Evaluating Data Reliability for Algorithmic Decision-Making
Xinyuan(Christine) Cui
September 29, 2025
You are a data analyst for the Texas Department of Human Services. The department is considering implementing an algorithmic system to identify communities that should receive priority for social service funding and outreach programs. Your supervisor has asked you to evaluate the quality and reliability of available census data to inform this decision.
Drawing on our Week 2 discussion of algorithmic bias, you need to assess not just what the data shows, but how reliable it is and what communities might be affected by data quality issues.
Submit by posting your updated portfolio link on Canvas. Your assignment should be accessible at your-portfolio-url/assignments/assignment_1/
Make sure to update your _quarto.yml navigation to include this assignment under an “Assignments” menu.
Create this assignment in your portfolio repository under an assignments/assignment_1/ folder structure. Update your navigation menu to include:
- text: Assignments
menu:
- href: assignments/assignment_1/your_file_name.qmd
text: "Assignment 1: Census Data Exploration"If there is a special character like comma, you need use double quote mark so that the quarto can identify this as text
# Load required packages (hint: you need tidycensus, tidyverse, and knitr)
"knitr" %in% rownames(installed.packages())[1] TRUElibrary(tidycensus)
library(tidyverse)
library(knitr)
# Set your Census API key
census_api_key("20068788c6e79d5716fbceb0dcd562ab23f74ca1", overwrite=TRUE, install = TRUE)[1] "20068788c6e79d5716fbceb0dcd562ab23f74ca1"State Selection: I have chosen Texas for this analysis because: Texas is an ideal state for this analysis due to its unique combination of rapid population growth, demographic diversity, and significant geographic variation. These factors present specific data quality challenges that are critical to evaluating algorithmic fairness.
Your Task: Use get_acs() to retrieve county-level data for your chosen state.
Requirements: - Geography: county level - Variables: median household income (B19013_001) and total population (B01003_001)
- Year: 2022 - Survey: acs5 - Output format: wide
Hint: Remember to give your variables descriptive names using the variables = c(name = "code") syntax.
# Write your get_acs() code here
county_data_texas <- get_acs(
geography = "county",
variables = c(
median_household_income = "B19013_001",
total_population = "B01003_001"
),
state = my_state,
year = 2022,
survey = "acs5",
output = "wide"
)
# Clean the county names to remove state name and "County"
# Hint: use mutate() with str_remove()
county_data_texas <- county_data_texas %>%
mutate(
# remove state name（Texas）
NAME = str_remove(NAME, ", Texas"),
# remove " County"
NAME = str_remove(NAME, " County")
)
# Display the first few rows
head(county_data_texas)# A tibble: 6 × 6
GEOID NAME median_household_inc…¹ median_household_inc…² total_populationE
<chr> <chr> <dbl> <dbl> <dbl>
1 48001 Anderson 57445 4562 58077
2 48003 Andrews 86458 16116 18362
3 48005 Angelina 57055 2484 86608
4 48007 Aransas 58168 6458 24048
5 48009 Archer 69954 8482 8649
6 48011 Armstro… 70417 14574 1912
# ℹ abbreviated names: ¹median_household_incomeE, ²median_household_incomeM
# ℹ 1 more variable: total_populationM <dbl>Your Task: Calculate margin of error percentages and create reliability categories.
Requirements: - Calculate MOE percentage: (margin of error / estimate) * 100 - Create reliability categories: - High Confidence: MOE < 5% - Moderate Confidence: MOE 5-10%
- Low Confidence: MOE > 10% - Create a flag for unreliable estimates (MOE > 10%)
Hint: Use mutate() with case_when() for the categories.
# Calculate MOE percentage and reliability categories using mutate()
county_data_texas <- county_data_texas %>%
mutate(
# Calculate MOE percentage
income_moe_pct = (median_household_incomeM / median_household_incomeE) * 100,
# Create reliability categories
income_reliability = case_when(
income_moe_pct < 5 ~ "High Confidence",
income_moe_pct >= 5 & income_moe_pct <= 10 ~ "Moderate Confidence",
income_moe_pct > 10 ~ "Low Confidence"
),
# Create a flag for unreliable estimates
unreliable_income = income_moe_pct > 10
)
# Create a summary showing count of counties in each reliability category
# Hint: use count() and mutate() to add percentages
county_data_texas %>%
count(income_reliability) %>%
mutate(
percentage = (n / sum(n)) * 100
)# A tibble: 4 × 3
income_reliability n percentage
<chr> <int> <dbl>
1 High Confidence 58 22.8
2 Low Confidence 113 44.5
3 Moderate Confidence 82 32.3
4 <NA> 1 0.394Your Task: Identify the 5 counties with the highest MOE percentages.
Requirements: - Sort by MOE percentage (highest first) - Select the top 5 counties - Display: county name, median income, margin of error, MOE percentage, reliability category - Format as a professional table using kable()
Hint: Use arrange(), slice(), and select() functions.
# Create table of top 5 counties by MOE percentage
library(knitr)
county_data_texas %>%
arrange(desc(income_moe_pct)) %>%
slice(1:5) %>%
select(
`County Name` = NAME,
`Median Income` = median_household_incomeE,
`Margin of Error` = median_household_incomeM,
`MOE Percentage` = income_moe_pct,
`Reliability Category` = income_reliability
) %>%
# Format as table with kable() - include appropriate column names and caption
kable(
caption = "Top 5 Counties with The Highest MOE Percentages",
digits = 2
)| County Name | Median Income | Margin of Error | MOE Percentage | Reliability Category |
|---|---|---|---|---|
| Jeff Davis | 38125 | 25205 | 66.11 | Low Confidence |
| Culberson | 35924 | 18455 | 51.37 | Low Confidence |
| King | 59375 | 29395 | 49.51 | Low Confidence |
| Kinney | 52386 | 23728 | 45.29 | Low Confidence |
| Dimmit | 27374 | 12374 | 45.20 | Low Confidence |
Data Quality Commentary: Based on these results, algorithmic funding decisions relying on the median income in counties like Jeff Davis and Culberson (with MOE>50%) would be highly unreliable. These counties risk being poorly served because the true income value could be far outside the estimate, potentially leading to the algorithm severely over-allocating or under-allocating crucial resources. The extreme uncertainty is primarily due to small population size and low survey response rates, making precise income estimation statistically impossible in these sparsely populated areas.
