class: center, middle, inverse, title-slide .title[ # Big Data and Economics ] .subtitle[ ## Neighborhoods and Upward Mobility ] .author[ ### Kyle Coombs ] .date[ ### Bates College |
EC/DCS 368
] --- name: toc <style type="text/css"> @media print { .has-continuation { display: block !important; } } </style> # Table of contents 1. [Prologue](#prologue) 2. [Geographical Variation in Upward Mobility](#geo-vary) 3. [Characteristics of High-Mobility Areas](#characteristics) 4. [Spatial Correlation and Decay](#spatial) --- class: inverse, center, middle name: prologue # Prologue <html><div style='float:left'></div><hr color='#EB811B' size=1px width=796px></html> --- # Housekeeping - Problem Set 2 is due on Friday - Your data description is due after break --- # Prologue - Today's lecture is a little different than the last few - We're talking about an application of big data to a big question: why do some people move up the income ladder and others don't? - This is a big question in economics and public policy - Chetty answered it using big data and spatial analysis - By big I mean: essentially all tax returns in the USA from 1989-2015 - He released summaries of the data publicly in 2018 as the Opportunity Atlas - These show tons of descriptive measures of income mobility at various levels of geography: state, county, and Census Tract - Problem Set 3 will involve using the Census Tract data to learn about income mobility in Lewiston --- <img src="pics/fading_dream.png" width="90%" style="display: block; margin: auto;" /> Source: [Chetty et al. (2014)](https://opportunityinsights.org/paper/recentintergenerationalmobility/) --- # Why is the "American Dream" Fading? - Why are children's chances of climbing the income ladder falling in the USA? - What can be done to reverse this trend? - Need to go beyond macroeconomic data to ansawer this question. Why? - Too many changes happening over time and across space to separate out the causal factors. - Also: only a handful of data points (classic macro problem) --- # Enter the Opportunity Atlas - Created in 2018, the Opportunity Atlas offers one measure of how income mobility differs by location in the USA - If some areas have more mobility than others, can we learn why and apply those lessons elsewhere? - Data sources: - Anonymized Census data (2000, 2010 ACS) covering U.S. population - Federal income tax returns from 1989-2015. - Method: Link parents based on dependent claiming on tax returns - Target sample: Children born between 1978-1983 (U.S. citizens and authorized immigrants who arrived as children) There's bound to be a messy with this much data, so they create an analysis sample - **Analysis sample**: 20.5 million children, 96% coverge of target sample --- # Toolkit to use these data - Data cleaning and wrangling - Data visualization - Spatial analysis (this week) - Regression analysis (after break) --- # Parent and Children Incomes in Tax Data - Parent household incomes: average income reported on Form 1040 tax return from 1994-2000 - Children incomes measured from tax returns in 2014-15 (ages 31-37) - But income levels differ over time! How do we compare them? - Use percentile ranks in the *national* distribution - Rank children relative to others born in same year and parents relative to other parents - [What is a percentile?](https://www.mentimeter.com/app/presentation/blkhnym4ou7ejzod9b1gc6b24id49nr5/xeit9p23asph) -- - **Income percentile**: The fraction of the national income distribution that a person's income exceeds - Take average income percentile of children by parental income percentile --- # Average Child Income Percentile by Parent Income Percentile <img src="09-oppatlas_files/figure-html/oppatlas-1.svg" style="display: block; margin: auto;" /> Source: [The Opportunity Atlas](https://opportunityinsights.org/data/) --- class: inverse, center, middle name: geo-vary # Geographic Variation in Upward Mobility <html><div style='float:left'></div><hr color='#EB811B' size=1px width=796px></html> --- # What is mobility for a given area? - Run this same regression of income ranks by Census tract, county, or commuting zone in the USA<sup>1</sup> - Census tracts are small geographic areas that contain 1,200-8,000 people - For simplicity, Chetty et al. (2018) report the average income percentile of children whose parents were at the 25th percentile of the national income distribution - This is a single measure of upward mobility that is easy to understand and compare across areas - It is not the only measure, but it is a good one - **Big data tip**: Sensibly summary statisics make big data more useful - The right statistic depends on the question you're asking - [Where do you think has the lowest upward mobility? The highest?](https://www.mentimeter.com/app/presentation/blkhnym4ou7ejzod9b1gc6b24id49nr5/9zp5t7zs6ot5) .footnote[<sup>1</sup> Technical detail: Weight each child by fraction of childhood (up to 23) in a given area to account for movement across areas during childhood] --- <img src="pics/op_atlas_map.png" width="90%" style="display: block; margin: auto;" /> *Note: Blue = More Upward Mobility, Red = Less Upward Mobility* Source: [The Opportunity Atlas](https://opportunityinsights.org/data/) --- class: inverse, center, middle name: characteristics # Characteristics of High-Mobility Areas <html><div style='float:left'></div><hr color='#EB811B' size=1px width=796px></html> --- # Why does upward mobility differ? Armed with a summary measure of upward mobility, we can ask: - Why do some areas have more upward mobility than others? - Spatial and correlational analysis is a good place to start - What are potential characteristics of high mobility areas? - Better jobs? - Better schools? - Institutional differences? - Culture? --- # Upward Mobility vs. Job Growth <img src="pics/metro_areas.png" width="90%" style="display: block; margin: auto;" /> --- # Actual correlates 1. Segregation: Greater racial and income segregation associated with lower levels of mobility 2. Income Inequality: Places with smaller middle class have less mobility 3. School Quality: Higher expenditure, smaller classes, higher test scores correlated with more mobility 4. Family Structure: - Areas with more single parents have lower mobility - Strong correlation even for kids whose *own* parents are married 5. Social Capital - It takes a village to raise a child - Chetty et al. (2023) leveraged Facebook Data to create the Social Capital Atlas --- class: inverse, center, middle name: spatial # Spatial Correlation and Decay <html><div style='float:left'></div><hr color='#EB811B' size=1px width=796px></html> --- # Big question: why don't people move? - If some areas have more mobility than others, why don't people move to those areas? - Is it rent? --- # The Price of Opportunity in Seattle Upward Mobility vs Median Rent by Neighborhood <img src="pics/seattle_opportunity.png" width="90%" style="display: block; margin: auto;" /> --- # Big question: why don't people move? - Initial experiments indicate benefits exist from moving (we'll see later) - If some areas have more mobility than others, why don't people move to those areas? - Is it rent? - Other costs of moving? - Maybe they do not want to move as far? - Overall, this is not a highly effective approach --- # Well what if we invest locally? - What if we invest in the areas that have low mobility? (place-based approach) - Would there be spillovers between locations? - It is tough to improve one neighborhood (e.g. a tract), let alone many at once - Do we have to improve them all at once to help people? - The answer to this question changes the policy approach --- # Spatial decay suggests localized effects <img src="pics/spatial_correlation_decay.png" width="90%" style="display: block; margin: auto;" /> --- # Overall Takeaways - Correlation evidence is suggestive, but not causal - Causality requires a more focused approach - We will build this toolkit in the next few lectures --- class: inverse, center, middle # Next lecture: Spatial Analysis and Opportunity Atlas <html><div style='float:left'></div><hr color='#EB811B' size=1px width=796px></html>