Tegan McCoy 2026-08-18
This project explores patterns of social vulnerability and economic conditions across the Pacific Division, and examines whether the New Markets Tax Credit (NMTC) and Low Income Housing Tax Credit (LIHTC) programs were associated with meaningful changes from 2010 to 2020. Using data from the U.S. Census Bureau and the CDC Social Vulnerability Index (SVI), we evaluated outcomes related to socioeconomic status, household characteristics, minority status, and housing and transportation access. We also incorporated economic measures such as median household income, median home value, and the House Price Index. To understand these patterns, we applied spatial analyses including choropleth and bivariate maps, correlation analyses, k-means clustering, and difference in differences regression models.
Across the project, several themes emerged. The spatial analyses showed clear clusters of higher vulnerability in parts of California, rural Alaska, western Oregon, and Honolulu County. At the same time, lower vulnerability tended to appear in coastal and metropolitan counties such as San Diego, Orange, and Anchorage Counties. Many counties displayed a mix of outcomes, containing both some of the most vulnerable tracts and some of the least vulnerable tracts when examining the four SVI themes individually and in combination. Correlation results showed that vulnerability was not strongly tied to whether a tract received NMTC or LIHTC funding. The difference in differences models found limited program effects, with significant results only for the socioeconomic SVI theme and median home values. These findings suggest that while the programs may support localized improvements, they do not appear to drive broader changes in vulnerability across the division. Future work would benefit from more detailed program data and additional years of SVI and economic information to better understand how these investments interact with long-term community conditions.
The data for this project comes from the U.S. Census Bureau and follows the CDC Social Vulnerability Index framework. For 2010, we used Census tract-level SVI indicators, and for 2020 we used Census block group data that was mapped to 2010 tracts using the NHGIS crosswalk so that the two years could be compared consistently. Nationally, the raw SVI dataset included 73,057 tracts in 2010 and 73,057 block groups in 2020. Within the Pacific Division, there were 10,867 tracts in 2010 and 10,867 tracts in 2020.
From Lab 02, we found that the most vulnerable areas in 2010 in the Pacific Division were concentrated in Walla Walla County, Washington, Clackamas County, Oregon, and Honolulu County, Hawaii. These tracts had high SVI scores across socioeconomic status, housing and transportation burden, and household characteristics. These counties consistently showed elevated vulnerability relative to other counties in the division. Interestingly, Honolulu County also appeared among the least vulnerable tracts, along with San Diego County, California, and Anchorage Municipality, Alaska.
For 2020, Walla Walla County, Washington, and Honolulu County, Hawaii, remained among the most vulnerable tracts, with Ventura County, California, joining them with high scores across the SVI index. For the least vulnerable tracts in 2020, San Diego County, California ranked as the least vulnerable, followed by Washington County, Oregon, and Orange County, California.
Eligibility for the NMTC and LIHTC programs was determined using the criteria we applied in Lab 04. We selected tracts that had never previously received funding and that met program guidelines. NMTC eligibility is based on poverty rates and area median income thresholds, while LIHTC eligibility is based on whether the tract contains or qualifies for affordable housing developments. Using these rules, 31,763 tracts in the Pacific Division were eligible for NMTC and 66,164 were eligible for LIHTC.
In addition to SVI indicators, we incorporated several economic measures to support our analyses. These included Median Household Income and Median Home Value from the Census Bureau, and the House Price Index from the Federal Housing Finance Agency. These indicators were used for county-level grouping in Labs 03 and 04, and for metro-level grouping in Lab 05, allowing us to compare economic and vulnerability patterns across different regional contexts.
In this project, we used several analytical methods to better understand patterns of social vulnerability across the Pacific Division. We started with spatial analyses, using choropleth and bivariate maps to visualize how different SVI themes and economic indicators were distributed across counties and tracts. These maps helped identify where vulnerability was concentrated, where economic conditions were stronger or weaker, and how different measures interacted across the region.
