Economic Impact of COVID-19 by Race

An analysis of pandemic-related unemployment on Native American communities

The goal of this analysis is to better understand how the Covid-19 pandemic impacted unemployment by race. Specifically, I want to better understand whether the Native American and Native Alaskan community was disproportionately impacted by the job loss crisis of 2020.

To accomplish this, I analyzed unemployment data from the Bureau of Labor Statistics. Combined with county racial data from the pooled 2014-2018 American Community Surveys, it paints an interesting picture on how the pandemic recession impacted race disproportionately.

Findings

• Asian residents had the lowest unemployment in 2019
• Native American communities had the highest unemployment pre-pandemic
• The worst unemployment increase happened in densely-populated Asian counties
• White residents saw very little unemployment increases from the pandemic
• Native American communities saw the smallest increase in unemployment
• Black communities saw a large increase in unemployment
• Linear models showed a significant relationship between race and unemployment

library(tidyverse)
library(tidycensus)
library(sf)
library(ggplot2)
library(readxl)
library(viridis)
library(plotly)
library(ggpubr)
library(grid)
library(gridExtra)
library(kableExtra)
library(stargazer)


#loading themes and functions

mapTheme <- function(base_size = 12) {
  theme(
    text = element_text( color = "black"),
    plot.title = element_text(size = 16,colour = "black"),
    plot.subtitle=element_text(face="italic"),
    plot.caption=element_text(hjust=0),
    axis.ticks = element_blank(),
    panel.background = element_blank(),axis.title = element_blank(),
    axis.text = element_blank(),
    axis.title.x = element_blank(),
    axis.title.y = element_blank(),
    panel.grid.minor = element_blank(),
    panel.border = element_rect(colour = "black", fill=NA, size=2),
    strip.text.x = element_text(size = 14))
}

# Load Quantile break functions

qBr <- function(df, variable, rnd) {
  if (missing(rnd)) {
    as.character(quantile(round(df[[variable]],1),
                          c(.01,.2,.4,.6,.8), na.rm=T))
  } else if (rnd == FALSE | rnd == F) {
    as.character(formatC(quantile(df[[variable]]), digits = 2),
                 c(.01,.2,.4,.6,.8), na.rm=T)
  }
}

qBr2 <- function(df, variable, rnd) {
  if (missing(rnd)) {
    as.character(quantile(round(df[[variable]],digits = 0),
                                 c(.01,.2,.4,.6,.8), na.rm=T))
  } else if (rnd == FALSE | rnd == F) {
    as.character(quantile(round(df[[variable]]),digits = 1),
                        c(.01,.2,.4,.6,.8), na.rm=T)
  }
}

q5 <- function(variable) {as.factor(ntile(variable, 5))}

# Load hexadecimal color palette

blue5 <- c("#f0f9e8","#bae4bc","#7bccc4","#43a2ca","#0868ac")
purple5 <- c("#feebe2","#fbb4b9","#f768a1","#c51b8a","#7a0177")
red5 <- c("#ffffb2","#fecc5c","#fd8d3c","#f03b20","#bd0026")
diverge5 <- c("#1a9641","#a6d96a","#ffffbf","#fdae61","#d7191c")

American Community Surveys & Bureau of Labor Statistics

The Bureau of Labor Statistics publishes unemployment rates by county each year. The averages for 2019 and 2020 were combined with demographic information by county from the American Community Surveys from 2014-2018. I performed some feature engineering to determine the percent of residents living in a county by race and whether they represented a majority (>50%). I chose to focus on survey respondents who identify solely as one race. This better captures Native Americans living in designated tribal areas.

