Is Higher Education A Luxury
Good?
An Analysis of the Veblen Effect on Demand
for Ivy League Universities Between 1998-2019
Jadah S. Francis-Ferrell
Universiteit van Amsterdam
A thesis presented for the degree of
BSc Economics and Business
11 June 2020
Statement of Originality
This document is written by Jadah S. Francis-Ferrell who declares to take full re-sponsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
In today’s capitalist society, consumerism is often debated over its environmental, social and economic implications. The topic is often discussed in terms of material-ism in relation to physical goods. However, one expenditure that is often overlooked during these economic discussions is education. Although there are still many places in the world where access to education is scarcely available and only afforded to a small wealthy minority, university programmes are not often referred to when dis-cussing the economics of “luxury” goods. In contrast, there are many parts of the world in which it is possible to obtain quality university education for little to no cost. Despite the availability of less costly choices, there is still a desire by many to purchase more expensive study options as a signal of wealth and status. This topic is investigated in this paper; assessing whether undergraduate study programmes at the top universities in America can be considered an example of a “luxury good”. In order to test this hypothesis, Veblen’s theory of “conspicuous consumption” is used to understand the trends in demand over the period from 1998 to 2019. The results show that these study programmes can be considered “luxury” goods, al-though there is no additional Veblen effect for Ivy League colleges. A suggestion for further research would be to increase the sample size.
Dedication
This thesis is dedicated to the wonderful Evelyn Moseley. Thank you for your patience, your wisdom, and your kindness. You have guided me throughout every challenge and reminded me to take each day as it comes.
Acknowledgements
I would like to express my sincere appreciation for everyone who has been a support to me throughout my academic journey thus far. Most importantly, Tracey Francis, who has been my sounding board throughout this whole process. I could not have done this without you.
A special thank you goes to my thesis supervisor, Konstantinos Ioannidis. I am incredibly grateful for all of your encouragement and feedback.
Contents
1 Introduction 6 2 Literature Review 9 3 Research Methodology 12 3.1 Data Collection . . . 12 3.2 Regression Model . . . 13 3.2.1 Statistical Model . . . 13 3.2.2 Hypotheses . . . 13 3.3 Research Method . . . 14 3.3.1 Research Variables . . . 14 3.3.2 Research Procedure . . . 15 4 Results 16 4.1 Panel Regression Validity . . . 164.1.1 Pre-regression Assumptions . . . 16 4.1.2 Post-regression Assumptions . . . 17 4.2 Regression Results . . . 17 5 Conclusion 19 5.1 Discussion . . . 19 5.2 Limitations . . . 20 5.3 Further Research . . . 20 A Appendix 24 A.1 Validity Tests . . . 24
A.1.1 Time Fixed Effects Test . . . 24
A.1.2 Hausman Test . . . 25
A.1.3 Breusch-Pagan Lagrange Multiplier Test . . . 25
A.2 Pre-regression Assumptions . . . 25
A.2.1 Linearity Test . . . 25
A.3 Post-regression Assumptions . . . 26
A.3.1 Modified Jarque Bera Test . . . 26
A.3.2 Wooldridge Test . . . 27
A.3.3 Modified Wald Test . . . 27
Section 1
Introduction
The concept of conspicuous consumption and Veblen goods has inspired many papers in Economics. The traditional law of supply and demand assumes that, ceteris paribus, demand for a good will decline as the price of said good increases. In contrast, the good will be more readily supplied when it can be sold at a higher price. An equilibrium is reached at a price point that is both profitable for suppliers and a reasonable opportunity cost for consumers (Rothbard, 2012). In reality, there are special cases of goods that display demand trends that do not fit this expectation. Thorstein Veblen identified the opposite pattern of spending amongst the upper class (and those aspiring to be so), who consume high cost goods to signal their wealth and status; he termed this “conspicuous consumption” (Veblen, 2009). A product that is consumed conspicuously, such that its demand rises in response to an increase in its price, is considered a Veblen good. An example of this is the iPhone (Kochaniak, 2016). Unlike Giffen goods, which are inferior, Veblen goods are usually coveted for their high quality. Furthermore, a luxury good can be characterised as a product that defies rational expectations, meaning that consumer demand grows more rapidly than consumer income. This contributes to the perceived unattainability of the good in comparison to lower cost alternatives and increases its perceived value in society.
In past research the Veblen paradox of the relationship between price and demand has been tested on more materialistic, physical goods (Mortelmans, 2005; Stepie´n, 2018). The Veblen effect of price increases on the demand for iPhones has been investigated (Kochaniak, 2016), which offers some support for Veblen’s hypothesis on consumer behaviour (Veblen, 2009). Several articles have been written criticising the value placed on designer handbags (Unger, 2016), such as the Herm`es Birkin (Times, 2008), which can be sold for over a hundred thousand dollars apiece. However, the existing literature scarcely reviews this concept in relation to non-physical goods. More specifically, there is limited literature and research focusing on the abnormal relationship between price and demand regarding higher education as a consumer good.
It can be argued that the existing literature scarcely reviews this concept in re-lation to demand for non-physical goods because it is more complex to assess. The demand for physical goods is usually measured directly by reviewing the producer’s financial statements and sales figures. In contrast, due to the limited research on this topic in relation to higher education, there is not one unanimously agreed standard for measuring consumer demand. Unlike most physical goods, for which markets
are clearing, supply for higher education is not equal to demand. In fact, many universities have strict admission requirements and tend to limit the number of admitted students in order to keep their acceptance rates within a certain thresh-old. Because of this, the number of full-time, freshman, undergraduate applications received annually for each university is a better representative of the demand for higher education. This total represents the number of students willing to purchase the good at a given price - the annual tuition fee. In relation to this, supply is a fraction of demand, and can be measured by the number of students that the university is willing to admit each year.