We then used correlation analyses to explore how SVI indicators related to economic outcomes such as median income, median home value, and the House Price Index. This step gave us a sense of which factors tended to move together and where the strongest relationships appeared. After that, we applied k-means clustering to group tracts or counties with similar characteristics. Clustering allowed us to identify patterns that might not be visible from individual maps or summary tables, and it helped highlight similarities and differences across the division.
To evaluate program impacts, we used a difference in differences regression model to compare trends in tracts that received NMTC or LIHTC investment with those that did not. This approach helped us examine whether funded tracts experienced different changes over time in their vulnerability or economic conditions. Taken together, these methods provided a structured way to understand the landscape of vulnerability in the Pacific Division and to assess how federal programs may relate to changes in community outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_SES with treat, post and cbsa (formula: SVI_FLAG_COUNT_SES ~ treat + post + treat * post + cbsa) where treat represents NMTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and moderate proportion of variance (R2 = 0.18, F(50, 7187) = 32.37, p < .001, adj. R2 = 0.18)
The effect of treat × post is statistically significant and negative (beta = -0.42, 95% CI [-0.73, -0.12], t(7187) = -2.72, p = 0.007; Std. beta = -0.03, 95% CI [-0.05, -8.08e-03])
Since the effect of treat x post is statistically significant, we can conclude that the NMTC program had a measurable impact on socioeconomic status-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_HHCHAR with treat, post and cbsa (formula: SVI_FLAG_COUNT_HHCHAR ~ treat + post + treat * post + cbsa) where treat represents NMTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and weak proportion of variance (R2 = 0.06, F(50, 7187) = 9.16, p < .001, adj. R2 = 0.05)
The effect of treat × post is statistically non-significant and negative (beta = -0.06, 95% CI [-0.29, 0.17], t(7187) = -0.51, p = 0.608; Std. beta = -5.87e-03, 95% CI [-0.03, 0.02])
Since the effect of treat x post is not statistically significant, we cannot conclude that the NMTC program had a measurable impact on household characteristics-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_REM with treat, post and cbsa (formula: SVI_FLAG_COUNT_REM ~ treat + post + treat * post + cbsa) where treat represents NMTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.34, F(50, 7187) = 74.72, p < .001, adj. R2 = 0.34)
The effect of treat × post is statistically non-significant and negative (beta = -0.05, 95% CI [-0.13, 0.04], t(7187) = -1.02, p = 0.309; Std. beta = -9.74e-03, 95% CI [-0.03, 9.01e-03])
Since the effect of treat x post is not statistically significant, we cannot conclude that the NMTC program had a measurable impact on racial and ethnic minority status-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_HOUSETRANSPT with treat, post and cbsa (formula: SVI_FLAG_COUNT_HOUSETRANSPT ~ treat + post + treat * post + cbsa) where treat represents NMTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.31, F(50, 7187) = 65.39, p < .001, adj. R2 = 0.31)
The effect of treat × post is statistically non-significant and negative (beta = -0.03, 95% CI [-0.24, 0.19], t(7187) = -0.26, p = 0.795; Std. beta = -2.54e-03, 95% CI [-0.02, 0.02])
Since the effect of treat x post is not statistically significant, we cannot conclude that the NMTC program had a measurable impact on housing and transportation access-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_OVERALL with treat, post and cbsa (formula: SVI_FLAG_COUNT_OVERALL ~ treat + post + treat * post + cbsa) where treat represents NMTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.27, F(50, 7187) = 52.77, p < .001, adj. R2 = 0.26)
The effect of treat × post is statistically non-significant and negative (beta = -0.56, 95% CI [-1.17, 0.05], t(7187) = -1.80, p = 0.072; Std. beta = -0.02, 95% CI [-0.04, 1.62e-03])
