of naturallogarithms (approximately equal to 2,718); and Pis the probability that an individual will make acertain choice, given the knowledge of X.

The most suitable technique of estimationwhen using logit is maximum likelihood.Although this technique requires the use of itera-tive algorithm, this procedure assumes the large-sample properties of consistency and asymptoticnormality of the parameter estimates so that con-ventional tests of significance are applicable.

The logit analyses center on the hypothesisthat a set of variables influence the decision to eatFAFH and the decision to eat various meats either


away from home, at home, or both. The logitmodels are specified as follows:

PROB = bO+ blurbanl + bzurban2 +bqregionl + b,region2 + b#egion4 +b#ace2 + b7race3 + b8ractA + b~ispl+ blOsexl + bllemployl + blzfstampl +b,,dietl + bl,hsize + b,~ogage +bl~ogincome + bl,weekend +b18quarterl + bl~uarter3 + bnquarter4

where PROB represents the following dependentvariables:

(1) equal to 1 if the individual consumed somenutrient from FAFH and Ootherwise;

(2) equal to 1 if the individual consumed beeffrom FAFH and O otherwise;

(3) equal to 1 if the individual consumed beeffrom FAH and O otherwise;

(4) equal to 1 if the individual consumed beeffrom all foods and O otherwise;

(5) equal to 1 if the individual consumed porkfrom FAFH and O otherwise;

(6) equal to 1 if the individual consumed porkfrom FAH and Ootherwise;

(7) equal to 1 if the individual consumed porkfrom all foods and O otherwise;

(8) equal to 1 if the individual consumed lamb,veal, and game from FAFH and O other-wise;

(9) equal to 1 if the individual consumed lamb,veal, and game from FAH and Ootherwise;

(10) equal to 1 if the individual consumed lamb,veal, and game from all foods and Oother-wise;

(11) equal to 1 if the individual consumed poultryfrom FAFH and O otherwise;

(12) equal to 1 if the individual consumed poultryfrom FAH and O otherwise;

(13) equal to 1 if the individual consumed poultryfrom all foods and O otherwise;

(14) equal to 1 if the individual consumed fishand shellfish from FAFH and O otherwise;

(15) equal to 1 if the individual consumed fishand shellfish from FAH and O otherwise;

(16) equal to 1 if the individual wmsumed fishand shellfish from all foods and O other-wise.

The independent variables include logagewhich is the logarithm of age; logincome which isthe logarithm of income; and quarterl, quarter3,and quarter4 which correspond to a set of binaryvariables that measure seasonality, (qaurterl = 1 ifJanuary -March; quarter3 = 1 if July-September;quarter4= 1 if October-December) (referencecategory, April-June). The rest of the indepen-dent variables are defined the same way as in thenutrient demand equations.

Quantity Models

To determine the impact of sociodemo-graphic and economic factors on FAFH consump-tion, Heckman procedure is used to estimateFAFH quantity models. Given the hypothesis thatthe demand for FAFH differs by type of facility,regression models are estimated separately for thenumber of meals purchased at restaurants, fastfood establishments, and other facilities. Basedon past studies and conditioned on the data avail-able in the 1987-88 NFCS data set, the modelspecification suggest estimation of the followingequations:

MEAL = bO+ blurbanl + bzurban2 +b~regionl + b~region2 + b~region4 +b#ace2 + b,race3 + b*race4 + b$iispl+ blOsexl + bllemployl + blzfstamp1 +bl~dietl + blghsize + bl~ogage +bl~ogincome + bl,weekend +blaquarterl + b1gquarter3 + bnquarter4+ bzl imratio

where MEAL represents the number of mealspurchased by an individual from the followingfood sources:

(1) away from home per unit of time (3 days);(2) restaurants per unit of time;(3) fast food facilities per unit of time; and(4) other away from home facilities per unit of

time.

The independent variables consist of thesame set of variables used in the logit models andare therefore defined the same way. Due to therelatively high proportion of zero observations inthe dependent variables, the Heckman procedureis used in estimating the models. An additional

Journal of Food Distribution Research February 93@age 85

variable (imratio) is, therefore, included as anexogenous variable. The variable “imratio” is theinverse of Mill’s ratio and is defined as the ratioof the value of the standard normal density func-tion to the value of the standard normal distribu-tion function. For this study, meals are defined toinclude only breakfast, brunch, lunch, dinner, andsupper. Snacks, infant feeding, and other eatingoccasions are, therefore, not considered as meals.In addition, restaurants only refer to those res-taurant facilities with waiter or waitress service.On the other hand, fast food facilities refer to self-service food facilities, cafeterias, and food facili-ties where food is ordered and picked up at thecounter. Other facilities include schools, day carecenters, vending machines, stores, and communityfeeding programs.


apple baum scholarship paper analysis of socio-economic and

Documents