Download to read offline
![3.[19 28]children influence on parents buying decisions in delhi (india)](https://cdn.slidesharecdn.com/ss_thumbnails/3-19-28childreninfluenceonparentsbuyingdecisionsindelhiindia-111125091146-phpapp01-thumbnail.jpg?width=640&height=640&fit=bounds)
Keith Final Draft
- 1. Determinants of Consumer Spendingon Apparel Keith Howard 12/15/2011 Paperpreparedunderthe directionof Dr.DavidCrary to fulfill part of the research requirement for the M.A. in Economic Trade and Development at Eastern Michigan University Thisresearchapplieseconometricanalysis todataobtainedfromthe Bureauof Labor Statistics ConsumerExpenditure Surveyto testandquantifydemographicandeconomicdeterminantsof spendingonapparel. The goal isto helpinvestorsandbusinessanalystsunderstandapparel salesgiven specificeconomicand demographicdatafordifferentregionsof the UnitedStates. The findings reinforce pastresultsonthe effectsof genderandincome onspendingonapparel;aswell asdiscover some newsignificanteffectslike thatof populationdensityof locationandthe effectsof race as a determiningfactor.
- 2. 1 Introduction: The factors whichdetermine apparel expenditure are sometimes assumed and sometimes implied. It is a widely believed assumption that women spend more on apparel than men, as well as the assumption that wealthier people tend to spend more on apparel1. What other factors are included in determining how much a household or an individual spends on clothing and accessories? The topic I will be researching and analyzing are the determining factors that make up consumer behavior pertaining to apparel consumption. I will use economic and demographic information to compile a model that will determine an estimate of apparel expenditures for individuals with different demographic and economic characteristics. The independent variables I will use will include income, education level, age, race, gender, average hours worked, marital status, and familial status. Many of these variables have been tested in similar studies described below. My study will focus on the information obtained through the Bureau of Labor Statistics (bls.gov) from their Consumption Expenditures Survey which gathers this information on behavioral spending habits of individuals on a quarterly basis. This data can be used to help understand the typical consumer’s spending on apparel. This can then be tailored for further studies of specific geographic areas2. Abernathy et al discusses how current retailing strategies include information technologies that analyze consumer sales data, and that this is the future of the retail and apparel industries. Bram & Ludvigson suggest that analyzing consumer sentiment “can help predict future movements in consumer spending.” My goal is to provide relevant 1 Heim states in his paper on consumer confidence and consumption spending, “If income declines…we expect decliningconsumption.”So the adverse is also assumed. 2 The Consumer Expenditures Survey does not includedata on zip code, but demographic research compiled in the future could investin obtainingthis information for a better fitof the model and assumptions.
- 3. 2 summaries of thedetermining factors of apparel expenditures. Then one could make comparisons of different demographical areas to help decide where optimal locations would be to invest in a store that primarily sells apparel. Past work has been completed to help show determining factors of apparel consumption, but not to understand the consumption of apparel given the above demographics. Vieira (2010) and Pentecost and Andrews (2010) both conducted surveys where they were fortunate enough to obtain data on fashion awareness. But, they also found that in annual data, this was an insignificant factor3. These papers did not include potential important variables that will be included in this research such as population density of the respondent’s region of residence, race, and income. The data used in this paper will help further the knowledge on this subject because it will build on earlier studies but use an expanded set of explanatory variables. The data used in this paper is cross-sectional, compiled by surveys conducted by employees of the Bureau of Labor Statistics, and is conducted on a quarterly basis over 5 quarters to substantially encompass all purchases made in a year. The data comes from the Consumer Expenditure Survey (CEX) of 2010. These are the most recent data available from the BLS4. Literature Review: Multiple studies have been written, with varying themes, regarding the determinants of consumer expenditures on apparel. Some studies focus on sports apparel purchases, some focus 3 At the 5% level. 4 This data was made public for purchaseon September 27th.