#pulling a variety of education, economic statistics
#focus is on Race, specifically Native American counties
county18 <- 
  get_acs(geography = "county", variables = c("B19013_001E","B25058_001E",
                                             "B06012_002E","C02003_001","C02003_002","C02003_003",
                                             "C02003_004","C02003_005","C02003_006","C02003_007",
                                             "C02003_008"), 
          year=2018, geometry=T, output="wide") %>%
  st_transform('ESRI:102005') %>%
  rename(MedHHInc = B19013_001E, MedRent = B25058_001E,
         TotalPoverty = B06012_002E, TotalPop = C02003_001E, TotalOneRace = C02003_002E,
         White = C02003_003E, AfrAmer = C02003_004E, AIAN = C02003_005E,
         Asian = C02003_006E, Hawaiian = C02003_007E, Other = C02003_008E) %>%
  dplyr::select(-starts_with("B"),-starts_with("C"))

#convert raw numbers into percents
ACSdata <- county18 %>%
  mutate(pctWhite = ifelse(TotalPop > 0, White / TotalPop,0), 
         pctAIAN = ifelse(TotalPop > 0, AIAN / TotalPop,0),
         pctBlack = ifelse(TotalPop > 0, AfrAmer / TotalPop,0),
         pctAsian = ifelse(TotalPop >0, Asian/TotalPop,0),
         pctPoverty = ifelse(TotalPop > 0, TotalPoverty / TotalPop, 0))


ACSdata$Majority_AIAN <- ifelse(ACSdata$pctAIAN > 0.5,1,0) %>%
  as.factor()

ACSdata$Majority_White <- ifelse(ACSdata$pctWhite > 0.5,1,0) %>%
  as.factor()

ACSdata$Majority_Black <- ifelse(ACSdata$pctBlack > 0.5,1,0) %>%
  as.factor()

#Multiple by 100, important for later when we
#interpret coefficients in our linear models

ACSdata$pctAIAN100 <- ACSdata$pctAIAN*100
ACSdata$pctWhite100 <- ACSdata$pctWhite*100
ACSdata$pctBlack100 <- ACSdata$pctBlack*100
ACSdata$pctAsian100 <- ACSdata$pctAsian*100

#reading in Bureau of Labor statistics from 2019 and 2020
#these unemployment stats have already been averaged
#assuming that 2019 is pre-pandemic, 2020 is pandemic
BLS2019 <- read_excel("county_unemploy_avg_2019.xlsx")
BLS2020 <- read_excel("county_unemploy_avg_2020.xlsx")

BLS <- rbind(BLS2019, BLS2020)

#combine state and county FIPS
BLS$GEOID <- paste0(BLS$State_FIPS_Code,BLS$County_FIPS_Code)

#removes NAs
BLS <- BLS %>%
  drop_na("State_FIPS_Code")

#creates the unemploy_YOY column
BLS <- BLS %>%
  gather(Variable,Value, -Year, -GEOID) %>%
  filter(Variable == "Unemploy_Rate") %>%
  spread(Year, Value) %>%
  mutate_at(c(3,4),as.numeric) %>%
  mutate(Unemploy_YOY = .[[4]] - .[[3]]) %>%
  select(GEOID, Unemploy_YOY) %>%
  left_join(.,BLS,by = "GEOID")


#merging datasets
merged <- ACSdata %>%
  right_join(.,BLS)

#Removing HI, AK, PR for mapping
#the main dataset, merged, was used in all linear modeling

lower48 <- merged[!(merged$State_FIPS_Code == "02" | merged$State_FIPS_Code == "15" | merged$State_FIPS_Code == "72"),]

Alaska <- merged[(merged$State_FIPS_Code == "02"),]

Hawaii <- merged[(merged$State_FIPS_Code == "15"),]



#Remember to filter by years!



test2019 <- lower48 %>%
  filter(Year == 2019)

test2020 <- lower48 %>%
  filter(Year == 2020)

Visualizing Unemployment

Unemployment in 2019 was concentrated in Appalachia, the Deep South, the Southwest and Northwest. By 2020, these same areas witnessed unemployment of 8% or higher.

However, other parts of America were affected worse by the pandemic’s economic fallout. Coastal counties in the Northeast, Michigan, California, and the Rio Grande saw the highest rate of unemployment increases - more than 4% in on year.