Your Task: Select 2-3 counties from your reliability analysis for detailed tract-level study.
Strategy: Choose counties that represent different reliability levels (e.g., 1 high confidence, 1 moderate, 1 low confidence) to compare how data quality varies.
# Use filter() to select 2-3 counties from your county_reliability data
# Store the selected counties in a variable called selected_counties
selected_counties <- county_data_texas %>%
filter(
income_moe_pct > 30 |
(income_moe_pct >= 5 & income_moe_pct <= 10) |
income_moe_pct < 2
) %>%
# select 3
group_by(income_reliability) %>%
slice(1)
# Display the selected counties with their key characteristics
# Show: county name, median income, MOE percentage, reliability category
selected_counties %>%
select(
`County Name` = NAME,
`Median Income` = median_household_incomeE,
`MOE Percentage` = income_moe_pct,
`Reliability Category` = income_reliability
) %>%
kable(
caption = "3 Counties Selected for Tract-Level Study",
digits= 2
)| County Name | Median Income | MOE Percentage | Reliability Category |
|---|---|---|---|
| Bexar | 67275 | 1.15 | High Confidence |
| Brooks | 30566 | 33.82 | Low Confidence |
| Anderson | 57445 | 7.94 | Moderate Confidence |
Comment on the output: I deliberately selected counties representing the more extreme ends of the High and Low Confidence ranges. This strategy ensures a clearer and more impactful comparison of data quality. By focusing on these distinct cases (one very high confidence, one moderate, and one very low confidence), the tract-level analysis will better expose the full spectrum of data reliability issues that could affect the department’s funding algorithms.
Your Task: Get demographic data for census tracts in your selected counties.
Requirements: - Geography: tract level - Variables: white alone (B03002_003), Black/African American (B03002_004), Hispanic/Latino (B03002_012), total population (B03002_001) - Use the same state and year as before - Output format: wide
Challenge: You’ll need county codes, not names. Look at the GEOID patterns in your county data for hints.
# Define your race/ethnicity variables with descriptive names
ethnicity_variables <- c(
white_alone = "B03002_003",
Black_African_American = "B03002_004",
Hispanic_Latino = "B03002_012",
total_population = "B03002_001"
)
# Use get_acs() to retrieve tract-level data
# Hint: You may need to specify county codes in the county parameter
selected_county_geoids <- selected_counties$GEOID
selected_county_codes <- str_sub(selected_county_geoids, start = -3) #the last 3 numbers of GEOID is county codes
tract_data_texas <- get_acs(
geography = "tract",
variables = ethnicity_variables,
state = my_state,
county = selected_county_codes,
year = 2022,
survey = "acs5",
output = "wide"
)
# Calculate percentage of each group using mutate()
# Create percentages for white, Black, and Hispanic populations
tract_data_texas <- tract_data_texas %>%
mutate(
pct_white = (white_aloneE / total_populationE) * 100,
pct_black = (Black_African_AmericanE / total_populationE) * 100,
pct_hispanic = (Hispanic_LatinoE / total_populationE) * 100
)
# Add readable tract and county name columns using str_extract() or similar
tract_data_texas <- tract_data_texas %>%
mutate(
Tract = str_extract(NAME, "Census Tract \\d+(?:\\.\\d+)?"),
County = str_extract(NAME, "[A-Za-z]+ County")
)