- 4. 3 on adolescent spendingon apparel, and some focus on online apparel shopping behavior. Demographic and psychographic variables are taken into consideration in much of the research (Scheerder, Vos, & Taks, 2011; Vieira, 2010; Beaudoin & Lachance, 2006; Sampson, 2009; Pentecost & Andrews, 2009). Demographic information such as age, income, and education are common throughout. A study compiled by Scheerder et al specifies expenditures on sports apparel with included variables of “sport-related lifestyle characteristics” such as active sport participation and sport spectatorship. This study continues with the theory that identification5 with a team, and not just active participation, is another determinant of sport apparel expenditures. Similarly, it was found in a previous study by Wicker et al that a one hour increase in weekly sport participation increases sport expenditures by €263 (roughly $354). Work completed by Sampson shows that there is a growing trend of “Lifestyles Of Health And Stability” (LOHAS) over the past decade. She contends that the growth of LOHAS has come from a consumer who is becoming more socially and environmentally aware of the impacts of the products one purchases. Her study focuses on the consumer’s purchasing behavior of green apparel. Attention was paid to consumer’s knowledge of green apparel companies, beliefs regarding the general environment, and attitudes towards green apparel, among other variables contributing towards the purchase of green apparel (Sampson, 2009). Awareness of a company’s intentions and practices can easily be compared to knowledge of brand awareness. Brand awareness is another psychological determinant that contributes to apparel expenditures (Beaudoin et al, 2006). Adolescents6, or Generation Y, pay more attention 5 This may be linked to one’s sense of belongingin a community. For example, someone who is not a loyal,or even a casual sportfan,and is fromthe Detroit area - might purchasea Detroit Tiger’s hat to identify with the area they are from. Itis my own theory that team logos are not only an identifier of fan-ship,butteam logos are also an identifier of one’s loyalty to their city or region. 6 At the time of this paper, adolescents = Generation Y.
- 5. 4 to fashion thanother generations (Beaudoin et al, 2006; Vieira, 2010; Pentecost et al, 2009). This can be measured through expenditures on apparel compared to income of the individual, and through behavioral surveys of one’s attitudes towards fashion. Where Sampson’s study focused on variables such as knowledge of green companies, Beaudoin et al compares psychological factors such as consumer confidence, self-esteem, and fashion innovativeness as variables contributing to consumption behavior. Fashion is also seen as a status symbol, especially with regards to high-end or designer apparel. Fogel et al found this to be a contributing factor in designer purchases. This is similar to findings from the Beaudoin et al study which includes self-esteem as a contributing factor. It is well documented and theorized that gender plays a role in apparel consumption (Scheerder, Vos, & Taks, 2011; Vieira, 2010; Beaudoin & Lachance, 2006; Sampson, 2009; Pentecost & Andrews, 2009). Females have a higher awareness of brands (Beaudoin et al, 2006), fashion trend awareness (Vieira, 2010), and frequency of purchases (Pentecost et al, 2009). Pentecost et al found that in a three week period, frequency of female apparel purchases were 71% to males’ 29%. So it would be wise to assume that gender has to be a variable of any model describing or forecasting apparel consumption. Vieira and Pentecost et al both theorize that fashion awareness7, is an important determinant that should be included in any model. Fashion, as defined in the papers above, is a basic term for apparel that follows current trends in the culture. Pentecost et al found that that fashion fanship was a significant positive factor of apparel expenditures weekly, and monthly. But adversely found that it was not statistically significant (p>0.05) when regressed with yearly expenditures (Pentecost et al, 2009). Vieira hypothesized that materialism played a significant 7 Vieira denotes this as “fashion apparel buyingbehavior”;whilePentecost et al refers to this as “fashion fanship”.
- 6. 5 role in differentiatingthose who buy fashion apparel and those who do not. It was found that materialism had no significant effect on fashion expenditures (Vieira, 2010). Data for these studies was typically collected by surveys given to college students (Vieira, 2010; Sampson, 2009), high school students (Beaudoin et al, 2006), surveys randomly given to people at popular malls (Pentecost et al, 2009), and in one case, primary and high school students took home surveys to their parents (Scheerder et al, 2011). A survey is the best method for this type of study because the data that is to be tested needs to include individual demographical and psychological factors as discussed above. One disadvantage to the above studies is that a few were given exclusively to younger generations. This can create some bias based on the hypothesis that I hold that older generations are going to behave differently given apparel spending habits. Another study had responses only from parents and guardians of school-age children (Scheerder et al, 2011), which only represents people who have children; and totally excludes the spending habits of people without children (who have more disposable income!) The study completed by Pentecost et al was given to random people at malls, and this is closest to a random sample of the population. But one thing stands out in this method: the survey was only given to people at malls – who obviously have a predisposition to spending on apparel! It is not a good estimation of the population at large. That is why I will use the Consumer Expenditure Survey, which I will discuss later. Also, research completed on spending habits of online apparel consumers found that they were a different type of shopper. The online shopper, (who is referred to as a hedonistic, recreational shopper) gets just as much pleasure from shopping online as in malls or brick-and-mortar8 8 Brick-and-mortar refers to physical retail establishments.Click-and-mortar retailers arethosethat have online shopping,as well as physical retail stores.