Consumption is the largest driver of American GDP (Mankiw, 2018), with a contribution of approximately 70% in 2019 (CEIC, 2020). In addition, American universities often make up a large proportion of the top-ranked universities in the world. For both of these reasons, American higher education institutes will be used to test the hypotheses of this paper. According to the United States Census Bureau, 18.9 million Americans were enrolled in university in 2019 (Alonzo, 2019). Obtaining a diploma has become a more common path chosen by students approaching the end of their high school studies as opposed to entering the labour market directly. Many students apply for college in the hopes of gaining knowledge, developing skills and increasing their future employability. The market for higher education is extremely varied, with options ranging from $0 to $60,000 (Lebowitz, 2015) per year and international study programmes offered across the globe. Despite the availability of “free education” options, many are still drawn to the more expensive, prestigious Ivy League institutions. The Ivy League is composed of the eight most elite universities in America: Harvard University, Yale University, Princeton University, Columbia University, Brown University, Dartmouth College, the University of Pennsylvania and Cornell University. This group initially gained popularity due to their strong athletic performance, but profited from this interest by stimulating demand for their academic faculties. They used a combination of marketing tactics, such as setting more selective admission requirements and steeper tuition fees. This gave the impression of exclusivity, and demand for these schools rose as they were perceived as less attainable than the average community college.
On one hand, it can be argued that the demand and price for these options are greater due to their intrinsic value based on the knowledge, quality, networks and career possibilities that they provide their students. On the other hand, those benefits could also be acquired through other means, such as independent studies, work experience or study programmes at other high-performing universities. The Massachusetts Institute of Technology and Stanford University are both examples of highly ranked, non-Ivy League colleges. It is possible that the additional attributes mentioned may simply reflect the perceived value that the rest of society places on those in possession of an Ivy League diploma. Therefore, supporting the definition of a Veblen good as something that is purchased to signal status to others. This debate is part of the foundation for the analysis in this paper.
The aim of this paper is to further develop upon existing literature by investi-gating the relationship between price and demand, specifically income elasticity of demand, in terms of university education. More specifically, this study will attempt to assess whether demand for freshman undergraduate programmes is positively cor-related with price. Additionally, an analysis will be performed to understand the Veblen effects of Ivy League status on the consumer demand level for education.
In order to assess this, a sample was selected of four Ivy League and four high-performing non-Ivy League universities. Data was collected on the annual tuition fees and number of applications received for each university over the period of 1998 to 2019. In addition to this, US annual GDP per capita was also sampled. These are used as proxies for price, demand, and aggregate household income, respectively. This data will be used to conduct an empirical study. Firstly, the existing theories and studies on consumption of luxury goods will be evaluated in the literature review. Subsequently, the methodology used for data collection and addressing the research question will be discussed. Following this, the results from the testing will be presented and elaborated on. Based on these findings, the conclusion will provide an answer to the research question, as well as proposals for further research.
Section 2
Literature Review
There is an array of research available on the topic of conspicuous consumption and Veblen goods, although most papers are written with a focus on more mate-rial products, such as designer handbags, smartphones, and sports cars. Thorstein Veblen’s theory will be discussed further in this section, as well as the findings of similar subsequent studies. The existing research suggests that the Veblen paradox does exist to some extent when looking at sales trends for physical goods. However, there is limited research investigating the Veblen effects of service goods, such as higher education.
Throughout the early 1900’s, student intake declined as university education became more vocationalised and “harnessed to [the] economic and nationalist agen-das” (Gould, 2003, p. 172) of the United States government during the war. In order to combat this falling demand, universities “set about convincing Americans that a college education. . . was a practical necessity for members and would be members of the managerial class” (Engell and Dangerfield, 2005, p. 57). In his research paper on Universities in the Marketplace, Bok infers that higher education has become more commercialised in recent years, as universities have seized the op-portunity to profit off of the desire for knowledge in an increasingly competitive, technology-driven society (Bok, 2003). Colleges have decided to focus more heavily on marketing “knowledge as a saleable product” (Newfield, 2003, pp. 39-40) and less as an essential human right.
Despite what some have criticised as an abuse of the education system (Veblen and Teichgraeber, 2015), Engell and Dangerfield discovered that the pressure to compete in the “ever-shifting landscape of the contemporary job market” (Engell and Dangerfield, 2005, p. 9) has resulted in higher demand for undergraduate majors. This is apparent in the 50% increase in enrolment between 1970 and 2000 (Knoedler, 2015). More specifically, Knoedler emphasises that the catalyst for this growth is the expected potential increase in employability. She highlights that “nearly two thirds of all undergraduate majors [choose] career-oriented subjects” (Knoedler, 2015, p. 336) and that interest in the arts and social sciences is swiftly declining. Engell and Dangerfield found support for this in the feedback from first-year students citing financial wealth as their main motivation for their choice of studies (Engell and Dangerfield, 2005).
In The Theory of the Leisure Class, Veblen introduces the term “conspicuous consumption” to describe the “vicarious consumption of goods” exhibited by the working class in order to portray an air of “pecuniary strength” (Veblen, 2009, p.