head(tract_data_texas)# A tibble: 6 × 15
GEOID NAME white_aloneE white_aloneM Black_African_Americ…¹
<chr> <chr> <dbl> <dbl> <dbl>
1 48001950100 Census Tract 950… 4061 682 213
2 48001950401 Census Tract 950… 1073 308 1569
3 48001950402 Census Tract 950… 1920 245 2419
4 48001950500 Census Tract 950… 1924 381 801
5 48001950600 Census Tract 950… 3287 674 711
6 48001950700 Census Tract 950… 626 174 1344
# ℹ abbreviated name: ¹Black_African_AmericanE
# ℹ 10 more variables: Black_African_AmericanM <dbl>, Hispanic_LatinoE <dbl>,
# Hispanic_LatinoM <dbl>, total_populationE <dbl>, total_populationM <dbl>,
# pct_white <dbl>, pct_black <dbl>, pct_hispanic <dbl>, Tract <chr>,
# County <chr>Your Task: Analyze the demographic patterns in your selected areas.
# Find the tract with the highest percentage of Hispanic/Latino residents
# Hint: use arrange() and slice() to get the top tract
top_hispanic_tract <- tract_data_texas %>%
# 按照 Hispanic/Latino 百分比 (pct_hispanic) 降序排列
arrange(desc(pct_hispanic)) %>%
slice(1)
top_hispanic_tract %>%
kable(
caption = "Tract with the Highest Percentage of Hispanic/Latino Residents",
)| GEOID | NAME | white_aloneE | white_aloneM | Black_African_AmericanE | Black_African_AmericanM | Hispanic_LatinoE | Hispanic_LatinoM | total_populationE | total_populationM | pct_white | pct_black | pct_hispanic | Tract | County |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 48029160100 | Census Tract 1601; Bexar County; Texas | 18 | 26 | 0 | 21 | 6931 | 1114 | 6949 | 1112 | 0.2590301 | 0 | 99.74097 | Census Tract 1601 | Bexar County |
# Calculate average demographics by county using group_by() and summarize()
# Show: number of tracts, average percentage for each racial/ethnic group
county_demographics_summary <- tract_data_texas %>%
group_by(County) %>%
summarize(
number_of_tracts = n(),
avg_pct_white = mean(pct_white, na.rm = TRUE),
avg_pct_black = mean(pct_black, na.rm = TRUE),
avg_pct_hispanic = mean(pct_hispanic, na.rm = TRUE),
total_population_E = sum(total_populationE, na.rm = TRUE),
)
county_demographics_summary %>%
select(
`County` = County,
`Tracts` = number_of_tracts,
`Avg White/%` = avg_pct_white,
`Avg Black/%` = avg_pct_black,
`Avg Hispanic/%` = avg_pct_hispanic,
`Total Pop` = total_population_E
) %>%
kable(
caption = "Average Demographics by County",
digits = 2
)| County | Tracts | Avg White/% | Avg Black/% | Avg Hispanic/% | Total Pop |
|---|---|---|---|---|---|
| Anderson County | 12 | 56.89 | 18.80 | 17.98 | 58077 |
| Bexar County | 375 | 25.50 | 6.42 | 62.54 | 2014059 |
| Brooks County | 2 | 8.42 | 0.88 | 89.16 | 7059 |
Your Task: Examine margins of error for demographic variables to see if some communities have less reliable data.
Requirements: - Calculate MOE percentages for each demographic variable - Flag tracts where any demographic variable has MOE > 15% - Create summary statistics
# Calculate MOE percentages for white, Black, and Hispanic variables
# Hint: use the same formula as before (margin/estimate * 100)
tract_data_texas <- tract_data_texas %>%
mutate(
# Calculate MOE percentage
white_moe_pct = (white_aloneM / white_aloneE) * 100,
black_moe_pct = (Black_African_AmericanM / Black_African_AmericanE) * 100,
hispanic_moe_pct = (Hispanic_LatinoM / Hispanic_LatinoE) * 100,
# Create a flag for tracts with high MOE on any demographic variable
# Use logical operators (| for OR) in an ifelse() statement
demographic_reliability = ifelse(
white_moe_pct > 15 | black_moe_pct > 15 | hispanic_moe_pct > 15,"High MOE",
"Acceptable"
)
)
# Create summary statistics showing how many tracts have data quality issues
# Calculate MOE percentage and reliability categories using mutate()
tract_data_texas %>%
count(demographic_reliability) %>%
mutate(
percentage = (n / sum(n)) * 100
)# A tibble: 1 × 3
demographic_reliability n percentage
<chr> <int> <dbl>
1 High MOE 389 100Your Task: Investigate whether data quality problems are randomly distributed or concentrated in certain types of communities.
# Group tracts by whether they have high MOE issues
# Calculate average characteristics for each group:
# - population size, demographic percentages
# Use group_by() and summarize() to create this comparison
# Create a professional table showing the patterns
tract_reliability_comparison <- tract_data_texas %>%
group_by(demographic_reliability) %>%
summarize(
Number_of_Tracts = n(),
Avg_Population = mean(total_populationE, na.rm = TRUE),
Avg_Pct_White = mean(pct_white, na.rm = TRUE),
Avg_Pct_Black = mean(pct_black, na.rm = TRUE),
Avg_Pct_Hispanic = mean(pct_hispanic, na.rm = TRUE)
)
tract_reliability_comparison %>%
select(
`Reliability Group` = demographic_reliability,
`Tract Count` = Number_of_Tracts,
`Avg Pop. Size` = Avg_Population,
`Avg White/%` = Avg_Pct_White,
`Avg Black/%` = Avg_Pct_Black,
`Avg Hispanic/%` = Avg_Pct_Hispanic
) %>% # KEY：使用管道符将数据框传递给 kable
kable(
caption = "Comparison of Tract Characteristics by Data Reliability",
digits = 2
)| Reliability Group | Tract Count | Avg Pop. Size | Avg White/% | Avg Black/% | Avg Hispanic/% |
|---|---|---|---|---|---|
| High MOE | 389 | 5344.97 | 26.39 | 6.78 | 61.29 |
Pattern Analysis: The analysis reveals an extreme, systemic data quality issue: 100% of the tracts in the selected focus areas fall into the High MOE category. This means data reliability is not merely poor in isolated cases, but universally insufficient across these tracts. The average population size for all these unreliable tracts is around 5,345 people, with a high average concentration of Hispanic residents (61.29%). This pattern strongly suggests that in small, high-minority counties, the reliability of demographic estimates completely breaks down at the tract level. The lack of sufficient sample size in these relatively small, rural, tracts makes the ACS data inherently unstable. Consequently, any algorithm relying on this demographic data in these counties would be making resource allocation decisions based on pure statistical uncertainty, disproportionately impacting Hispanic-majority communities with unreliable information.
Your Task: Write an executive summary that integrates findings from all four analyses.
Executive Summary Requirements: 1. Overall Pattern Identification: What are the systematic patterns across all your analyses? 2. Equity Assessment: Which communities face the greatest risk of algorithmic bias based on your findings? 3. Root Cause Analysis: What underlying factors drive both data quality issues and bias risk? 4. Strategic Recommendations: What should the Department implement to address these systematic issues?
Executive Summary:
The analysis reveals a two-type, systematic breakdown in data reliability: At the county level, high MOE is concentrated in sparsely populated counties like Jeff Davis and Culberson. These areas immediately fail basic reliability thresholds for household income estimates. At the county level, 100% of all tracts examined fell into the High Uncertainty category, showing an MOE greater than 15% on demographic variables. This means data quality completely collapses at the neighborhood level within these regions, making reliable demographic information universally insufficient for use in funding or allocation decisions.
Communities with the highest risk of algorithmic bias are those that are small, rural, and possess a high concentration of minority residents. The tracts flagged as universally unreliable show an average of 61.29% Hispanic residents. Algorithms relying on this fundamentally flawed data risk systematically under-serving these communities, as any resource allocation based on the estimate would carry a high degree of statistical uncertainty, effectively rendering the data unusable for precise needs assessment.