- 7. 6 venues (Cowart &Goldsmith, 2007). All types of shoppers are captured in the Consumer Expenditure Survey, not just people who shop in face-to-face contact with retailers. My dependent variable will be consumer spending on apparel for an average quarter9. This is similar to the research compiled previously which I have discussed above. But most of the abovementioned research had different timeframes in which the total apparel spending was measured. The major differences will be in the explanatory variables I will use. Though I am restricted to using socio-economic data, my data is better suited to making good assumption of the national population because it is a much more diverse random sample of consumers in the United States. I am missing some important qualitative variables such as fashion-awareness, brand awareness, and self-esteem10, which I would theorize would have some effect on the dependent variable. The research completed in this paper will include the largest data set of any of the papers found relating to this same subject. The Consumer Expenditure Survey captures data from almost 150,000 respondents, of which I have broken into 2 groups of interest: Married couples and their family members living in their household; and single individuals living on their own with no dependents in their households. Theoretical Model: The dependent variable in my model will be quarterly spending on apparel. The explanatory variables will be income, education level, age, race, gender, employment status, marital status, and familial status. I believe these variables determine a household’s and an 9 The Consumer Expenditure Survey takes randomsurveys of respondent’s spendinghabits coveringeach quarter of the calendar year. 10 Though Pentecost et al found similar explanatory variables insignificant.
- 8. 7 individual’s habits ofspending on apparel. The models use variables that have been found to have significance in explaining spending on apparel, such as: income, age, and gender. I will also add familial status, average hours worked per week, education, and race because these socio-economic factors, I believe, will each influence spending in their own ways. Someone with kids will increase their spending on apparel, someone who is employed will have a positive effect on spending, and race and education may have something to do with spending – but I will have to determine the effect after the data is analyzed (to see if there is any significance of those variables in the model). Single Person Model: Ylog_tot_apparel = β0 + β1Xmega + β2Xmetro + β3Xcity + β4Xtown + β5Xrural + β6Xprimary + β7Xsome_hs + β8Xdiploma + β9Xsome_coll + β10Xassoc + β11Xbach + β12Xmasters + β13Xphd + β14Xwhite + β15Xblack + β16Xasian + β17Xpac_is + β18Xmulti_r + β19Xage_ref + β20Xage_ref_sq + β21Xfemale + β22Xlog_fsalaryx + µ Married Household Model: Ylog_tot_apparel = β0 + β1Xmega + β2Xmetro + β3Xcity + β4Xtown + β5Xrural + β6Xprimary + β7Xsome_hs + β8Xdiploma + β9Xsome_coll + β10Xassoc + β11Xbach + β12Xmasters + β13Xphd + β6Xsp_primary + β7Xsp_some_hs + β8Xsp_diploma + β9Xsp_some_coll + β10Xsp_assoc + β11Xsp_bach + β12Xsp_masters + β13Xsp_phd + β14Xsp_white + β15Xsp_black + β16Xsp_asian + β17Xsp_pac_is + β18Xsp_multi_r + β19Xage2 + β20Xage2_sq + β21Xfemale + β22Xlog_fsalaryx + µ Variable Description: mega: dummy variable, MSA larger than 4 million metro: dummy variable, MSA 1.2 – 4 million city: dummy variable, MSA 0.33 – 1.19 million town: dummy variable, MSA 125 – 329.9 thousand rural: dummy variable, MSA less than 125 thousand “sp_” before these variables denotes spouse characteristic primary: dummy variable, only primary school completed some_hs: dummy variable, some high school completed diploma: dummy variable, high school gradute some_coll: dummy variable, only some college completed
- 9. 8 assoc: dummy variable,associate’s degree bach: dummy variable, bachelor’s degree masters: dummy variable, master’s degree phd: dummy variable, phd “sp_” before these variables denotes spouse characteristic white: dummy variable, respondent is white black: dummy variable, respondent is black asian: dummy variable, respondent is Asian pac_is: dummy variable, respondent is Pacific Islander multi_r: dummy variable, respondent is multi-racial “sp_” before these variables denotes spouse characteristic age_ref: age of respondent age_ref_sq: quadratic form of age_ref variable female: dummy variable, 0 if male, 1 if female log_fsalaryx: log form of annual salary of respondent/household no_earnr: number of earners in the household inc_hrs1: average number of hours worked by respondent inc_hrs2: average number of hours worked by spouse of respondent age2: age of spouse of respondent age2_sq: quadratic form of age2 variable perslt18: number of people in household under age of 18 fam_size: number of people in household Expectations: Using these models to understand the determinants of possible sales given specific demographics will help investors decide where the best area will be for the store they want to open. Also, changing demographics for the same area can affect the dependent variable in these models. For example, imagine a store that has been in business in a small suburban neighborhood that has grown to a more populous affluent community. The demographics of that area would undoubtedly change as well, which would have an effect on the amount households