Since the effect of treat x post is not statistically significant, we cannot conclude that the NMTC program had a measurable impact on socioeconomic, household characteristics, racial and ethnic minority status, and housing and transportation access-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict MEDIAN_INCOME with treat, post and cbsa (formula: MEDIAN_INCOME ~ treat + post + treat * post + cbsa) where treat represents NMTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and weak proportion of variance (R2 = 0.09, F(50, 7185) = 13.83, p < .001, adj. R2 = 0.08)
The effect of treat × post is statistically non-significant and positive (beta = 0.06, 95% CI [-7.16e-03, 0.14], t(7185) = 1.76, p = 0.078; Std. beta = 0.02, 95% CI [-2.22e-03, 0.04])
Since the effect of treat x post is not statistically significant, we cannot conclude that the NMTC program had a measurable impact on Median Income-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict MEDIAN_HOME_VALUE with treat, post and cbsa (formula: MEDIAN_HOME_VALUE ~ treat + post + treat * post + cbsa) where treat represents NMTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.73, F(50, 6737) = 362.45, p < .001, adj. R2 = 0.73)
The effect of treat × post is statistically significant and positive (beta = 0.11, 95% CI [5.76e-03, 0.22], t(6737) = 2.07, p = 0.039; Std. beta = 0.01, 95% CI [6.80e-04, 0.03])
Since the effect of treat x post is statistically significant, we can conclude that the NMTC program had a measurable impact on Median Home Value-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict HOUSE_PRICE_INDEX with treat, post and cbsa (formula: HOUSE_PRICE_INDEX ~ treat + post + treat * post + cbsa) where treat represents NMTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.32, F(45, 1974) = 20.40, p < .001, adj. R2 = 0.30)
The effect of treat × post is statistically non-significant and positive (beta = 0.05, 95% CI [-0.13, 0.23], t(1974) = 0.55, p = 0.583; Std. beta = 0.01, 95% CI [-0.03, 0.05])
Since the effect of treat x post is not statistically significant, we cannot conclude that the NMTC program had a measurable impact on House Price Index-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_SES with treat, post and cbsa (formula: SVI_FLAG_COUNT_SES ~ treat + post + treat * post + cbsa) where treat represents LIHTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and moderate proportion of variance (R2 = 0.18, F(50, 7187) = 32.37, p < .001, adj. R2 = 0.18)
The effect of treat × post is statistically significant and negative (beta = -0.42, 95% CI [-0.73, -0.12], t(7187) = -2.72, p = 0.007; Std. beta = -0.03, 95% CI [-0.05, -8.08e-03])
Since the effect of treat x post is statistically significant, we can conclude that the LIHTC program had a measurable impact on socioeconomic status-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_HHCHAR with treat, post and cbsa (formula: SVI_FLAG_COUNT_HHCHAR ~ treat + post + treat * post + cbsa) where treat represents LIHTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and weak proportion of variance (R2 = 0.06, F(50, 7187) = 9.16, p < .001, adj. R2 = 0.05)
The effect of treat × post is statistically non-significant and negative (beta = -0.06, 95% CI [-0.29, 0.17], t(7187) = -0.51, p = 0.608; Std. beta = -5.87e-03, 95% CI [-0.03, 0.02])
Since the effect of treat x post is not statistically significant, we cannot conclude that the LIHTC program had a measurable impact on household characteristics-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_REM with treat, post and cbsa (formula: SVI_FLAG_COUNT_REM ~ treat + post + treat * post + cbsa) where treat represents LIHTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.34, F(50, 7187) = 74.72, p < .001, adj. R2 = 0.34)
The effect of treat × post is statistically non-significant and negative (beta = -0.05, 95% CI [-0.13, 0.04], t(7187) = -1.02, p = 0.309; Std. beta = -9.74e-03, 95% CI [-0.03, 9.01e-03])
Since the effect of treat x post is not statistically significant, we cannot conclude that the LIHTC program had a measurable impact on racial and ethnic minority status-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_HOUSETRANSPT with treat, post and cbsa (formula: SVI_FLAG_COUNT_HOUSETRANSPT ~ treat + post + treat * post + cbsa) where treat represents LIHTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.31, F(50, 7187) = 65.39, p < .001, adj. R2 = 0.31)
The effect of treat × post is statistically non-significant and negative (beta = -0.03, 95% CI [-0.24, 0.19], t(7187) = -0.26, p = 0.795; Std. beta = -2.54e-03, 95% CI [-0.02, 0.02])
