1 Alex Morgano Ladji Bamba Lucas Van Cleef Computer Skills for Economic Analysis E226 11/6/2015 Dr. Myers Abstract This essay focuses on the causality between specific questions that deal with people s current economic conditions, which include factors such as their current thoughts on spending and saving, and their beliefs about how their economic condition will be in the future. Here we utilize notable research to dive into the causality of these variables. What we find in both the qualitative and quantitative portions of the literature and research conducted by Carbone and Hay (2004) is that people who are either financially well off, or feel they will be financially well off in the future have greater consumption expenditures than those who don t feel this way. Through the use of the income related questions in our own data, we ask the question, do those who are employed, and believe that they will be more financially better off in future periods, intend to consume more in the current period? What we find when we dive into this question is that first, our data is perhaps not at all representative in its scope nor free of outlying situations. And second, our data does not directly confer with the conclusions that the Carbone and Hay (2004) data provide.
2 The Effect of Employment Status on Consumption Behavior Of the diverse array of questions included in our survey, several focus on the respondents economic circumstances, including their employment status, current intentions regarding spending and saving, and their beliefs about their future economic condition. The responses to these questions are related in the sense that they all deal with the individual s income, with employment having a positive effect on income, and spending and saving being theoretically positively dependent on income. While these variables either influence or are influenced by income, the literature included in this analysis provides evidence to suggest that an individual s predictions about their future income interplays with their current employment status to either positively or negatively affect current consumption attitude. In this analysis we attempt to answer for our sample, do those who are employed, and believe that they will be more financially better off in future periods, intend to consume more in the current period? Literature Review In their experimental analysis, Carbone and Hey (2004) investigate the relationship between an individual s employment status and their consumption habits in the current and future periods. In particular they find evidence that many individuals, when employed, have a tendency to overconsume in the current period such that lifetime consumption is suboptimal. The model that they derive regarding an individual s consumption in the current period is C(t)=a+bW (P. 663) with C falling between 0 and W. The long form of this model as given by Carbone and Hey (2004, p. 663) is specified as such: In the model, the intercept a is positively affected by p, which is a proportion representing an individual s confidence that they will remain employed in future periods. Similarly q represents an individual s belief that they will find employment in a future period. Both p and q have a positive effect on consumption, and decrease in each successive t as an individual s expectations of future employment decrease with age. y represents an individual s income, which will have a positive effect on consumption, and e is a binomial variable which is 1 if the respondent is employed and a 0 if they are not. The slope b of C is solely dependant on r, which is the proportion of an individual s wealth that will be devoted to consumption. Carbone and Hey (2004) indicate that the values of a x and b x will increase in each successive period as the individual has less life ahead of them and thus less reason to save their wealth. Data While we do not have data for each variable included in Carbone and Hey (2004) s model for consumption, we do have access to variables that either work as proxies for relevant variables, or whose interactions with another variable embodies some of the same effects as a missing relevant variable. In terms of C, consumption, we will be using the variable large in the year0915 dataset for which a yes response indicated that the respondent would be making a large consumption (over 500$) in the
3 near future. For our purposes, a response of yes for Large indicates a pro-consumption attitude in the current period, and no indicates a consumption-averse attitude in the current period. To proxy for both P and Q, we included the interaction variable ConfidentlyEmployed between the previously existing emp and finbetter, which indicates not only whether or not the individual is employed, but also captures the effects of their predictions about their employment status in future periods. If a respondent is both employed, and believes that their future financial circumstances will either improve or remain the same, they are considered Confident. If they are employed but believe that they will be worse off in future periods, we consider them to be Unconfident. Finally, for those who are unemployed but believe they will have better financial luck in future periods we have included Hopeful to account for consumption based on the belief of higher future income. Finally, we include a second interaction variable ConsumptionPropensity which indicates how likely an individual is to spend more than they save. This is analogous to the r coefficient in Carbone and Hey (2004) s model. ConsumptionPropensity is considered high in cases where the respondent answers they they are either going to spend more and save less or equal, or if they are going to spend equal and save less than this time last year, medium if their response to variables spend and savings as the same, and low if the individual indicates that they will spend more and save less or equal, or spend equally and save less than this time last year. We expect the employed to have a higher consumption propensity than the unemployed, the confident than the employed, and the old than the young based off of the findings of Carbone and Hey (2004). Table A-1: ConfidentlyEmployed by Large Results While we predicted that confidently employed individuals would have the largest demand for consumption, the frequency table between ConfidentlyEmployed and Large reveals that within our survey, the portion of the group Confident who expect to make a large purchase in the current period is extremely similar to the Unemployed, and slightly less than the Unconfident. Those who were Hopeful, Unconfident, and Out of LF were more likely to outright answer that they would not be making a large purchase than Confident as expected. Still, The unemployed group, who does not believe that their income will be higher in future periods, remains the least prone to outright reject the notion of making a large purchase in the current period.