- 10. 9 and individuals willspend on apparel. This model can be easily modified to predict possible apparel sales as well. But for the purpose of this paper, I am going to analyze the explanatory variables. I expect that the denser the MSA11 of the respondent, the more they will spend on apparel. This variable captures part of the effect of fashion-awareness because people who live in more populous areas tend to be more inclined to define themselves through their apparel12, and therefore more conscious of their personal impression on other people. So, I hypothesize that as the MSA gets smaller compared to the control variable of mega, less and less will be spent on apparel. H1: Xmega > Xmetro H2: Xmega > Xcity H3: Xmega > Xtown H4: Xmega > Xrural The variables for education and race will be interesting to look at because there is no real theory behind whether they will increase of decrease spending. Though I hypothesize the variable black will have a larger effect on apparel spending than the control variable white. This could be because of a combination of two main factors: African-Americans are more likely to live in urban environments13 and the African-American culture is more fashion-conscious14. 11 The term “MSA” will bediscussed in full in theData Description section of the paper. 12 This is anecdotal to the author. 13 i.e. Detroit is over 90% African-American. 14 Purely anecdotal.
- 11. 10 Hispanic origin ofrace is its own separate dummy variable15. Thus, the effect of this characteristic cannot be captured in the race variable because Hispanic is not an option. Most Hispanic people select white as their race in the normal race variable. So the effect of Hispanic is captured in the race variable as white, more often than it will be captured in other race selections. This could drive up the average purchases of apparel of white respondents (if my hypothesis that Hispanic and black respondents spend more on apparel), and/or could drive down the average apparel expenditures because the average income of Hispanic respondents is lower than the average Caucasian white respondents. There is a possibility that the effects of these characteristics offset each other, but probably not to the exact same amount. H5: Xblack > Xwhite I also expect that as age increases, spending on apparel will decrease. This is an easy variable to predict because so much of the research discussed above has mentioned this and used this variable in previous research. The variable female is also an easy variable to predict the effect on apparel spending. This will have a positive effect on the dependent variable. H6: Xfemale > 0 The amount of time a person works on average during a week will probably have some effect on spending on apparel in a positive way. This variable affects apparel spending because the more hours a person works, the more they will have to spend on apparel. It could be a 15 In the CEX, the racequestion was separatefrom the Hispanic origin question.
- 12. 11 confounding variable becauseof the correlation to income, but I feel it is important to keep in the model because it helps capture employment status. Also, I will only use it in the household model because in the individual model it will be highly correlated with income. Because of the research complied above16, I will include the variable perslt18, which captures the number of people in the household under the age of 18 that are not the respondent or the spouse. I hypothesize that this will have an effect on apparel spending because the more individuals in the household who are under 18 means more fashion-aware individuals who will likely want to spend more on apparel (or at least spend more off their parent’s income on apparel). H7: Xperslt18 > 0 I also hypothesize that the more earners there are in the household model, the more disposable income the family will have, which will lead to an increase in apparel spending. Also, the more people in the household, I believe there will be an increase in spending on apparel as well. H8: Xno_earnr > 0 H9: Xfam_size > 0 16 Which found that younger people spent more on apparel than older people.
- 13. 12 Since my modelhas a logarithmic dependent variable, a co-efficient over 0 shows that as that variable increases by one unit, and then we will see an increase in spending by that percentage, ceteris paribus. Also, previous economic theory suggests that as income increases, so will consumption. This is represented by the following hypothesis: H10: Xlog_fsalaryx > 0 Data Description: The sample is a random survey completed by the BLS each quarter that captures all purchases and demographic information for the survey respondents. This data is cross-section data. The survey collects the expenditure data or each respondent for the past and current quarters. I am using the data for one quarter of purchases for each respondent. The reason the BLS collects data for each quarter is so that it can more accurately assess expenditures by quarter, instead of trying to find respondents who keep records of their annual purchases. The time period this data was collected ranges from the first quarter of 2010 through the first quarter of 2011. It is compiled this way because the respondent is asked about the current quarter and the previous quarter in each survey. Each respondent is surveyed for their purchases per quarter, and the dependent variable reflects apparel purchases in the quarter in which the survey was completed after. For example, if a survey was completed in Q1 of 2010, then the respondent’s answers reflect Q4 of 2009.