33). He argues that a consumer’s “utility” can be heightened by their ability to present themselves as “wealthy”. Veblen suggests that, rationally speaking, the working class should only consume necessities required for survival. Therefore “lux-ury” purchases should be reserved solely for the “leisure class”. To some extent the consumption of higher value goods is driven by intrinsic benefits, such as enhanced “comfort and wellbeing” (Veblen, 2009, pp. 35-36). However, Veblen also observed an ulterior motive for this level of indulgence: the need to present oneself as of equal status to one’s wealthier counterparts. One signal of affluence is the ability of the consumer to distinguish between high and low value goods.
Moreover, when applied to education, even those from low income households may choose to attend more prestigious universities. This is because they believe that they benefit from “acquiring” status by integrating into the communities of the upper class (Veblen, 2009, p. 38). To some extent, the competitive nature of the human psyche is evident as the working class “struggle to outdo one another” by assimilating with the elites. Although bragging rights may be one incentive for the demand for Ivy League schools, Veblen does not dispute that part of the utility received from this good is derived from the core intrinsic benefits of studying.
As an extension of his earlier work, in 1918, Veblen utilised his theories on cap-italism and consumer behaviour to critique the American higher education system. He refers to “higher learning” as “the current body of science and scholarship”, but makes the following criticism: “through indoctrination with utilitarian ideals of earning and spending, as well as by engendering spendthrift and sportsmanlike habits, such a business-like management diverts the undergraduate students from going in for the disinterested pursuit of knowledge, and so from entering on what is properly university work” (Veblen, 2005, p. 66). Veblen highlights the issue that university education was becoming increasingly more driven by commercial gain, and less by its principal value of transferring knowledge. This is not only apparent in the pricing strategies of these businesses, but also in the quality of the content being taught to their students. He argues that “business proficiency is put in the place of learning” (Veblen, 2005, p. 142), which is a criticism that can be applied to many of the Ivy League business schools. The colleges boast of alumni including some of the top business people and politicians in the world, and their “corporate connection” (Lucas, 1994, p. 255) with leading organisations who recruit directly from their campuses. Many students may be more attracted to the increased em-ployability that a degree from Harvard or Yale will offer them as opposed to the intrinsic benefits mentioned. To summarise, “the hand of business control (domi-nates) practically every aspect of the modern university” (Lucas, 1994, p. 200).
Nevertheless, as mentioned in Going to College on My iPhone, if the purpose of attending university is obtaining information (even if solely for business motives), there are “new and cheaper ways” for students to do this (Knoedler, 2015, p. 337). As other industries adapt to the technological advancements of modern society, ed-ucation is becoming more accessible with options for online universities and virtual courses. These options are in some cases more affordable and convenient than “tra-ditional” in-person tuition. However, it does not yet appear that this has had a significant substitution effect on the demand for elite Ivy League universities. In response to the growing number of “online non-profit universities (that provide the same courses for free)” (Knoedler, 2015, p. 341), some of the Ivy League have responded by offering their own Massive Open Online Courses (MOOCs) and
dig-ital programmes. Although, many of these remote options are charged at a higher price than the face-to-face courses, and the MOOCs are used as a marketing tool to attract students to the traditional courses on campus.
In 2016, Kochaniak selected the iPhone to study the “Veblen law mechanisms” (Kochaniak, 2016, p. 3). She chose the iPhone for her analysis because she believed that it matched the basic criteria of a Veblen good. She argued that the increasingly high prices of iPhones in comparison to other alternatives was economically irrational based on the level of homogeneity in the smartphone market. She claimed that the various “models offered by manufacturers do not differ significantly” (Kochaniak, 2016, p. 1) in terms of features and uses. Therefore, she assumed that demand for this good is driven by the perception of “the brand’s logo”, which its consumers believe to offer them a “special status” in the way they are regarded by society (Kochaniak, 2016, p. 2). This perspective can similarly be applied to Ivy League universities when reviewing the market for higher education. In theory, if the pur-pose of a university is purely to provide knowledge to students, then there is little differentiation between the options on the market. Based on Kochaniak’s logic, ra-tional consumers should substitute the more expensive Ivy League universities for cheaper alternatives, but this is not necessarily the case. Irrespective of the high cost for these options relative to their market averages, there still appears to be a high level of demand.
Kochaniak found a “positive correlation” between the growth in iPhone demand and price, “which goes against the law of supply and demand” and led her to believe that “Veblen characteristics” exist in the smartphone market (Kochaniak, 2016, pp. 4-5). Based on the significant positive results found for price elasticity of demand from 2010 to 2014, she concludes that there is sufficient evidence to “consider the iPhone as an example of a luxury good”.
Although it is not the focus of this paper, the effects of the 2008-2009 Great Recession on demand for higher education cannot be ignored. The impact of this shock on the education market was studied in great depth (Long, 2014). On the one hand, the decline in income during this period in combination with rising tuition fees should, in theory, have a negative impact on applications. However, the increase in unemployment during this period provided motivation for young people to obtain additional qualifications to improve their employability. Although, many chose part-time study programs to balance alongside other work commitments. Long found that these effects differed between the various regions and demographics within America, but concluded that on average the crisis had a positive effect on university demand. The literature discussed generally summarises the development of higher edu-cation as a “money-making” (Bok, 2003, p. 15) tool for institutions and, to some extent, explains the psychology behind the increasing demand for undergraduate degree programmes. However, this paper will use a detailed quantitative analysis to assess these trends explicitly within the market of education and Ivy League universities.