#2019 unemployment rate by county
ggplot() +
  geom_sf(data = test2019, aes(fill = q5(Unemploy_Rate))) +
  scale_fill_manual(values = blue5,
                  labels = paste(qBr2(test2019, "Unemploy_Rate"),"%"),
                  name = "Unemployment Rate\n(Quintile Breaks)") +
  labs(title = "US Unemployment Rate by County", subtitle = "2019",
       caption = "Source: Bureau of Labor Statistics, 2019") +
  mapTheme() + theme(plot.title = element_text(size=22))

#2020 unemployment rate by county
ggplot() +
  geom_sf(data = test2020, aes(fill = q5(Unemploy_Rate))) +
  scale_fill_manual(values = purple5,
                    labels = paste(qBr2(test2020, "Unemploy_Rate"),"%"),
                    name = "Unemployment Rate\n(Quintile Breaks)") +
  labs(title = "US Unemployment Rate by County", subtitle = "2020",
       caption = "Source: Bureau of Labor Statistics, 2020") +
  mapTheme() + theme(plot.title = element_text(size=22))

#2020-2019 unemployment rate by county
ggplot() +
  geom_sf(data = test2020, aes(fill = q5(Unemploy_YOY))) +
  scale_fill_manual(values = diverge5,
                    labels = paste("+",qBr2(test2020, "Unemploy_YOY"),"%"),
                    name = "Unemployment Rate Change\n(Quintile Breaks)") +
  labs(title = "US Unemployment Rate Change by County", subtitle = "2020 - 2019",
       caption = "Source: Bureau of Labor Statistics, 2019-2020") +
  mapTheme() + theme(plot.title = element_text(size=22))

Visualizing Tribal Density

These maps show the density of Native Americans across the United States. New Mexico, Arizona, and California currently have the highest percentage of Native Americans by county. Alaska and Hawaii also have a large native population. Visually, these counties don’t appear strongly correlated with the previous pandemic unemloyment maps.

Yellow counties in the lower 48 states had roughly 80,000 residents identify solely as Native American.

#AIAN Lower 48 Population Density
ggplot() +
  geom_sf(data = test2020) +
  geom_sf(data = test2020, aes(fill = AIAN), color = NA) +
  scale_fill_viridis()+
  labs(title = "Native American Population by County", 
       caption = "Source: American Community Surveys, 2014-2018\nAmericans who identified solely as Native American") +
  mapTheme()

#Alaskan Native Population Density
ggplot() +
  geom_sf(data = Alaska) +
  geom_sf(data = Alaska, aes(fill = AIAN), color = NA) +
  scale_fill_viridis()+
  labs(title = "Alaskan Native Population by County", 
       caption = "Source: American Community Surveys, 2014-2018\nAmericans who identified solely as Alaskan Native") +
  mapTheme()

#Hawaiian Native Population Density
ggplot() +
  geom_sf(data = Hawaii) +
  geom_sf(data = Hawaii, aes(fill = AIAN), color = NA) +
  scale_fill_viridis()+
  labs(title = "Hawaiian Native Population by County", 
       caption = "Source: American Community Surveys, 2014-2018\nAmericans who identified solely as Native Hawaiian") +
  mapTheme()

Charting Unemployment and Race

This box plot displays the drastic increase in unemployment from 2019 to 2020. Median unemployment rose from 3.7% to 6.5% by county. All quartiles also increased.

#Unemployment 2019 v. 2020
box1 <- ggplot(merged, aes(x=Year, y=Unemploy_Rate, fill=Year)) + geom_boxplot() +
  labs(title = "US County Unemployment (%)",
       subtitle = "2019 vs. 2020",
       caption = "Source: BLS (2019,2020")

box1 %>% ggplotly()

The following histograms show the count of counties by racial composition. Note that there are about 3100 counties in the U.S. The average county has about 80% white residents and 15% black residents. Counties have very few Native American and Asian residents.

#County Spread AIAN
ggplot(merged %>%
         filter(Year == 2020), aes(x=pctAIAN)) + geom_histogram(bins = 40, fill = "steelblue") +theme_classic() +
  geom_vline(aes(xintercept = mean(pctAIAN)), col='black', size=0.5) +
  labs(title = "Histogram of County Demographics",
       subtitle = "County Percent American Indian/ Alaskan Native Residents",
       caption = "Source: BLS (2019-2020), ACS (2014-2018)")

#County Spread White
ggplot(merged %>%
         filter(Year == 2020), aes(x=pctWhite)) + geom_histogram(bins = 40, fill = "red") +theme_classic() + 
  geom_vline(aes(xintercept = mean(pctWhite)), col='black', size=0.5) +
  labs(title = "Histogram of County Demographics",
       subtitle = "County Percent White Residents",
       caption = "Source: BLS (2019-2020), ACS (2014-2018)")