- 14. 13 The Consumer ExpenditureSurvey collects similar variables as the research completed previously, such as age, gender, and spending on apparel. But it does not collect qualitative data that is very subjective such as brand awareness or fashion-awareness. Most of the variables are qualitative demographic variables that are dummy variables. The continuous variables are age_ref, log_fsalaryx, no_earnr, inc_hrs1, inc_hrs2, age2, perslt18, fam_size. There are some non-normal distributions found in the summary statistics. These were in the continuous variables fsalaryx and tot_apparel. The variable fsalaryx was only used if greater than zero so a logarithmic representation could be made from this data. Logarithmic variables were created for fsalaryx, and tot_apparel. These were named log_fsalaryx and log_tot_apparel, respectively. Also, quadratic forms of age and age2 were created. The variables that represent the population density of the respondent are mega, metro, city, town, and rural. These represent the sizes of the Metropolitan Statistical Areas (MSA’s). An MSA is a Primary Sampling Unit which is a group of counties which have not only geographical connections, but economic connections as well. Simply put, it is a measured characteristic of a small region in the United States, or metropolitan area. For instance, one of the MSA’s (PSU’s) in the sample data is represented by the Detroit-Ann Arbor-Flint MSA. This includes the counties within this triangulation of locations. In the sample data, there are 21 MSA’s (PSU’s) with a population greater than 1.5 million, which would put them into the metro or mega descriptive variables depending on population size. The apparel expenditures variable contains any spending on apparel from men’s socks, women’s dresses, accessories17, kids shoes, to girl’s outerwear, men’s suits, women’s lingerie, 17 Such as gloves,sunglasses,scarves,ties,belts,etc…
- 15. 14 etc., etc. Basicallyanything a person can wear. In the CEX these can all be separated into their own categories and measured; but for the purposes of this paper I have included them all together to measure general apparel expenditures. The typical respondent in the Single Person Model spends $505.17 per quarter on apparel, lives in an MSA larger than 4 million18, is white19, has some college education20 (but no degree), is 41.8 years old, female21, and has an annual salary of $41,467.65. The typical respondent in the Household Model spends $874.01 per quarter on apparel, is 45.3 years old and has a spouse age 45, lives in a household of 3.7 people with 2 income earners and 1.3 people under the age of 18, works an average of 41.2 hours per week with a spouse who works an average of 41.3 hours per week, lives in an MSA of more than 4 million people22, has a bachelor’s degree23 with a spouse who also holds a bachelor degree24, is white25 with a spouse who is also white26, and with an annual household salary of $95,706.47. 18 37% likelihood 19 79% likelihood 20 27% likelihood 21 60% likelihood 22 41% likelihood 23 25% likelihood 24 26% likelihood 25 85% likelihood 26 84% likelihood
- 16.
- 17. 16 Correlations are foundin the Single Model as the level of education increases with salary. For example, the correlation between education and income begins with no_sch at -0.0331, and increases gradually to masters at .2128. Also from the Single Model, the correlation with population density is strongest with mega and fsalaryx at .2522. Econometric Results: Regression was completed for both models and a few variables were found to be insignificant in both. For the Single Person model these were inc_hrs127, no_sch and native28. In the Household model the variables found to have low significance were age2, age2_sq29, and perslt1830. Also, White’s Robust standard errors we used to correct for any non-normal distributions left after other measures had been taken. The final regression models are as follows: Single Person Model: Ylog_tot_apparel = β0 + β2Xmetro + β3Xcity + β4Xtown + β5Xrural + β6Xprimary + β7Xsome_hs + β8Xdiploma + β9Xsome_coll + β10Xassoc + β11Xbach + β12Xmasters + β13Xphd + β15Xblack + β16Xasian + β17Xpac_is + β18Xmulti_r + β19Xage_ref + β20Xage_ref_sq + β21Xfemale + β22Xlog_fsalaryx + µ Married Household Model: Ylog_tot_apparel = β0 + β2Xmetro + β3Xcity + β4Xtown + β5Xrural + β6Xprimary + β7Xsome_hs + β8Xdiploma + β9Xsome_coll + β10Xassoc + β11Xbach + β12Xmasters + β13Xphd + β6Xsp_primary + β7Xsp_some_hs + β8Xsp_diploma + β9Xsp_some_coll + β10Xsp_assoc + β11Xsp_bach + β12Xsp_masters + β13Xsp_phd + β15Xsp_black + β16Xsp_asian + β17Xsp_pac_is + β18Xsp_multi_r + β19Xblack + β20Xasian + β21Xpac_is + β22Xmulti_r + β23Xage_ref + β24Xage_ref_sq + β25Xfemale + β26Xlog_fsalaryx + β27Xinc_hrs1 + β28Xinc_hrs2 + β29Xfam_size + β30Xno_earnr + µ 27 T-stat of -0.91. 28 no_sch and native were found to be highly non-normal distribution and collinear,with lowt-stats. 29 age2 t-stat of -0.94, and age2_sq t-stat of 1.7 was so barely significantthatI dropped both from the model. 30 T-stat 0.45.