4 Table A-2: ConfidentlyEmployed by ConsumptionPropensity Counterintuitive to our hypothesis, the group Confident from our sample does not have the highest percentage of respondents with high consumption propensity. Rather, they have the lowest with the group Unconfident following them with a similarly low number of High responses. All groups other than the Unemployed have a majority of respondents falling under a Medium propensity to consume, whereas the Unemployed are most prone to Low consumption propensity. Table A-3: Age by Consumption Propensity Finally, we compared age groups with spending propensity in order to see if Carbone and Hey s (2004) theory that older individuals have a lower propensity to consume than younger ones. This is not true of our sample in which those aged 18-24 actually had the highest occurrence of High consumption propensity. The oldest two groups actually had more than half of their members report a medium consumption propensity. Ultimately the responses included in our survey do not reinforce our hypothesis that the confidently employed are the most prone to consumption of the groups represented in our analysis. We also find that amongst our sample, most respondents in general were
5 within the Medium consumption propensity group, with age not having the positive effect on the incidences of High responses as we had predicted. Our results would actually indicate that the unemployed, who do not see their personal incomes increasing in future periods, are actually the group with the most consumption planned within our sample. It is worth noting that our sample is mostly comprised of those within the 18-24 age group was taken by college students, so it is likely that many respondents were students as well. Because of this, employment status may be less indicative of consumption behavior, as many students rely not on their own employment and income, but rather that of their parents. Limitations and Conclusion As previously mentioned we are likely inherently limited in our analysis due to sampling biases. This is true because the data was collected by students for whom it may have been most convenient to include the large number of fellow students that they would come into contact with on a regular basis. If this is true the responses would be biased as students are less likely to be engaged in high income work than non-students. The presence of this bias seems likely as the number of those 18-24 in our data exceeds 2,000 whereas the full sample size is 3,660. A second obvious limitation in our analysis comes from the fact that the finbetter variable first appeared in the survey in 2014. Because of this, our first two tables only contain roughly 1041 observations. Our first table contains only 1039 observations, which may be due to a mistake in recording the data made by one or more surveyors. Finally, our data is not perfectly analogous to the relevant variables in the original model. (Carbone and Hey, 2004) Our interaction variables may have measured similar influences on an individual s consumption habits, but we took liberties in certain interpretations, such that a response of yes to finbetter in combination with employment meant that the individual had confidence that they would remain employed in future periods. The expectation of being financially better off could come from any number of things including a general belief that all people are made better off in real terms as the economy grows. Finally, our proxy for consumption is limited in that it is a categorical variable. We were unable to measure actual numerical expectations of consumption by respondents, only whether they intended to make a large purchase soon. This is greatly affected by the proximity of our survey to the gift giving holiday of Christmas. Additionally, an individual can still expect to devote a large amount of wealth to consumption, without having individual incidence of consumption exceed 500$ For all of the aforementioned reasons and more, the explanatory value of our analysis is limited. Despite these limitations, this analysis does provide valuable and accurate information about the sampled individuals. While our results did not confer with those of Carbone and Hey (2004) they still did provide a look inside to the spending habits and employment conditions of those sampled.
6 References Carbone, Enrica, and John D. Hey. 2004. The Effect of Unemployment on Consumption: An Experimental Analysis. The Economic Journal 114 (497). Wiley: 660 83. http://www.jstor.org/stable/3590300.