The underlying factor driving both data quality issues and the resulting bias risk is the insufficient statistical sample size combined with geographical isolation. In sparsely populated, often rural counties, the ACS cannot collect enough household responses to produce stable estimates for either income or demographics. Because high-minority communities are overrepresented in these low-population areas in Texas, the data gaps systematically align with ethnic lines, translating a statistical problem into an undeniable equity problem.
To reduce this systemic risk and ensure equitable allocation, the Department should implement the following strategic measures: 1. Implement a strict rule requiring that any data used by the allocation algorithm must have an MOE percentage below a pre-defined threshold (e.g., 10%). Funding allocation logic must be halted if this threshold is breached. 2. For all counties or tracts flagged as “High Uncertainty” (MOE>15%), the algorithm must default to using supplementary, non-ACS data rather than relying on unreliable census estimates. 3. Program the algorithm to automatically flag and defer funding decisions for any tract with high data uncertainty to a human review committee, ensuring local context and qualitative data are applied where quantitative data is weakest.
Your Task: Create a decision framework for algorithm implementation.
# Create a summary table using your county reliability data
# Include: county name, median income, MOE percentage, reliability category
county_data_texas %>%
select(
`County Name` = NAME,
`Median Income` = median_household_incomeE,
`MOE Percentage` = income_moe_pct,
`Reliability Category` = income_reliability
) %>%
# Add a new column with algorithm recommendations using case_when():
# - High Confidence: "Safe for algorithmic decisions"
# - Moderate Confidence: "Use with caution - monitor outcomes"
# - Low Confidence: "Requires manual review or additional data"
mutate(
`Algorithm Recommendation` = case_when(
# High Confidence: MOE < 5%
`Reliability Category` == "High Confidence" ~ "Safe for algorithmic decisions",
# Moderate Confidence: MOE 5-10%
`Reliability Category` == "Moderate Confidence" ~ "Use with caution - monitor outcomes",
# Low Confidence: MOE > 10%
`Reliability Category` == "Low Confidence" ~ "Requires manual review or additional data"
)
) %>%
# Format as a professional table with kable()
kable(
caption = "Algorithmic Implementation Framework",
digits = 2
)| County Name | Median Income | MOE Percentage | Reliability Category | Algorithm Recommendation |
|---|---|---|---|---|
| Anderson | 57445 | 7.94 | Moderate Confidence | Use with caution - monitor outcomes |
| Andrews | 86458 | 18.64 | Low Confidence | Requires manual review or additional data |
| Angelina | 57055 | 4.35 | High Confidence | Safe for algorithmic decisions |
| Aransas | 58168 | 11.10 | Low Confidence | Requires manual review or additional data |
| Archer | 69954 | 12.13 | Low Confidence | Requires manual review or additional data |
| Armstrong | 70417 | 20.70 | Low Confidence | Requires manual review or additional data |
| Atascosa | 67442 | 6.39 | Moderate Confidence | Use with caution - monitor outcomes |
| Austin | 73556 | 6.47 | Moderate Confidence | Use with caution - monitor outcomes |
| Bailey | 69830 | 18.79 | Low Confidence | Requires manual review or additional data |
| Bandera | 70965 | 8.05 | Moderate Confidence | Use with caution - monitor outcomes |
| Bastrop | 80151 | 6.09 | Moderate Confidence | Use with caution - monitor outcomes |
| Baylor | 52716 | 25.14 | Low Confidence | Requires manual review or additional data |
| Bee | 50283 | 10.00 | Moderate Confidence | Use with caution - monitor outcomes |
| Bell | 62858 | 2.80 | High Confidence | Safe for algorithmic decisions |
| Bexar | 67275 | 1.15 | High Confidence | Safe for algorithmic decisions |
| Blanco | 79717 | 9.35 | Moderate Confidence | Use with caution - monitor outcomes |
| Borden | 80625 | 24.88 | Low Confidence | Requires manual review or additional data |
| Bosque | 63868 | 6.05 | Moderate Confidence | Use with caution - monitor outcomes |
| Bowie | 56628 | 4.12 | High Confidence | Safe for algorithmic decisions |
| Brazoria | 91972 | 3.31 | High Confidence | Safe for algorithmic decisions |
| Brazos | 57562 | 3.45 | High Confidence | Safe for algorithmic decisions |
| Brewster | 47747 | 10.97 | Low Confidence | Requires manual review or additional data |
| Briscoe | 35446 | 27.64 | Low Confidence | Requires manual review or additional data |
| Brooks | 30566 | 33.82 | Low Confidence | Requires manual review or additional data |
| Brown | 53792 | 4.75 | High Confidence | Safe for algorithmic decisions |
| Burleson | 71745 | 6.53 | Moderate Confidence | Use with caution - monitor outcomes |
| Burnet | 71482 | 8.09 | Moderate Confidence | Use with caution - monitor outcomes |
| Caldwell | 66779 | 6.96 | Moderate Confidence | Use with caution - monitor outcomes |
| Calhoun | 62267 | 9.92 | Moderate Confidence | Use with caution - monitor outcomes |
| Callahan | 63906 | 3.63 | High Confidence | Safe for algorithmic decisions |
| Cameron | 47435 | 3.41 | High Confidence | Safe for algorithmic decisions |
| Camp | 53968 | 7.64 | Moderate Confidence | Use with caution - monitor outcomes |
| Carson | 83199 | 4.51 | High Confidence | Safe for algorithmic decisions |
| Cass | 54303 | 6.82 | Moderate Confidence | Use with caution - monitor outcomes |
| Castro | 59886 | 17.03 | Low Confidence | Requires manual review or additional data |
| Chambers | 106103 | 8.28 | Moderate Confidence | Use with caution - monitor outcomes |