- 18. 17 Linear regression Number ofobs= 12509 Single Person log_tot_apparel Coef. t-stat age_ref*** -0.0487852 -10.78 age_ref_sq*** 0.0005161 10.28 female*** 0.4632305 22.87 log_fsalaryx*** 0.2721552 24.81 metro* -0.0478414 -1.85 city*** -0.1831202 -4.84 town*** -0.1113173 -4.28 rural*** -0.1590931 -4.85 primary_sch*** -0.293145 -3.19 some_hs*** 0.254782 3.70 some_coll*** 0.1074565 3.65 assoc*** 0.2637531 7.08 bach*** 0.217131 7.20 masters*** 0.2484496 7.43 phd*** 0.4595739 7.98 black*** 0.0760165 2.44 asian*** 0.1357596 3.65 pac_is*** -1.35049 -18.64 multi_r* -0.1138039 -1.75 _cons 3.420335 30.87 Adj. R-squared = 0.1446 Root MSE = 1.0344 d.f. = 12490 significance at 10% * 1.645 significance at 5% ** 1.960 significance at 1% *** 2.576
- 19. 18 I had previouslyexpected that the denser the respondent’s population surroundings, the more that respondent would spend on apparel. All of the coefficients for population are negative. But, the two densely populated MSA-types are substantially less than the three sparsely populated MSA-types. Linear regression Number of obs = 46168 Married Household log_tot_apparel Coef. t-stat log_tot_apparel Coef. t-stat metro*** -0.198951 -16.87 black*** -0.1417449 -3.21 city*** -0.1036346 -5.39 asian*** -0.2924466 -9.70 town*** -0.1933415 -15.43 pac_is*** -0.4509522 -4.63 rural*** -0.3486247 -18.51 multi_r* -0.0445095 -1.37 primary_sch*** 0.1930799 4.07 sp_black*** 0.1599659 3.58 some_hs*** 0.1057396 3.59 sp_asian* -0.0471836 -1.43 some_coll*** 0.1341075 8.57 sp_pac_is* -0.105139 -1.41 assoc*** 0.0656294 3.72 sp_multi_r* -0.0635888 -1.54 bach*** 0.1496439 9.69 age_ref_sq*** -0.0003361 -9.09 masters*** 0.0681685 3.88 no_earnr*** 0.0512558 5.87 phd*** 0.1550854 6.36 inc_hrs1*** 0.0044886 11.67 sp_primary_sch*** -0.1882662 -5.92 inc_hrs2*** 0.0025496 6.68 sp_some_hs*** 0.0903885 3.28 fam_size*** 0.0888853 21.27 sp_some_coll** 0.0296262 1.89 age_ref*** 0.0369061 10.98 sp_assoc*** 0.0663848 3.70 log_fsalaryx*** 0.2152439 24.48 sp_bach*** 0.126177 8.80 _cons 2.123106 20.99 sp_masters*** 0.1990366 11.08 sp_phd*** 0.2181962 7.87 Adj. R-squared = 0.1167 Root MSE = 0.9595 d.f. = 46164 significance at 10% * 1.645 significance at 5% ** 1.960 significance at 1% *** 2.576
- 20. 19 H1: Xmega >Xmetro Accept H2: Xmega > Xcity Accept H3: Xmega > Xtown Accept H4: Xmega > Xrural Accept In both the Household Model and the Single Person Model the variables for location are found to be significant. In the Single Person Model, an individual who lives in an MSA that falls into the metro category only spends on average about 5% less on apparel each quarter than an individual who lives in an MSA that is considered mega. While individuals in less populous areas spend 11% - 18% less on apparel, on average31. In the Household Model, the comparison shows a starker contrast from mega to the smaller MSAs. The difference ranges from 10% to 35% less32, with the largest difference between mega and rural at 35% less spent on apparel, on average. I also expected that as the respondent’s education level increased, so would their spending on apparel because of the implied higher income level. In both models, what is found is that each higher level education variable increases spending on apparel. This may mean that the education dummy variables are capturing part of the income of the respondents33. I also expected that respondents who are black would spend more on apparel. And a respondent who is black increases apparel (7.6%) compared with a white respondent, as well as respondents who are Asian (13.6%). H5: Xblack > Xwhite Accept 31 City = 18% less,Town = 11% less,Rural =16% less 32 Metro = 20% less,City = 10% less,Town = 19% less,Rural = 35% less 33 Refer to section on correlations above.