Section 3
Research Methodology
3.1
Data Collection
In order to answer the research question, this paper will first focus on the trend in the demand for higher education in a quantifiable form: the total number of freshman applications for undergraduate programs at the sampled universities. The quantity of applications will be used to represent the sales volume, as this symbolises the demand for the product more accurately than the number of students admitted to the study programme. The number of admissions is restricted by the supplier, which in this case is the respective university. The price of the product is represented by the annual tuition fee, excluding administration costs. This information will be combined to determine whether education can be considered a luxury and if a Veblen effect is present in its demand. The study will examine sales over a given time period, from 1998 to 2019. The data will be gathered from multiple sources and collated into one database for the purpose of this study. To test these hypotheses and determine whether education is indeed an example of a luxury, Veblen good, the annual sales price and volume must be analysed in combination with the average household income of American consumers.
American colleges are not legally required to publicly disclose their admissions data, which means sourcing historical data can prove challenging in many cases. Due to these limitations, only four of the Ivy League universities were sampled for this paper: Cornell University, Dartmouth College, Princeton University and Yale University. In order to have a fair and balanced study, four non-Ivy League univer-sities were also selected: Northwestern University, Stanford University, University of Michigan and the University of California, Berkeley. The latter four selections were made due their top-twenty placements in most rankings for the best colleges in America. These universities are all of comparable quality to the Ivy League when considering the following metrics: academic reputation, employer reputation, faculty/student ratio, citations per faculty, international faculty and ratio of inter-national students. For this reason, these universities are used as a control when testing the third hypothesis - to determine the Veblen effect of Ivy League status.
The data was obtained from annual reports and commission data published in the universities’ archives, articles, government reports and directly from the univer-sities’ admissions and registrar offices. In order to determine the correlation between price and demand, the annual tuition fees and applications for freshman (first-time, first-year) undergraduate programmes has been collected. Next, to test the income
elasticity of demand, growth in tuition fees will be compared to the changes in aver-age household income. As this research is focused on American universities, annual GDP per capita in the USA has been used as a proxy for average household in-come. These figures were sourced from statista.com. The following variables are measured: year (academic year), apls (applicants), fee (tuition fee), uni (univer-sity), hinc (household income), crisis (1=before 2018, 0=after 2018) and ivy (1=Ivy, 0=non-Ivy).
3.2
Regression Model
3.2.1
Statistical Model
logaplsi,j = βi,j,0+β1loghinci,j + β2logf eei,j + β3ivyi,j + β4crisisi,j + i,j
The variables used in this model are explained in further detail in section 3.3.1. A panel data set is used in this study, therefore the number of applications received is measured for each university over each academic year.
3.2.2
Hypotheses
For the purpose of examining the relationship between price and demand for edu-cation, following the theories discussed in the literature review, this paper will test three hypotheses.
Hypothesis 1: Education is a luxury good.
Elastic demand is characterised by significant increases or decreases in demand rel-ative to incremental changes in income. The theory of income elasticity of demand states that a positive income elasticity of demand is standard for most normal goods. If the measure for responsiveness of demand to a change in consumer income is less than one, but greater than zero, education can be considered a “necessity”. Items that are “necessities”, as opposed to “luxuries” tend to express relatively inelastic demand. Based on the definition of a luxury good, it is expected that income elas-ticity of demand with respect to income is greater than 1.
Null hypothesis: H0 : β1 = 0
Alternative Hypothesis: H1 : β1 > 0
The null hypothesis implies that education is a sticky or inferior good with an inelastic response to changes in income based on the model parameters. In contrast, the alternative hypothesis states that a significant positive linear relationship exists between income and demand, such that education is a normal good. A significance level of 5% will be used to test these hypotheses.
Hypothesis 2: Education is a Veblen good.
In order to measure the Veblen effect of higher education, the relationship between tuition price increases and the growth in the number of university applications will
be measured and tested to identify the existence of a significant correlation. If a Veblen effect is present, it is expected that there is a positive correlation between price and demand. This hypothesis is based upon the Veblen Effects in a Theory of Conspicuous Consumption (Bagwell and Bernheim, 1996). When combining the existing literature with the visible growth in university applications and prices over the years, it is expected that higher education will meet the criteria of a Veblen good.
Null hypothesis: H0 : β2 = 0
Alternative Hypothesis: H1 : β2 > 0
The null hypothesis suggests that there is insignificant evidence to infer that there is a positive linear relationship between price and demand for education. However, the alternative hypothesis suggests that a significant positive linear relationship exists between income and price, implying that university education is in fact a Veblen good. These hypotheses will both be tested at a 5% significance level.
Hypothesis 3: The Veblen effect on demand for education increases for Ivy League colleges.
The Ivy League affiliation can be considered a desirable status symbol for many prospective students. As discussed in the second section, this paper predicts that trends in demand for Ivy League universities will exhibit stronger Veblen tendencies in comparison to other competitors in the market.
Null hypothesis: H0 : β3 = 0
Alternative Hypothesis: H1 : β3 > 0
The null hypothesis implies that there is insignificant evidence to suggest that there is a stronger positive linear relationship between price and demand for Ivy League universities in comparison to non-Ivy League schools. On the other hand, the alternative hypothesis states that there is significant proof that growth in consumer demand is driven by the pursuit of Ivy League status, ceteris paribus. Both of these hypotheses will be tested at a 5% significance level.