| Cherokee | 56971 | 8.45 | Moderate Confidence | Use with caution - monitor outcomes |
| Childress | 56063 | 29.34 | Low Confidence | Requires manual review or additional data |
| Clay | 75227 | 7.11 | Moderate Confidence | Use with caution - monitor outcomes |
| Cochran | 41597 | 16.97 | Low Confidence | Requires manual review or additional data |
| Coke | 40230 | 13.80 | Low Confidence | Requires manual review or additional data |
| Coleman | 51034 | 7.95 | Moderate Confidence | Use with caution - monitor outcomes |
| Collin | 113255 | 1.16 | High Confidence | Safe for algorithmic decisions |
| Collingsworth | 52045 | 22.02 | Low Confidence | Requires manual review or additional data |
| Colorado | 63352 | 8.17 | Moderate Confidence | Use with caution - monitor outcomes |
| Comal | 93744 | 2.83 | High Confidence | Safe for algorithmic decisions |
| Comanche | 57383 | 13.32 | Low Confidence | Requires manual review or additional data |
| Concho | 55750 | 27.87 | Low Confidence | Requires manual review or additional data |
| Cooke | 66374 | 8.36 | Moderate Confidence | Use with caution - monitor outcomes |
| Coryell | 63281 | 3.62 | High Confidence | Safe for algorithmic decisions |
| Cottle | 47625 | 37.82 | Low Confidence | Requires manual review or additional data |
| Crane | 71364 | 32.89 | Low Confidence | Requires manual review or additional data |
| Crockett | 64103 | 33.95 | Low Confidence | Requires manual review or additional data |
| Crosby | 50268 | 10.40 | Low Confidence | Requires manual review or additional data |
| Culberson | 35924 | 51.37 | Low Confidence | Requires manual review or additional data |
| Dallam | 71969 | 9.75 | Moderate Confidence | Use with caution - monitor outcomes |
| Dallas | 70732 | 0.86 | High Confidence | Safe for algorithmic decisions |
| Dawson | 45268 | 27.64 | Low Confidence | Requires manual review or additional data |
| Deaf Smith | 51942 | 6.01 | Moderate Confidence | Use with caution - monitor outcomes |
| Delta | 68491 | 27.75 | Low Confidence | Requires manual review or additional data |
| Denton | 104180 | 1.30 | High Confidence | Safe for algorithmic decisions |
| DeWitt | 61100 | 7.86 | Moderate Confidence | Use with caution - monitor outcomes |
| Dickens | 46638 | 13.62 | Low Confidence | Requires manual review or additional data |
| Dimmit | 27374 | 45.20 | Low Confidence | Requires manual review or additional data |
| Donley | 51711 | 12.45 | Low Confidence | Requires manual review or additional data |
| Duval | 50697 | 20.28 | Low Confidence | Requires manual review or additional data |
| Eastland | 52902 | 12.21 | Low Confidence | Requires manual review or additional data |
| Ector | 70566 | 4.20 | High Confidence | Safe for algorithmic decisions |
| Edwards | 40809 | 27.09 | Low Confidence | Requires manual review or additional data |
| Ellis | 93248 | 2.66 | High Confidence | Safe for algorithmic decisions |
| El Paso | 55417 | 1.92 | High Confidence | Safe for algorithmic decisions |
| Erath | 59654 | 6.73 | Moderate Confidence | Use with caution - monitor outcomes |
| Falls | 45172 | 15.35 | Low Confidence | Requires manual review or additional data |
| Fannin | 65835 | 6.09 | Moderate Confidence | Use with caution - monitor outcomes |
| Fayette | 72881 | 5.03 | Moderate Confidence | Use with caution - monitor outcomes |
| Fisher | 60461 | 8.19 | Moderate Confidence | Use with caution - monitor outcomes |
| Floyd | 49321 | 8.99 | Moderate Confidence | Use with caution - monitor outcomes |
| Foard | 41944 | 20.94 | Low Confidence | Requires manual review or additional data |
| Fort Bend | 109987 | 2.64 | High Confidence | Safe for algorithmic decisions |
| Franklin | 67915 | 4.37 | High Confidence | Safe for algorithmic decisions |
| Freestone | 55902 | 10.53 | Low Confidence | Requires manual review or additional data |
| Frio | 56042 | 30.34 | Low Confidence | Requires manual review or additional data |
| Gaines | 73299 | 13.82 | Low Confidence | Requires manual review or additional data |
| Galveston | 83913 | 2.78 | High Confidence | Safe for algorithmic decisions |
| Garza | 56215 | 34.96 | Low Confidence | Requires manual review or additional data |
| Gillespie | 70162 | 8.15 | Moderate Confidence | Use with caution - monitor outcomes |
| Glasscock | 112188 | 27.86 | Low Confidence | Requires manual review or additional data |
| Goliad | 58125 | 25.84 | Low Confidence | Requires manual review or additional data |
| Gonzales | 64255 | 8.43 | Moderate Confidence | Use with caution - monitor outcomes |
| Gray | 54563 | 7.16 | Moderate Confidence | Use with caution - monitor outcomes |
| Grayson | 66608 | 3.60 | High Confidence | Safe for algorithmic decisions |
| Gregg | 63811 | 3.91 | High Confidence | Safe for algorithmic decisions |
| Grimes | 63484 | 9.10 | Moderate Confidence | Use with caution - monitor outcomes |
| Guadalupe | 88111 | 3.93 | High Confidence | Safe for algorithmic decisions |
| Hale | 50721 | 8.97 | Moderate Confidence | Use with caution - monitor outcomes |
| Hall | 43873 | 10.96 | Low Confidence | Requires manual review or additional data |
| Hamilton | 54890 | 17.27 | Low Confidence | Requires manual review or additional data |
| Hansford | 62350 | 19.70 | Low Confidence | Requires manual review or additional data |
| Hardeman | 60455 | 15.33 | Low Confidence | Requires manual review or additional data |
| Hardin | 70164 | 5.15 | Moderate Confidence | Use with caution - monitor outcomes |
| Harris | 70789 | 0.68 | High Confidence | Safe for algorithmic decisions |
| Harrison | 63427 | 4.81 | High Confidence | Safe for algorithmic decisions |