- 21. 20 The only properway to test the female variable is to test only individuals, which is why a model was created just for single people. The Single Person Model shows that females spend much more on average than males: roughly 46% more34. This corresponds with previous research discussed above regarding the spending habits of women compared to men. H6: Xfemale > 0 Accept The variable perslt18 was found to be insignificant with the lowest t-stat of roughly 0.45, so it was left out of the final model. It had a co-efficient that increased household spending on apparel by only 0.5% for each member of the household that was under the age of 18. H7: Xperslt18 > 0 Reject In the Household Model both no_earnr and fam_size were found to increase spending on apparel, and both were found to be significant. As the number of earners in the household increases, spending on apparel increases 5%. As the number of individuals in the household increases, spending on apparel increases almost 9%. H8: Xno_earnr > 0 Accept H9: Xfam_size > 0 Accept The co-efficient for log_fsalaryx in the Single Person model is 0.27, and similarly in the Household model is 0.22; which both show very low income elasticity. This could be due to part of the income effect being picked up by other descriptive variables, such as the slightly 34 With a very significantt-statof 22.87
- 22. 21 correlated variable megaand masters35. Also, low income elasticity is normal for necessity goods36. H10: Xlog_fsalaryx > 0 Accept As inc_hrs1 and inc_hrs_2 increase by an average of 1 more hour worked per week, the amount spent on apparel increases by 0.45% and 0.25%, respectively. This means if the average hours worked per week by the respondent37 by 10 hours, spending on apparel will increase 4.5% (which is probably due to correlation with the variable fsalaryx of .1876) Summary Research was completed on this subject because past research was very selective in the sample of the population at large that was used to obtain data. The data used in this paper was obtained from a very reliable source and gives a better representation of the U.S. population as a whole. The assumptions made from the data are more indicative of the populations of question that these models can be used for. This paper had the opportunity to focus on some interesting hypotheses which the past research hadn’t completed. The hypotheses of population density and race are the two most interesting to understand. Though sometimes small differences, the two variables show a difference in priorities of cultures within our country. What exactly can be inferred from these findings I will leave to sociologists and psychologists. I believe the research I have completed uses the available data to make valid assumptions of consumer behavior. The data is objective and contains relevant and current information. The 35 Correlations of 0.3103 and 0.2128, respectively.Both come from the Single Model and are the only two which stand out in the correlation tables discussed earlier. 36 Income elasticity of demand states that positive income elasticity lower than 1 are usually dueto normal goods that are also necessity goods – such as apparel. 37 Referring to inc_hrs1.
- 23. 22 missing variables whichhad been used in previous research, such as brand awareness and fashion-ability, are subjective and hard to quantify in a fair and scientific manner. Who is to say how fashionable they believe themselves to be on a scale of one-to-ten? But, if it were possible to obtain this information from each respondent in a fair and objective way, this would have much to do with the behavior of consumer’s spending on apparel38. The models used in this research only received Adjusted R2 statistics of .144639 and .116740. Though these are very low indicators41 the impacts of the variables used in each model are still significant, if not highly significant. But more goes into the respondent’s choice of spending on apparel than can be measured by basic demographics. This analysis betters our understanding of the determinants that go into a consumer’s apparel spending behavior. It furthers the research completed before that looked only at specific groups of people and uses a data set that is much more representative of the United States population as a