#County Spread Black
ggplot(merged %>%
         filter(Year == 2020), aes(x=pctBlack)) + geom_histogram(bins = 40, fill = "green")+theme_classic()+ 
  geom_vline(aes(xintercept = mean(pctBlack)), col='black', size=0.5) +
  labs(title = "Histogram of County Demographics",
       subtitle = "County Percent Black Residents",
       caption = "Source: BLS (2019-2020), ACS (2014-2018)")

#County Spread Asian
ggplot(merged %>%
         filter(Year == 2020), aes(x=pctAsian)) + geom_histogram(bins = 40, fill = "orange") +
  theme_classic() + geom_vline(aes(xintercept = mean(pctAsian)), col='black', size=0.5) +
  labs(title = "Histogram of County Demographics",
       subtitle = "County Percent Asian Residents",
       caption = "Source: BLS (2019-2020), ACS (2014-2018)")

Scatterplots show the correlation between unemployment and race. Regression line slopes display the strength of relationship between a particular’s county’s racial composition and unemployment. A positive slope implies increasing unemployment, while a negative slope implies lower unemployment at higher racial densities.

Findings:
1. In 2019, Asian residents had the lowest unemployment
      + In 2020, they had the highest unemployment slopes
2. High density Native American counties had the largest unemployment(8-10%)
      + However, this relationship wasn’t as drastic in 2020
3. White counties had the lowest unemployment in 2020
4. Black counties struggles with unemployment only increased from 2019 to 2020

In 2020, a 10% increase in a county’s Native American composition is associated with a .3% increase in unemployment

#Unemployment 2019 and 2020 by AIAN
ggplot(data = merged, aes(x = pctAIAN, y = Unemploy_Rate)) +
  geom_point()+
  geom_smooth(method='lm', formula= y~x)+
  stat_regline_equation(label.y = 20,label.x = .25, aes(label = ..eq.label..)) +
  stat_regline_equation(label.y = 18,label.x = .25, aes(label = ..rr.label..))+
  facet_wrap(~ Year, nrow=2) +
  labs(title = "Unemployment by Year",
       subtitle = "County Percent American Indian/ Alaskan Native Residents",
       caption = "Source: BLS (2019-2020), ACS (2014-2018)")

#Unemployment 2019 and 2020 by White
ggplot(data = merged, aes(x = pctWhite, y = Unemploy_Rate)) +
  geom_point()+
  geom_smooth(method='lm', formula= y~x)+
  stat_regline_equation(label.y = 20,label.x = .25, aes(label = ..eq.label..)) +
  stat_regline_equation(label.y = 18,label.x = .25, aes(label = ..rr.label..))+
  facet_wrap(~ Year, nrow=2) +
  labs(title = "Unemployment by Year",
       subtitle = "County Percent White Residents",
       caption = "Source: BLS (2019-2020), ACS (2014-2018)")

#Unemployment 2019 and 2020 by Black
ggplot(data = merged, aes(x = pctBlack, y = Unemploy_Rate)) +
  geom_point()+
  geom_smooth(method='lm', formula= y~x)+
  stat_regline_equation(label.y = 20,label.x = .25, aes(label = ..eq.label..)) +
  stat_regline_equation(label.y = 18,label.x = .25, aes(label = ..rr.label..))+
  facet_wrap(~ Year, nrow=2) +
  labs(title = "Unemployment by Year",
       subtitle = "County Percent Black Residents",
       caption = "Source: BLS (2019-2020), ACS (2014-2018)")

#Unemployment 2019 and 2020 by Asian
#no counties where Asian population is more than 50%
ggplot(data = merged, aes(x = pctAsian, y = Unemploy_Rate)) +
  geom_point()+
  geom_smooth(method='lm', formula= y~x)+
  stat_regline_equation(label.y = 20,label.x = .2, aes(label = ..eq.label..)) +
  stat_regline_equation(label.y = 18,label.x = .2, aes(label = ..rr.label..))+
  xlim(0,0.5) +
  facet_wrap(~ Year, nrow=2) +
  labs(title = "Unemployment by Year",
       subtitle = "County Percent Asian Residents",
       caption = "Source: BLS (2019-2020), ACS (2014-2018)")

Another way to examine this relationship is by the unemployment change year-over-year. This truly gets to the heart of the study - were certain racial groups disproportionately impacted by unemployment from the pandemic?