3.3
Research Method
3.3.1
Research Variables
In order to answer this research question, a panel regression model was used. The academic year, year, is measured by the time parameter, and university, uni, has been specified as the panel identification variable. Each university has been ran-domly allocated a number from 1 to 8 for the purpose of testing via multiple dimen-sion analysis.
The dependent variable, logapls, measures the number of applications received for each university over time. This is representative of the changes in demand. In addition to this, there are four independent variables included in the model. One
example being logfee, representing the tuition fee charged for freshman programs at each university over the years. As price is a key element of supply and demand theory this variable has been included to test hypothesis one. In order to determine whether education is a luxury good, it is important to understand the relationship between price and demand. Similarly, understanding the response of demand to changes in income is essential for discerning consumer behaviour and the presence of a Veblen effect. Hence why household income, loghinc, was included as an inde-pendent variable. Furthermore, to distinguish whether these effects are specific to Ivy League universities, ivy was included as a dummy variable in the model. This variable is useful for testing the third hypothesis.
Finally, the last independent variable included is the dummy variable crisis. Long identified that the economic crisis of 2008-2009 had a broad impact on the higher education market (Long, 2014, pp. 229-231). The overall effect of the crisis on university demand is varied. Demand increased due to higher levels of unemploy-ment, but simultaneously decreased due to the economic strain on households. The impact was found to vary between different demographics and course types. This relationship is not the focus of this paper, and will not be evaluated in great detail. Nevertheless, the variable has been included to account for any shocks in demand as a result of the Great Recession that cannot be explained by the other variables in this model.
3.3.2
Research Procedure
A panel regression is utilised in an attempt to answer the research question posed. This paper will first focus on the behaviour of demand for education in relation to annual price changes. Once the price elasticity of demand is understood, this will be used to determine whether university education can be considered a luxury good. Following this, the regression results will be analysed to ascertain which variables have a significant influence on demand for higher education. More specifically, to evaluate if a Veblen effect is present; and whether that effect is amplified for Ivy League colleges.
The first step in determining the most suitable regression method was to first perform a Wald test, as seen in Appendix A.1.1. At a significance level of 5%, there is insufficient evidence to reject the null hypothesis that the coefficients of all years are jointly equal to zero. For this reason, time fixed effects are excluded from the model. Subsequently, a Hausman test was performed to determine whether a fixed or random effect model is most appropriate. There was insufficient evidence to reject the null hypothesis. Therefore, it cannot be concluded that the difference in coefficients is systematic, so a random effects model is used. Lastly, the Breusch-Pagan Lagrange Multiplier test was used to decide between a simple OLS regression and a random effects panel regression. The null hypothesis states that there is no panel effect and as such the variance across universities is zero. Based on Appendix A.1.3, the null hypothesis can be rejected. Therefore, there is sufficient evidence to suggest that significant differences exist across universities, so a random effects panel regression was chosen.
Section 4
Results
4.1
Panel Regression Validity
4.1.1
Pre-regression Assumptions
There are several assumptions that need to be verified before a generalised least square regression is performed; the first being the absence of outliers. Figure A.2 shows that there are no irregularly small or large observations in the sample that would indicate measurement error. Based on this, no observations have been ex-cluded from the sample.
The second assumption is that a linear relationship exists between the dependent variable and independent variables, thus reducing the possibility of functional form misspecification. The graph matrix in Figure A.1 shows that there is a positive linear relationship between logfee and loghinc in relation to logapls, respectively. The dummy variables, ivy and crisis, were not included as they are binary. Overall, the assumption of linearity holds.
Finally, it is assumed that no perfect multicollinearity exists between the ex-planatory variables. A high level of multicollinearity leads to variance inflation and causes a downward bias on the statistical significance of the variables. A correlation matrix of the variables included in the model was created to identify any multi-collinearity. A correlation value of 0.80 or higher is generally accepted as a sign of the existence of multicollinearity. The values found are presented in table 4.1 below.
logapls loghinc logfee ivy crisis logapls 1.0000
loghinc 0.6875 1.0000
logfee 0.4247 0.9046 1.0000
ivy -0.4119 0.0000 0.1660 1.0000
crisis -0.6473 -0.8202 -0.7900 -0.0000 1.0000
Table 4.1: Correlation Matrix
The high level of multicollinearity found between household income and tuition fees can be partially explained by increasing inflation rates over time. If households become wealthier on average, it is expected that they have more disposable income. As such, businesses maximise their profits by increasing their prices. Despite the interaction between the two variables, neither were excluded from the model as they
both are very relevant in understanding the elasticity of demand when answering the research question.
The negative correlation between the crisis and logfee variables suggests that tuition fees were generally lower before the financial crisis, and positive thereafter. It is expected that prices increase annually in line with inflation, so the overall trend is not surprising. What is important, for the purpose of this study, is to understand the impact of the crisis on demand in spite of the increasing prices. Vice versa, in order to understand any fluctuations in price elasticity of demand over time it is vital to account for any shocks in the model. The matrix also shows that average household income was lower before the crisis. Looking at the data on GDP per capita, this makes sense. At the peak of the crisis, GDP per capita fell. However, this has increased above the pre-crisis level as the economy has gradually recovered, so this is not unexpected. By nature of the crisis, it is expected that there would be a strong relationship between these variables. This relationship does not negate the necessity of either of these variables in the model.
