Recipient Organization
AUBURN UNIVERSITY
108 M. WHITE SMITH HALL
AUBURN,AL 36849
Performing Department
(N/A)
Non Technical Summary
Myriad policy decisions depend on the implicit or explicit comparisons of individual cost of living values across households with different compositions. People living in large households benefit from the economies of scale of shared or joint consumption. One-person households cannot achieve intrahousehold economies of scale, and it creates cost of living disparities among households of different compositions. Therefore, establishing robust cost of living estimates not confounded with the household size is of particular interest to researchers and policymakers. How much would a two-person household need to attain the same standard of living as a one-person household with a monthly expenditure of $1,000? To answer this question, researchers need to empirically estimate the equivalence scales which compare economic wellbeing of households of different compositions. However, equivalence scales are difficult to construct because household utility cannot be directly measured, which results in econometric identification problems. This project will be the first comprehensive study that empirically estimates the equivalence scales among U.S. households over the past 50 years. We will use (i) utility-based objective methods and household-level scanner data to estimate objective equivalence scales, (ii) life satisfaction surveys from a nationally representative sample to measure subjective equivalence scales, and (iii) Bayes Theorem to develop a new equivalence scale that combines the information content of both methods to produce a third equivalence scale estimate which is more flexible and efficient. The estimated equivalence scales could be widely applied, inter alia, to define the poverty line, evaluate social inequality and poverty, calibrate social safety net payments, measure the costs of raising children, and calculate payments for life insurance, alimony, and legal compensation for wrongful death.
Animal Health Component
50%
Research Effort Categories
Basic
50%
Applied
50%
Developmental
(N/A)
Goals / Objectives
Myriad policy decisions depend on the implicit or explicit comparisons of individual cost of living values across households with different compositions. People living in large households benefit from the economies of scale of shared or joint consumption. One-person households cannot achieve intrahousehold economies of scale, and it creates cost of living disparities among households of different compositions. Therefore, establishing robust cost of living estimates not confounded with the household size is of particular interest to researchers and policymakers. How much would a two-person household need to attain the same standard of living as a one-person household with a monthly expenditure of $1,000? To answer this question, researchers need to empirically estimate the equivalence scales which compare economic wellbeing of households of different compositions. However, equivalence scales are difficult to construct because household utility cannot be directly measured, which results in econometric identification problems. This project will be the first comprehensive study that empirically estimates the equivalence scales among U.S. households over the past 50 years.There are three types of approaches to calculating equivalence scales: expert judgment method; utility-based objective method; and survey-based subjective method. The expert judgment method relies on individuals or panels of experts' views on a reasonable scale. While the scales resulting from this process may be based on some empirical evidence, they are more likely to reflect the subjective opinions of experts. The most widely used equivalence scales (e.g., OECD Scale, OECD-modified Scale, and Square Root Scale) including the one implied by the official U.S. poverty lines are examples of such a process. We identify the following gaps in equivalence scales that are currently in use:• There is a lack of research and evidence to justify the current widely-used equivalence scales in the U.S. The method and process for determining the equivalence scales in use is not transparent since the last published estimate of the equivalence scale by the U.S. government was in 1968, over 50 years ago.• The widely used expert approach requires normative judgments about the goods and services needed to meet a particular standard of living. However, the baskets of goods and services consumed reflect, to some extent, the customs, habits, and social expectations of different consumers. Thus, the expert approach is inevitably arbitrary.• When used to estimate the costs of children, the expert method ignores shared household costs (e.g., food, transportation, housing) and only include those expenditures which can clearly be attributed to children (e.g., clothing, health, personal care). Therefore, it produces a biased estimate of the cost of raising children.We seek to address the limitations above by proposing three objectives:Objective 1: Use utility-based objective methods to estimate objective equivalence scales: We will identify both resource allocation and scale economies for each family member without the need to observe the consumption choices of single-person households. Specifically, the structural parameters of interest will be identified by comparing the Engel curves of private assignable goods across family members and household compositions.Objective 2: Use survey-based subjective methods to estimate subjective equivalence scalesWe will use a holistic approach, eliciting preferences to construct the subjective equivalence scale with the help of incentivized survey instruments:We will also elicit incentivized beliefs to reduce the magnitude of inaccuracies in happiness and life satisfaction measures.Objective 3: Use a Bayesian method to combine both objective and subjective methods to produce reliable equivalence scale estimates: We will use Bayesian information processing rules to combine the two equivalence scale measures derived in the first two objectives into a single, efficient measure of intra-household economies of scale.