| Hartley | 78065 | 27.00 | Low Confidence | Requires manual review or additional data |
| Haskell | 52786 | 16.33 | Low Confidence | Requires manual review or additional data |
| Hays | 79990 | 3.73 | High Confidence | Safe for algorithmic decisions |
| Hemphill | 67798 | 27.70 | Low Confidence | Requires manual review or additional data |
| Henderson | 59778 | 4.35 | High Confidence | Safe for algorithmic decisions |
| Hidalgo | 49371 | 2.31 | High Confidence | Safe for algorithmic decisions |
| Hill | 60669 | 6.01 | Moderate Confidence | Use with caution - monitor outcomes |
| Hockley | 53283 | 7.51 | Moderate Confidence | Use with caution - monitor outcomes |
| Hood | 80013 | 4.72 | High Confidence | Safe for algorithmic decisions |
| Hopkins | 63766 | 5.67 | Moderate Confidence | Use with caution - monitor outcomes |
| Houston | 51043 | 10.50 | Low Confidence | Requires manual review or additional data |
| Howard | 67243 | 6.53 | Moderate Confidence | Use with caution - monitor outcomes |
| Hudspeth | 35163 | 23.20 | Low Confidence | Requires manual review or additional data |
| Hunt | 66885 | 4.17 | High Confidence | Safe for algorithmic decisions |
| Hutchinson | 62211 | 7.55 | Moderate Confidence | Use with caution - monitor outcomes |
| Irion | 54708 | 16.96 | Low Confidence | Requires manual review or additional data |
| Jack | 58861 | 13.22 | Low Confidence | Requires manual review or additional data |
| Jackson | 67176 | 17.91 | Low Confidence | Requires manual review or additional data |
| Jasper | 48818 | 9.84 | Moderate Confidence | Use with caution - monitor outcomes |
| Jeff Davis | 38125 | 66.11 | Low Confidence | Requires manual review or additional data |
| Jefferson | 57294 | 2.91 | High Confidence | Safe for algorithmic decisions |
| Jim Hogg | 42292 | 13.64 | Low Confidence | Requires manual review or additional data |
| Jim Wells | 46626 | 12.69 | Low Confidence | Requires manual review or additional data |
| Johnson | 77058 | 3.06 | High Confidence | Safe for algorithmic decisions |
| Jones | 59361 | 10.04 | Low Confidence | Requires manual review or additional data |
| Karnes | 57798 | 14.47 | Low Confidence | Requires manual review or additional data |
| Kaufman | 84075 | 4.11 | High Confidence | Safe for algorithmic decisions |
| Kendall | 104196 | 8.26 | Moderate Confidence | Use with caution - monitor outcomes |
| Kenedy | 45455 | 25.12 | Low Confidence | Requires manual review or additional data |
| Kent | 68553 | 15.57 | Low Confidence | Requires manual review or additional data |
| Kerr | 66713 | 6.23 | Moderate Confidence | Use with caution - monitor outcomes |
| Kimble | 62386 | 22.58 | Low Confidence | Requires manual review or additional data |
| King | 59375 | 49.51 | Low Confidence | Requires manual review or additional data |
| Kinney | 52386 | 45.29 | Low Confidence | Requires manual review or additional data |
| Kleberg | 52487 | 9.50 | Moderate Confidence | Use with caution - monitor outcomes |
| Knox | 48750 | 9.82 | Moderate Confidence | Use with caution - monitor outcomes |
| Lamar | 58246 | 4.88 | High Confidence | Safe for algorithmic decisions |
| Lamb | 54519 | 8.61 | Moderate Confidence | Use with caution - monitor outcomes |
| Lampasas | 73269 | 7.44 | Moderate Confidence | Use with caution - monitor outcomes |
| La Salle | 62798 | 26.04 | Low Confidence | Requires manual review or additional data |
| Lavaca | 58530 | 7.71 | Moderate Confidence | Use with caution - monitor outcomes |
| Lee | 66448 | 10.40 | Low Confidence | Requires manual review or additional data |
| Leon | 57363 | 12.04 | Low Confidence | Requires manual review or additional data |
| Liberty | 59605 | 6.69 | Moderate Confidence | Use with caution - monitor outcomes |
| Limestone | 53102 | 7.12 | Moderate Confidence | Use with caution - monitor outcomes |
| Lipscomb | 71625 | 12.92 | Low Confidence | Requires manual review or additional data |
| Live Oak | 55949 | 17.98 | Low Confidence | Requires manual review or additional data |
| Llano | 64241 | 8.81 | Moderate Confidence | Use with caution - monitor outcomes |
| Loving | NA | NA | NA | NA |
| Lubbock | 61911 | 3.54 | High Confidence | Safe for algorithmic decisions |
| Lynn | 52996 | 7.09 | Moderate Confidence | Use with caution - monitor outcomes |
| McCulloch | 53214 | 16.82 | Low Confidence | Requires manual review or additional data |
| McLennan | 59781 | 3.19 | High Confidence | Safe for algorithmic decisions |
| McMullen | 60313 | 41.18 | Low Confidence | Requires manual review or additional data |
| Madison | 65768 | 9.56 | Moderate Confidence | Use with caution - monitor outcomes |
| Marion | 48040 | 4.95 | High Confidence | Safe for algorithmic decisions |
| Martin | 70217 | 27.10 | Low Confidence | Requires manual review or additional data |
| Mason | 77583 | 15.82 | Low Confidence | Requires manual review or additional data |
| Matagorda | 56412 | 6.28 | Moderate Confidence | Use with caution - monitor outcomes |
| Maverick | 48497 | 10.10 | Low Confidence | Requires manual review or additional data |
| Medina | 73060 | 4.04 | High Confidence | Safe for algorithmic decisions |
| Menard | 40945 | 17.91 | Low Confidence | Requires manual review or additional data |
| Midland | 90123 | 5.80 | Moderate Confidence | Use with caution - monitor outcomes |
| Milam | 56985 | 5.92 | Moderate Confidence | Use with caution - monitor outcomes |
| Mills | 59315 | 9.26 | Moderate Confidence | Use with caution - monitor outcomes |
| Mitchell | 49869 | 12.52 | Low Confidence | Requires manual review or additional data |
| Montague | 63336 | 8.60 | Moderate Confidence | Use with caution - monitor outcomes |