Based on the correlation matrix, it is clear that some multicollinearity exists. Nevertheless, there is insufficient evidence to remove any of the parameters included in the model as they are all relevant in deriving an accurate conclusion to the research question.
4.1.2
Post-regression Assumptions
Subsequent to the running of the regression, there are a number of other assumptions that need to be tested. Firstly, a modified Jarque Bera test is used to examine the normality of the residuals in the random effect model. This test performs a joint analysis on the normality of random errors, e, and the normality of total errors, u, for the panel regression. The results in Table A.1 show that there is insufficient evidence to reject the null hypothesis that total errors are normally distributed. In contrast, at a 5% significance level, there is evidence to reject the null hypothesis for the normal distribution of random errors. Nevertheless, as the null hypothesis was not rejected for total errors, no adjustments were made to the regression.
For the purpose of measuring the heteroskedasticity of the variables in the model, the Modified Wald test was used as no other appropriate tests were found for random effect panel data. Because of this, the data is treated as fixed effect for testing purposes. At a 5% significance level there is sufficient statistical proof to infer that heteroskedastic errors are present for all variables other than logfee.
In order to verify the final condition of the absence of autocorrelation in the panel data, a Wooldridge test was executed. There is sufficient evidence to reject the null hypothesis, and therefore the assumption of no autocorrelation. Based on the post-regression test results, clustered standard errors were included in the post-regression to counter the effects of autocorrelation.
4.2
Regression Results
logapls Coef. St.Err. t-value p-value [95% Conf. Interval] Sig. loghinc 1.225 0.289 4.24 0.000 0.659 1.792 *** logfee -0.026 0.174 -0.15 0.881 -0.367 0.314 ivy -0.395 0.197 -2.01 0.045 -0.781 -0.009 ** crisis -0.250 0.025 -9.94 0.000 -0.299 -0.200 *** constant -2.406 1.316 -1.83 0.067 -4.985 0.172
Mean dependent var 10.194 SD dependent var 0.484 Overall r-squared 0.668 Number of obs 176.000 Chi-square 3301.54 Prob >chi2 0.000 R-squared within 0.936 R-squared between 0.370 *** p<0.01, ** p<0.05, * p<0.1
Table 4.2: Regression Results
As predicted the absolute value elasticity of demand with respect to income is greater than one. This implies that demand for higher education is elastic and sensitive to changes in household income. More specifically, the linear effect is estimated to be positive. This implies that consumers are willing to spend more than they receive in additional income on education. One possible explanation for this is that many financially-able families set up trust funds or savings accounts for their children at birth in preparation for future college expenses. Alternatively, many families also apply for student loans and financial aid to contribute to the high cost of attending university. These reasons could partially explain why consumer demand grows at a rate faster than income. Overall, this result is significant at a 5% significance level. Therefore, there is sufficient evidence to suggest that education is a luxury good.
In order to identify the existence of a Veblen effect in the demand for education, a positive correlation between price and demand was expected. The results of the regression show that there is in fact a negative linear relationship between price and demand. Therefore, the null hypothesis cannot be rejected. Moreover, the result is not significant at a 5% significance level. Hence why it is not possible to determine that education is a Veblen good based on the data tested.
Surprisingly, the regression does not show that there is a positive linear relation-ship between Ivy League status and demand for undergraduate places. This result is significant at the 5% significance level. Therefore, there is insufficient evidence to reject the null hypothesis. In other words, the Ivy League affiliation does not stimulate demand for university admissions and it cannot be concluded that Ivy League colleges are an example of Veblen goods.
Section 5
Conclusion
This paper studies the relationship between price and demand for freshman under-graduate programmes at American universities. Annual tuition fees are used as a proxy for price and the number of applications received are used to represent the level of demand for each college. The research sample is composed of the price and demand for eight universities over the period from 1998 to 2019. In this section, the research findings and limitations of the study will be discussed. Finally, suggestions for further research will be proposed.
5.1
Discussion
The purpose of this paper was to understand the behavioural economics behind the demand for university education. The research develops upon the existing literature by combining different theories and concepts to generate a unique research question. The aim of this paper was to determine whether university education could be considered a luxury good. Moreover, the purpose was to examine whether it could be concluded that education meets the criteria of a Veblen good. Although there is limited research on this specific topic, certain predictions were made based on the findings of similar investigations. The first expectation was that education is indeed a luxury good. Secondly, it was predicted that freshman study programs could be considered Veblen goods. Additionally, a prediction was made that the Veblen effects would be stronger for Ivy League universities due to their elite status in society.
After evaluating the necessity of the explanatory variables included in the statis-tical model, the hypotheses were tested using a panel regression. The results showed that income has a significant positive effect on demand and confirmed that under-graduate study programs can be considered luxury goods. However, the findings were not able to support the second and third hypotheses to prove the existence of a Veblen effect on the demand for education. The effect of Ivy League status on demand was unexpected, in that the results show that demand for these universities is lower than non-Ivy League colleges, ceteris paribus. This does not support the idea that Ivy League courses are Veblen goods within the education market. Based on this result, the recommendation would be for Ivy League schools to charge a fee in line with other top universities in the country to avoid a decline in demand due to the substitution effect.
were similar for both Ivy League and non-Ivy League universities throughout the period studied and grew at a relatively constant rate over time. The one exception identified is the University of California, Berkeley, which consistently charges a fee slightly below the average over time and is more volatile with its pricing when compared to the other universities sampled. Following this, appendix item A.4 also shows that demand for this university was slightly higher than the average of the other universities sampled. It is clear that there is a correlation between the lower price of this option and its higher demand. Nevertheless, further testing is required to determine if any causal relationship exists.