Since the effect of treat x post is not statistically significant, we cannot conclude that the LIHTC program had a measurable impact on housing and transportation access-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict SVI_FLAG_COUNT_OVERALL with treat, post and cbsa (formula: SVI_FLAG_COUNT_OVERALL ~ treat + post + treat * post + cbsa) where treat represents LIHTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.27, F(50, 7187) = 52.77, p < .001, adj. R2 = 0.26)
The effect of treat × post is statistically non-significant and negative (beta = -0.56, 95% CI [-1.17, 0.05], t(7187) = -1.80, p = 0.072; Std. beta = -0.02, 95% CI [-0.04, 1.62e-03])
Since the effect of treat x post is not statistically significant, we cannot conclude that the LIHTC program had a measurable impact on socioeconomic, household characteristics, racial and ethnic minority status, and housing and transportation access-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict MEDIAN_INCOME with treat, post and cbsa (formula: MEDIAN_INCOME ~ treat + post + treat * post + cbsa) where treat represents LIHTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and weak proportion of variance (R2 = 0.09, F(50, 7185) = 13.83, p < .001, adj. R2 = 0.08)
The effect of treat × post is statistically non-significant and positive (beta = 0.06, 95% CI [-7.16e-03, 0.14], t(7185) = 1.76, p = 0.078; Std. beta = 0.02, 95% CI [-2.22e-03, 0.04])
Since the effect of treat x post is not statistically significant, we cannot conclude that the LIHTC program had a measurable impact on Median Income-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict MEDIAN_HOME_VALUE with treat, post and cbsa (formula: MEDIAN_HOME_VALUE ~ treat + post + treat * post + cbsa) where treat represents LIHTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.73, F(50, 6737) = 362.45, p < .001, adj. R2 = 0.73)
The effect of treat × post is statistically significant and positive (beta = 0.11, 95% CI [5.76e-03, 0.22], t(6737) = 2.07, p = 0.039; Std. beta = 0.01, 95% CI [6.80e-04, 0.03])
Since the effect of treat x post is statistically significant, we can conclude that the LIHTC program had a measurable impact on Median Home Value-related social vulnerability and economic outcomes.
We fitted a linear model (estimated using OLS) to predict HOUSE_PRICE_INDEX with treat, post and cbsa (formula: HOUSE_PRICE_INDEX ~ treat + post + treat * post + cbsa) where treat represents LIHTC program participation, post is the year of 2020 after starting period of 2010, and cbsa controls for metro-level effects.
The model explains a statistically significant and substantial proportion of variance (R2 = 0.32, F(45, 1974) = 20.40, p < .001, adj. R2 = 0.30)
The effect of treat × post is statistically non-significant and positive (beta = 0.05, 95% CI [-0.13, 0.23], t(1974) = 0.55, p = 0.583; Std. beta = 0.01, 95% CI [-0.03, 0.05])
Since the effect of treat x post is not statistically significant, we cannot conclude that the LIHTC program had a measurable impact on House Price Index-related social vulnerability and economic outcomes.
Across the Pacific Division, the SVI infographics and maps from Lab 03 showed steady improvement in several socioeconomic indicators between 2010 and 2020, including reductions in poverty and unemployment and substantial gains in health insurance coverage. Despite these improvements, housing affordability remained a major contributor to vulnerability, especially in areas such as California’s Central Valley, Southern California, rural Alaska, western Oregon, and Honolulu County. Lower vulnerability tended to cluster in coastal and metropolitan counties like San Diego, Orange, and Anchorage. These patterns highlight how structural and geographic factors shape community conditions even when overall economic indicators improve.
The correlation analyses from Lab 04 showed that SVI measures were not strongly aligned with whether a tract received NMTC or LIHTC funding. Many of the tracts with the highest vulnerability scores did not appear to receive funding, while some lower-vulnerability tracts did. This suggests that program eligibility and actual investment do not always overlap with the communities facing the greatest challenges. The spatial findings reinforce this mismatch by showing persistent high-vulnerability clusters in areas that may not be receiving targeted development support.