whole. I find these results both internally and externally valid due to the extensive nature of the Bureau of Labor Statistic’s Consumer Expenditures Survey. The next steps that should be taken in continuing this research would be to create a more in-depth character survey of the respondents. Being able to obtain data on fashion sense, brand awareness, self-esteem, uniqueness, and trend awareness would help to further this research immensely. But these are very subjective topics, and it will be hard to scientifically and accurately gather this information. This research is beneficial to investors and business analysts 38 Contrary to pastresearch which found these variables insignificant. I think this was probably dueto some sampling error. 39 For the SinglePerson Model 40 For the Household Model 41 Shows largevariationsin expenditures
- 24. 23 due to thevalid assumptions that can be drawn from the data within, but more work can be done to help complete a better model with more variables. Summary Statistics: Single Person OBS = 12509 Variable Mean Std. Dev. Min Median Max Skew Kurtosis JB Stat tot_apparel 505.1699 1339.8440 1 285 38600 22.4650 628.2575 204816969.516 mega 0.3716 0.4833 0 1 0.5310 1.2824 metro 0.2408 0.4276 0 1 1.2120 2.4699 city 0.0552 0.2283 0 1 3.8960 16.1810 town 0.2091 0.4067 0 1 1.4300 3.0459 rural 0.1175 0.3220 0 1 2.3760 6.6455 no_sch 0.0024 0.0494 0 1 20.1637 407.5739 primary_sch 0.0132 0.1141 0 1 8.5358 73.8599 some_hs 0.0381 0.1914 0 1 4.8264 24.2942 diploma 0.1717 0.3772 0 1 1.7407 4.0299 some_coll 0.2668 0.4423 0 1 1.0543 2.1117 assoc 0.0944 0.2924 0 1 2.7747 8.6992 bach 0.2640 0.4408 0 1 1.0706 2.1461 masters 0.1152 0.3192 0 1 2.4110 6.8128 phd 0.0341 0.1815 0 1 5.1332 27.3503 white 0.7941 0.4044 0 1 -1.4544 3.1154 black 0.1291 0.3353 0 1 2.2129 5.8968 native 0.0052 0.0721 0 1 13.7164 189.1386 asian 0.0556 0.2292 0 1 3.8788 16.0451 pac_is 0.0014 0.0373 0 1 26.7161 714.7514 multi_r 0.0146 0.1201 0 1 8.0794 66.2768 age_ref 41.7969 15.8069 17 41 87 0.2368 1.9639 676.413 female 0.5985 0.4902 0 1 -0.4020 1.1616 fsalaryx 41467.6500 39911.6800 1 34000 279006 2.7197 15.0430 91013.491 log_income 10.1219 1.2215 10.4341 -1.2801 5.6315 7025.654 log_tot_ap~l 5.5817 1.1176 5.6525 -0.2254 3.9247 551.640
- 25.
- 26. 25 Bibliography: Abernathy, Frederick H.,John T. Dunlop, Janice H. Hammond, David Weil. 1995. “The Information-Integrated Channel: A Study of the U.S. Apparel Industry in Transition.” Brookings Papers on Economic Activity. Microeconomics, Vol. 1995 (1995), pp. 175-246 Beaudoin, Pierre, Marie J. Lachance. 2006. “Determinants of Adolescents’ Brand Sensitivity to Clothing.” Family and Consumer Sciences Research Journal. Vol. 34, No. 4, June 2006 pg. 312-331. Bram, Jason, Sydney Ludvigson. 1998. “Does Consumer Confidence Forecast Household Expenditure? A Sentiment Index Horse Race.” Economic Policy Review. Cowart, Kelly O., Ronald E. Goldsmith. 2007. “The influence of consumer decision-making styles on online apparel consumption by college students.” International Journal of Consumer Studies. 31 (2007) 639–647. Fogel, Joshua, Mayer Schneider. 2009. “Understanding designer clothing purchases over the internet.” Journal of Fashion Marketing and Management. Vol. 14 No. 3, 2010 p. 367-396 Heim, John J. 2010. “The Impact of Consumer Confidence on Consumption and Investment Spending.” Journal of Applied Business and Economics vol.11(2) Pentecost, Robin and Andrews, Lynda. 2010. “Fashion retailing and the bottom line : The effects of generaional cohorts, gender, fashion fanship, attitudes and impulse buying on fashion expenditure.” Journal of Retailing and Consumer Services, 17(1). pp. 43-52. Sampson, Laura Kathryn. 2009. “Consumer Analysis of Purchasing Behavior for Green Apparel.” North Carolina State University. Raleigh, North Carolina. Scheerder, Jeroen, Steven Vos, Marijke Taks. 2011. “Expenditures on Sport Apparel: Creating Consumer Profiles through Interval Regression Modeling.” European Sport Management Quarterly, 11:3, 251-274. Vieira, Valter Afonso. 2010. “Determinants of the Purchase or Non-Purchase of Fashion Apparel: an Exploratory Study.” R.Adm., São Paulo, v.45, n.2, p.130-141. “2010 CONSUMER EXPENDITURE INTERVIEW SURVEY PUBLIC USE MICRODATA” User’s Documentation. Bureau of Labor Statistics. September 27, 2011.