5.2
Limitations
There are some limitations of this study that can be used to develop further research. The first being the lack of publicly available university admissions data. In order to improve internal validity, it would be useful to increase the sample size by obtaining admissions figures for more universities and over longer periods of time. There are increasing initiatives within the education sector to improve the disclosure and availability of data within the market. This has improved in recent years, however finding accurate, consistent and reliable data for most universities before 1998 is still a challenge. Likewise, the availability of data is limited to the universities who are willing to participate in improving transparency within the market. There are several universities that opt out of publicly disclosing more data than is legally required.
Another limitation of this study is potential omitted variable bias. One factor that is not included in the statistical model is the university reputation or exclusivity. There are many non-Ivy League schools that have comparable reputations to the Ivy League and this can be observed in their very low admissions rates. Despite this, this variable was not included in the model due to the simultaneous causality between admissions rates and demand. If the non-Ivy League colleges sampled in this study are found to have an equal or better reputation than the Ivy League, this could explain the rejection of the third hypothesis.
5.3
Further Research
One suggestion for future research would be to increase the sample size and scope, therefore including mid and low performing universities in the selection. This would help to identify whether the findings can be generalised over a larger population or are specific to the top-ranking schools in America. Following this reasoning, defining and including a variable that measures university reputation/exclusivity would be interesting for making inferences about the psychological influences on demand.
Many of the non-Ivy League universities in this sample benefit, not only from their high quality, but also their recognition as household names. It would be useful to test the conclusions on other high-performing universities with less brand recogni-tion, such as Babson College - a business school with strong academic performance. The average acceptance rate for Ivy League universities in 2019 was 8.6% (Coach, 2019). In contrast, Babson admitted 26% of its applicants. The average salary six years after graduation from Babson is approximately equal to that of Yale at
$75,000 per year. However, demand for Babson represents a fraction of only 17% of the number of applications received for Yale (NCES, 2020). It would be useful to compare the trends in demand against the findings in this paper. In theory, if consumers are rational they should favour high quality universities that are more affordable and with higher acceptance rates. This would help to further understand the value placed on reputation and Ivy League status.
Furthermore, a qualitative analysis could be performed in the form of surveys and interviews for a random selection of university applicants and hiring managers. This information could be used to better understand the behavioural economics behind employers’ demand for Ivy League alumni, as well as the rationale behind the demand of the students who pursue different study programmes. Taking into consideration Engell and Dangerfield’s (2005) findings about the shift in demand away from liberal arts and science majors toward business studies; the research from this paper can be further developed by adding regressors for each undergraduate major to the model and testing whether the Veblen effect is specific to certain specialisations. There are several examples of successful billionaires and business owners who have not completed any form of higher education (Lebowitz, 2015). Additional factors may be discovered that are relevant parameters for demand that could improve the model.
Finally, if the motivation for pursuing education is to improve social standing by assimilating with upper class peers; the success of this strategy should be evaluated. Consumption choices do not automatically transfer the consumer to a status that they do not originally “belong”, and therefore may be considered wasteful (Veblen, 2009, p. 40). Defining and assessing the success of this strategy would make an interesting research topic.
Bibliography
Alonzo, Frances (2019). Census Bureau Reports Nearly 77 Million Students Enrolled in U.S. Schools. url: https://www.census.gov/newsroom/press-releases/ 2019/school-enrollment.html.
Bagwell, Laurie Simon and B. Douglas Bernheim (1996). “Veblen Effects in a Theory of Conspicuous Consumption”. In: The American Economic Review 86.3, p. 349. doi: https://www.jstor.org/stable/2118201?seq=1.
Bok, Derek (2003). Universities in the Marketplace: The Commercialization of Higher Education. Princeton, NJ: Princeton University Press.
CEIC (2020). United States Private Consumption: % of GDP. url: https://www. ceicdata.com/en/indicator/united- states/private- consumption-- of-nominal-gdp.
Coach, Ivy (2019). Ivy League Admissions Statistics. url: https://www.ivycoach. com/2019-ivy-league-admissions-statistics/from.
Engell, James and Anthony Dangerfield (2005). Saving Higher Education in the Age of Money. Charlottesville, VA: University of Virginia Press.
Gould, Eric (2003). The University in a Corporate Culture. New Haven, CT: Yale University Press.
Knoedler, Janet T. (2015). “Going to College on My iPhone”. In: Journal of Eco-nomic Issues 49.2, pp. 329–354. doi: https://www.tandfonline.com/doi/ abs/10.1080/00213624.2015.1042729?journalCode=mjei20.
Kochaniak, Ewelina (2016). “IPHONE AS AN EXAMPLE OF THE PRESENT LUXURY GOOD – ANALYSIS OF SALES AND PRICES IN 2010-2014”. In: Contemporary Economy 7.1, pp. 1–8. doi: https : / / doaj . org / article / b2c338262f5b4864ba4c16c24c147f05.
Lebowitz, Shana (2015). Why Going to the most expensive colleges might not be worth it. url: https : / / www . businessinsider . com . au / are expensive -colleges-worth-it-2015-12.