The diff-in-diff models from Lab 06 add context to these patterns. Both NMTC and LIHTC were associated with statistically significant improvements in only two outcomes: the socioeconomic SVI theme and median home values. For all other SVI themes, overall vulnerability, median income, and the house price index, the effects of program participation were not statistically significant. These results suggest that while individual projects may support localized improvements, the programs’ overall impact on broad tract-level vulnerability patterns appears limited within the Pacific Division.
Taken together, these findings point to several recommendations. Stakeholders may want to refine how tax credit programs identify and prioritize high-need tracts, particularly in regions where vulnerability remains concentrated and persistent. Aligning investments more closely with the areas identified in the spatial analyses could enhance the reach and effectiveness of program efforts. Future research would benefit from more detailed project-level information, including investment size, project type, and timing, as well as additional years of vulnerability and economic data to examine longer-term trends. These steps could help clarify how federal tax-credit programs interact with local conditions and how they might be adjusted to better support communities facing the highest levels of social vulnerability.
Analyses were conducted using the R Statistical language (version 4.3.1; R Core Team, 2023) on Windows 11 x64 (build 26100)
CDFI Fund (2023). FY 2023 NMTC Public Data Release: 2003-2021 Data File Updated - Aug 21, 2023. https://www.cdfifund.gov/documents/data-releases
Centers for Disease Control and Prevention/ Agency for Toxic Substances and Disease Registry/ Geospatial Research, Analysis, and Services Program. (2022). CDC/ATSDR Social Vulnerability Index 2020 Methodology. https://web.archive.org/web/20241028180954/https://www.atsdr.cdc.gov/placeandhealth/svi/documentation/SVI_documentation_2020.html
FHFA (n.d.). HPI® Census Tracts (Developmental Index; Not Seasonally Adjusted). https://www.fhfa.gov/DataTools/Downloads/Pages/House-Price-Index-Datasets.aspx#atvol
HUD User (n.d.). 2010, 2011, and 2012 QCT data for all of the census tracts in the United States and Puerto Rico (qct_data_2010_2011_2012.xlsx). https://www.huduser.gov/portal/datasets/qct.html#year2010
HUD User (2023). Low-Income Housing Tax Credit (LIHTC): Property Level Data. https://www.huduser.gov/portal/datasets/lihtc/property.html
Novogradac New Markets Tax Credit Resource Center. (2017). New Markets Tax Credit Low-Income Community Census Tracts - American Community Survey 2011-2015. https://www.novoco.com/resource-centers/new-markets-tax-credits/data-tables
Steven Manson, Jonathan Schroeder, David Van Riper, Katherine Knowles, Tracy Kugler, Finn Roberts, and Steven Ruggles. IPUMS National Historical Geographic Information System: Version 18.0 [2020 → 2010 Block Groups → Census Tracts Crosswalks National File]. Minneapolis, MN: IPUMS. 2023. http://doi.org/10.18128/D050.V18.0
U.S. Bureau of Labor Statistics (n.d.). CPI Inflation Calculator. https://data.bls.gov/cgi-bin/cpicalc.pl
U.S. Bureau of Labor Statistics (n.d.). QCEW County-MSA-CSA Crosswalk (For NAICS-Based Data). https://www.bls.gov/cew/classifications/areas/county-msa-csa-crosswalk.htm
U.S. Census Bureau. (2011). 2006-2010 American Community Survey 5-year. https://www.census.gov/newsroom/releases/archives/american_community_survey_acs/cb11-208.html
U.S. Census Bureau. (2013). 2008-2012 American Community Survey 5-year. https://www.census.gov/newsroom/press-kits/2013/20131217_acs_5yr.html
U.S. Census Bureau. (2022). 2016-2020 American Community Survey 5-year. https://www.census.gov/newsroom/press-releases/2022/acs-5-year-estimates.html