Project Methods
Objective 1: Using utility-based objective methods to estimate objective equivalence scalesWe will utilize the indifference scale to capture the well-being change of each household member. An indifference scale is a variant of equivalence scales, which measures the fraction of household expenditure that puts an individual living alone on the same indifference curve that she (or he) would attain living in a multi-person household (Browning, Chiappori, and Lewbel 2013). Since indifference scales and resource shares for women and men among the population of two-person childless households have been estimated in Li and Dorfman (2021), this project will consist of three research questions: First, we will attempt to provide the indifference scale for women and men living with one child compared with those living without a child. Second, we will estimate the indifference scales for women, men, and children living in multi-children households compared with those living in one-child households. Third, we will further consider the heterogeneity across genders, ages, races, and other demographic variables and estimate the indifference scale separately. For Objective 1, we will use IRI Consumer Network household scanner data and IRI MedProfiler consumer health data (2015-2021) through the third-party agreement with the USDA ERS. The IRI Consumer Network survey is a national representative household panel survey that contains self-reported information on food purchases among over 120,000 households. We can identify the expenditure on private goods for each household member and the characteristics of each household and individuals in the dataset.Objective 2Using survey-based subjective methods to estimate subjective equivalence scalesCalibrating the equivalence scale using subjective life satisfaction and happiness measures was first introduced by van Praag (1968) and later modified by van Praag and Sar (1988) and van Praag (1991). The existing methodologies in the literature for the construction of the subjective equivalence measure rely on a set of assumptions. The primary assumption is that individuals accurately assess their standard of living or welfare when they are asked to report their life satisfaction or happiness. However, it has been shown that reported life satisfaction measures are prone to measurement errors stemming from different individual characteristics (Layard et al., 2008). Controlling a broad spectrum of demographic and household characteristics can alleviate the estimation bias inflicted by potential measurement errors. However, a more fundamental issue is biases and measurement errors that often are prevalent in the stated preference methods. Different survey techniques developed in behavioral economics literature will be used to reduce biases and minimize measurement errors in survey responses.We will work with Qualtrics (a professional Survey Company) to conduct our survey with a nationally representative sample comprised of United States residents. The sample will be balanced across age, gender, income, and residential locational characteristics to mimic the entire U.S. population. We will employ a non-probability sampling method to ensure that we have a balanced representation of the United States population across each category. We will target survey responses in the range of 1500 to 2000, in line with existing studies focusing on constructing inferences for the entire United States population (McCall et al., 2017).Objective 3: Use a Bayesian method to combine both objective and subjective methods to produce reliable equivalence scale estimatesFinal methods for combining the two equivalence scales have not been determined, but we will use Bayesian information processing rules (Zellner 2002) to combine the two equivalence scale measures derived in the first two objectives into a single, efficient measure of intra-household economies of scale. Essentially, one (likely the subjective equivalence scale) will be treated as the prior information while the other (the utility- and data-based equivalence scale) will enter in the role typically referred to as the likelihood function. Principles of maximum entropy will be employed to ensure that no distributional assumptions are employed that "add" information not actually present in our two measures. Given that the distributional aspects of our two measures are almost surely not compatible with solving analytically for the combined measures (the "posterior" distribution), numerical Bayesian techniques will be relied upon. Numerical Bayesian approaches utilize pseudo-random number generation to approximate the integrals at the heart of Bayes' Theorem in cases for which the calculus involved cannot be solved in a straightforward manner. These techniques, developed mostly in the 1980s and 1990s are not fairly standard and can be employed in feasible time frames with today's computing power, even for complex problems such as the one here. In this case, Gibbs sampling is the most likely approach to utilize, as that relies on pseudo-random draws from conditional distributions of subsets of the parameters of interest. Because we should know the distributions of both the subjective and objective equivalence scales, the conditional posterior distributions of parameter subsets will likely be tractable.