| Montgomery | 95946 | 3.45 | High Confidence | Safe for algorithmic decisions |
| Moore | 59041 | 6.55 | Moderate Confidence | Use with caution - monitor outcomes |
| Morris | 51532 | 6.71 | Moderate Confidence | Use with caution - monitor outcomes |
| Motley | 66528 | 8.43 | Moderate Confidence | Use with caution - monitor outcomes |
| Nacogdoches | 51153 | 4.19 | High Confidence | Safe for algorithmic decisions |
| Navarro | 56261 | 7.70 | Moderate Confidence | Use with caution - monitor outcomes |
| Newton | 38871 | 16.91 | Low Confidence | Requires manual review or additional data |
| Nolan | 47437 | 7.72 | Moderate Confidence | Use with caution - monitor outcomes |
| Nueces | 64027 | 2.26 | High Confidence | Safe for algorithmic decisions |
| Ochiltree | 62240 | 17.77 | Low Confidence | Requires manual review or additional data |
| Oldham | 71103 | 11.17 | Low Confidence | Requires manual review or additional data |
| Orange | 71910 | 7.89 | Moderate Confidence | Use with caution - monitor outcomes |
| Palo Pinto | 65242 | 4.40 | High Confidence | Safe for algorithmic decisions |
| Panola | 58205 | 18.34 | Low Confidence | Requires manual review or additional data |
| Parker | 95721 | 3.91 | High Confidence | Safe for algorithmic decisions |
| Parmer | 65575 | 13.74 | Low Confidence | Requires manual review or additional data |
| Pecos | 59325 | 17.15 | Low Confidence | Requires manual review or additional data |
| Polk | 57315 | 5.19 | Moderate Confidence | Use with caution - monitor outcomes |
| Potter | 47974 | 3.95 | High Confidence | Safe for algorithmic decisions |
| Presidio | 29012 | 23.98 | Low Confidence | Requires manual review or additional data |
| Rains | 60291 | 10.78 | Low Confidence | Requires manual review or additional data |
| Randall | 78038 | 3.45 | High Confidence | Safe for algorithmic decisions |
| Reagan | 70319 | 12.82 | Low Confidence | Requires manual review or additional data |
| Real | 46842 | 32.96 | Low Confidence | Requires manual review or additional data |
| Red River | 44583 | 9.73 | Moderate Confidence | Use with caution - monitor outcomes |
| Reeves | 57487 | 22.06 | Low Confidence | Requires manual review or additional data |
| Refugio | 54304 | 4.88 | High Confidence | Safe for algorithmic decisions |
| Roberts | 62667 | 14.22 | Low Confidence | Requires manual review or additional data |
| Robertson | 59410 | 17.70 | Low Confidence | Requires manual review or additional data |
| Rockwall | 121303 | 3.78 | High Confidence | Safe for algorithmic decisions |
| Runnels | 55424 | 5.97 | Moderate Confidence | Use with caution - monitor outcomes |
| Rusk | 61661 | 9.85 | Moderate Confidence | Use with caution - monitor outcomes |
| Sabine | 47061 | 16.90 | Low Confidence | Requires manual review or additional data |
| San Augustine | 45888 | 9.69 | Moderate Confidence | Use with caution - monitor outcomes |
| San Jacinto | 54839 | 12.97 | Low Confidence | Requires manual review or additional data |
| San Patricio | 63842 | 6.88 | Moderate Confidence | Use with caution - monitor outcomes |
| San Saba | 54087 | 16.38 | Low Confidence | Requires manual review or additional data |
| Schleicher | 53774 | 15.15 | Low Confidence | Requires manual review or additional data |
| Scurry | 58932 | 21.39 | Low Confidence | Requires manual review or additional data |
| Shackelford | 60924 | 14.04 | Low Confidence | Requires manual review or additional data |
| Shelby | 49231 | 9.96 | Moderate Confidence | Use with caution - monitor outcomes |
| Sherman | 66169 | 27.93 | Low Confidence | Requires manual review or additional data |
| Smith | 69053 | 3.19 | High Confidence | Safe for algorithmic decisions |
| Somervell | 87899 | 33.73 | Low Confidence | Requires manual review or additional data |
| Starr | 35979 | 8.42 | Moderate Confidence | Use with caution - monitor outcomes |
| Stephens | 44712 | 18.85 | Low Confidence | Requires manual review or additional data |
| Sterling | 63558 | 22.25 | Low Confidence | Requires manual review or additional data |
| Stonewall | 66591 | 32.48 | Low Confidence | Requires manual review or additional data |
| Sutton | 56778 | 22.74 | Low Confidence | Requires manual review or additional data |
| Swisher | 40290 | 13.69 | Low Confidence | Requires manual review or additional data |
| Tarrant | 78872 | 1.01 | High Confidence | Safe for algorithmic decisions |
| Taylor | 61806 | 3.90 | High Confidence | Safe for algorithmic decisions |
| Terrell | 52813 | 21.03 | Low Confidence | Requires manual review or additional data |
| Terry | 42694 | 10.93 | Low Confidence | Requires manual review or additional data |
| Throckmorton | 55221 | 21.09 | Low Confidence | Requires manual review or additional data |
| Titus | 57634 | 8.12 | Moderate Confidence | Use with caution - monitor outcomes |
| Tom Green | 67215 | 4.68 | High Confidence | Safe for algorithmic decisions |
| Travis | 92731 | 1.19 | High Confidence | Safe for algorithmic decisions |
| Trinity | 51165 | 11.37 | Low Confidence | Requires manual review or additional data |
| Tyler | 50898 | 10.34 | Low Confidence | Requires manual review or additional data |
| Upshur | 60456 | 7.74 | Moderate Confidence | Use with caution - monitor outcomes |
| Upton | 55284 | 21.72 | Low Confidence | Requires manual review or additional data |
| Uvalde | 55000 | 15.28 | Low Confidence | Requires manual review or additional data |
| Val Verde | 57250 | 8.01 | Moderate Confidence | Use with caution - monitor outcomes |
| Van Zandt | 62334 | 8.10 | Moderate Confidence | Use with caution - monitor outcomes |