Long, Bridget (2014). “The Financial Crisis and College Enrollment: How Have Students and Their Families Responded?” In: How the Financial Crisis and Great Recession Affected Higher Education. National Bureau of Economic Research, Inc, pp. 209–233. url: https://EconPapers.repec.org/RePEc:nbr:nberch: 12862.
Lucas, Christopher J. (1994). American higher education : a history. New York, NY: St. Martin’s Press.
Mankiw, N. Gregory (2018). Macroeconomics (10 ed.) New York, NY: Worth Pub-lishers.
Mortelmans, Dimitri (2005). “Sign Values of Distinction, The Concept of Luxury”. In: Semiotica 2005.157, pp. 497–520. doi: https://doi.org/10.1515/semi. 2005.2005.157.1-4.497.
NCES (2020). College Navigator. url: https://nces.ed.gov/collegenavigator/ ?id=164580#outcome.
Newfield, Christopher (2003). Ivy and Industry: Business and the Making of the American University. Durham, NC: Duke University Press.
Rothbard, Murray N. (2012). “Competition and the economists”. In: The Quarterly Journal of Austrian Economics 15.4, pp. 396–409. doi: https://link.gale. com/apps/doc/A320733939/AONE?u=amst&sid=AONE&xid=10ce3c38.
Stepie´n, Beata (2018). “Snobbish Bandwagoners: Ambiguity of Luxury Goods’ Per-ception”. In: Journal of Management and Business Administration, Central Eu-rope 26.1, pp. 79–99. doi: https : / / cemj . eu / resources / html / article / details?id=168259&language=en.
Times, The New York (2008). The Birkin and Herm`es. url: https://www.nytimes. com/2008/03/02/style/02iht-birkin.10611850.html.
Unger, Brooke (2016). Demand Curve. url: https://www.1843magazine.com/ style/demand-curve.
Veblen, Thorstein (2005). The Higher Learning in America. New York, NY: Cosmo Classics.
— (2009). The Theory of The Leisure Class. Auckland, New Zealand: S.l. : The Floating Press.
Veblen, Thorstein and Richard F. Teichgraeber (2015). The Higher Learning in America: A Memorandum on the Conduct of Universities by Business Men, the Annotated Edition. Baltimore, MD: Johns Hopkins University Press.
Appendix
A.1
Validity Tests
A.1.1
Time Fixed Effects Test
testparm i.year ( 1) 1999.year = 0 ( 2) 2000.year = 0 ( 3) 2001.year = 0 ( 4) 2002.year = 0 ( 5) 2003.year = 0 ( 6) 2004.year = 0 ( 7) 2005.year = 0 ( 8) 2006.year = 0 ( 9) 2007.year = 0 (10) 2008.year = 0 (11) 2009.year = 0 (12) 2010.year = 0 (13) 2011.year = 0 (14) 2012.year = 0 (15) 2013.year = 0 (16) 2014.year = 0 (17) 2015.year = 0 (18) 2016.year = 0 (19) 2017.year = 0 (20) 2018.year = 0 F( 20, 147)=1.40 Prob >F =0.128
A.1.2
Hausman Test
Test: Ho: difference in coefficients not systematic
chi2(3) = (b − B)0[(Vb − VB)−1](b − B)
= −14.00 chi2 < 0
A.1.3
Breusch-Pagan Lagrange Multiplier Test
Test: Var(u) = 0 chi2 = (01) = 750.96 P rob > chi2 = 0.0000
A.2
Pre-regression Assumptions
A.2.1
Linearity Test
Figure A.2: Scatter Plot to Test for Outliers
A.3
Post-regression Assumptions
A.3.1
Modified Jarque Bera Test
Observed Bootstrap Normal-based Coef. Std. Err. z P>kzk [95% Conf. Interval] Skewness e -.0010328 .0004975 -2.08 0.038 -.0020078 -.0000578
Kurtosis e .0001688 .0001114 1.51 0.130 -.0000496 .0003872 Skewness u .0015814 .0027096 0.58 0.559 -.0037292 .0068921 Kurtosis u -.0001062 .0010885 -0.10 0.922 -.0022396 .0020273 Joint test for Normality on e: chi2(2) =6.61 Prob >chi2 = 0.0368 Joint test for Normality on u: chi2(2) =0.35 Prob >chi2 = 0.8394
A.3.2
Wooldridge Test
Wooldridge test for autocorrelation in panel data
H0: no first-order autocorrelation
F(1, 7)=30.054 Prob >F =0.0009
A.3.3
Modified Wald Test
Modified Wald test for groupwise heteroskedasticity in the residuals of a fixed effect regression model:
H0: σ(i)2 = σ2, for all i
chi2(4) = 2316.75 P rob > chi2 = 0.0000
logapls Coef. Std. Err. z P>kzk [95% Conf. Interval] logfee -.0260933 .114693 -0.23 0.820 -.2508875 .1987009 loghinc 1.225478 .1590244 7.71 0.000 .9137963 1.53716 ivy -.3950887 .1525151 -2.59 0.010 -.6940127 -.0961646 crisis -.249672 .0269672 -9.26 0.000 -.3025267 -.1968173 cons -2.406245 .7951786 -3.03 0.002 -3.964766 -.8477234 sigma u .21004272 sigma e .0919019
rho .83932017 (fraction of variance due to u i)
A.4
Graphical Analyses
Figure A.4: Number of Applicants Received for Each University Over Time