| Victoria | 66308 | 3.55 | High Confidence | Safe for algorithmic decisions |
| Walker | 47193 | 6.32 | Moderate Confidence | Use with caution - monitor outcomes |
| Waller | 71643 | 5.98 | Moderate Confidence | Use with caution - monitor outcomes |
| Ward | 70771 | 12.60 | Low Confidence | Requires manual review or additional data |
| Washington | 70043 | 9.41 | Moderate Confidence | Use with caution - monitor outcomes |
| Webb | 59984 | 3.11 | High Confidence | Safe for algorithmic decisions |
| Wharton | 59712 | 6.84 | Moderate Confidence | Use with caution - monitor outcomes |
| Wheeler | 58158 | 14.24 | Low Confidence | Requires manual review or additional data |
| Wichita | 58862 | 3.24 | High Confidence | Safe for algorithmic decisions |
| Wilbarger | 50769 | 18.36 | Low Confidence | Requires manual review or additional data |
| Willacy | 42839 | 13.05 | Low Confidence | Requires manual review or additional data |
| Williamson | 102851 | 1.42 | High Confidence | Safe for algorithmic decisions |
| Wilson | 89708 | 4.89 | High Confidence | Safe for algorithmic decisions |
| Winkler | 89155 | 16.06 | Low Confidence | Requires manual review or additional data |
| Wise | 85385 | 6.07 | Moderate Confidence | Use with caution - monitor outcomes |
| Wood | 61748 | 6.02 | Moderate Confidence | Use with caution - monitor outcomes |
| Yoakum | 80317 | 8.81 | Moderate Confidence | Use with caution - monitor outcomes |
| Young | 65565 | 16.86 | Low Confidence | Requires manual review or additional data |
| Zapata | 35061 | 10.72 | Low Confidence | Requires manual review or additional data |
| Zavala | 49243 | 28.78 | Low Confidence | Requires manual review or additional data |
Key Recommendations:
Your Task: Use your analysis results to provide specific guidance to the department.
Counties categorized as High Confidence (those with median income MOE below 5%) are suitable. These typically include large, urban counties with stable estimates, such as Harris or Dallas County. They are appropriate because their data is statistically stable, meaning resource allocations based on the income estimates are highly reliable (>95% certainty), minimizing the risk of systemic misallocation.
Counties categorized as Moderate Confidence (median income MOE between 5% and 10%) should be deployed with caution. The Department must implement real-time monitoring to compare algorithmic allocation decisions against actual service uptake and ground-level case worker reports. A 5%-10% uncertainty range requires human oversight to validate outcomes and trigger manual review if the allocation pattern appears to deviate significantly from expected need.
All Low Confidence counties (MOE>10%), especially those with extreme uncertainty like Jeff Davis (MOE≈66%) and Culberson (MOE≈53%), must have standard algorithmic application immediately suspended. For these areas, where tract-level data is universally unreliable and serves high Hispanic populations, the Department must transition to manual case worker review for needs assessment, or invest in small, targeted administrative surveys as a more reliable proxy for need than the ACS income estimates.
How do the tracts flagged with High MOE cluster geographically, and is there a measurable correlation between these clusters and existing disparities in access to Department resources or social infrastructure?
Beyond the Hispanic population, are High MOE issues equally prevalent across other small, concentrated minority groups (e.g., specific Native American or Asian subgroups), or is the sample size challenge uniquely impacting the Hispanic population in the Texas border regions?
Data Sources: - U.S. Census Bureau, American Community Survey 2018-2022 5-Year Estimates - Retrieved via tidycensus R package on September 19, 2025
Reproducibility: - All analysis conducted in R version 4.5.1 - Census API key required for replication - Complete code and documentation available at: https://github.com/MUSA-5080-Fall-2025/portfolio-setup-ChristineCui12
Methodology Notes:
The analysis employed several deliberate choices to focus the data quality evaluation. First, all estimates were sourced from the American Community Survey (ACS) 2018-2022 5-Year Estimates to ensure the most geographically detailed and statistically stable data available for tract-level analysis. Second, the three focus counties were intentionally selected based on the MOE analysis to represent the extremes of data reliability (High, Moderate, Low Confidence MOE ranges) rather than simply picking counties near the confidence boundaries. This was done to maximize the observed contrast in data quality patterns during the tract-level deep dive. Third, a conservative 15% MOE threshold was set for the tract-level demographic variables. If the MOE for any of the three major demographic groups (White, Black, Hispanic) exceeded this threshold, the entire tract was flagged as High MOE.
Limitations:
The findings are restricted to Texas and the three specifically selected counties. The conclusion that 100% of tracts in the focus area are unreliable is an extreme finding that may not be generalizable to counties in other parts of the state or country. Besides, the use of 5-Year ACS estimates means the data represents a rolling average from 2018 to 2022. This dampens volatility but may mask rapid economic changes or population shifts that have occurred since the beginning of the period.
Before submitting your portfolio link on Canvas:
Remember: Submit your portfolio URL on Canvas, not the file itself. Your assignment should be accessible at your-portfolio-url/assignments/assignment_1/your_file_name.html