The largest negative slope belongs to the Native American & Native Alaskan communities, followed by counties with more white residents. Both of these groups witnessed the smallest increase in unemployment from 2019 to 2020 relative to other racial groups.

The largest increase in pandemic-related unemployment was for Asian residents. The 2nd largest increase in unemployment by county was for Black residents.

A 10% increase in a county’s composition of Asian residents was associated with a 1.7% increase in unemployment from 2019 to 2020

That is a massive increase for Asian citizens. However, I need to prove that these relationships were statistically significant.

#Unemploy_YOY by Race
grid.arrange(ncol = 2, top = textGrob("Unemployment Change by County Race (2020 - 2019)",
                                      gp=gpar(fontsize=20,font=3)),
  ggplot(data = merged %>%
           filter(Year == 2020), aes(x = pctAIAN, y = Unemploy_YOY)) +
    geom_point()+
    geom_smooth(method='lm', formula= y~x)+
    stat_regline_equation(label.y = 15,label.x = .15, aes(label = ..eq.label..)) +
    stat_regline_equation(label.y = 13,label.x = .15, aes(label = ..rr.label..))+
    xlim(0,1.0),
  ggplot(data = merged %>%
           filter(Year == 2020), aes(x = pctWhite, y = Unemploy_YOY)) +
    geom_point()+
    geom_smooth(method='lm', formula= y~x)+
    stat_regline_equation(label.y = 15,label.x = .15, aes(label = ..eq.label..)) +
    stat_regline_equation(label.y = 13,label.x = .15, aes(label = ..rr.label..)),
  ggplot(data = merged %>%
           filter(Year == 2020), aes(x = pctBlack, y = Unemploy_YOY)) +
    geom_point()+
    geom_smooth(method='lm', formula= y~x)+
    stat_regline_equation(label.y = 15,label.x = .15, aes(label = ..eq.label..)) +
    stat_regline_equation(label.y = 13,label.x = .15, aes(label = ..rr.label..))+
    xlim(0,1.0),
  ggplot(data = merged %>%
           filter(Year == 2020), aes(x = pctAsian, y = Unemploy_YOY)) +
    geom_point()+
    geom_smooth(method='lm', formula= y~x)+
    stat_regline_equation(label.y = 15,label.x = .15, aes(label = ..eq.label..)) +
    stat_regline_equation(label.y = 13,label.x = .15, aes(label = ..rr.label..))+
    xlim(0,0.5) + 
    labs(caption = "Source: BLS (2019-2020), ACS (2014-2018)"))

Baseline Unemployment

Majority Native American counties (>50% of population) had the highest pre-pandemic unemployment at 7.5%. This increased by 1.2% as as result of the pandemic. This was a minor increase compared to majority white counties (+2.7%) and majority Black counties (+3.5%). This is displayed again in the second chart where I examine year over year increases. Note that no counties in the U.S. are majority Asian.

merged %>%
  st_drop_geometry() %>%
  select(Year, Unemploy_Rate, Majority_AIAN, Majority_White, Majority_Black) %>%
  gather(Variable, Value, -Majority_AIAN, -Majority_White, -Majority_Black, -Year) %>%
  mutate(Value = as.numeric(Value)) %>%
  group_by(Year, Majority_AIAN, Majority_White, Majority_Black) %>%
  summarise(mean = mean(Value, na.rm = T)) %>%
  spread(Year, mean) %>%
  kable(caption = "Unemployment 2019 - 2020 (%)") %>%
  kable_styling("striped", full_width = F)

Unemployment 2019 - 2020 (%)

Majority_AIAN	Majority_White	Majority_Black	2019	2020
0	0	0	5.068919	8.869231
0	0	1	6.079208	9.283505
0	1	0	3.973475	6.594917
1	0	0	7.467857	8.650000

Majority Native American counties increased in unemployment by 1.2% vs. 2.8% for when they were the minority race

foo <- merged %>%
  st_drop_geometry() %>%
  select(Majority_AIAN, Unemploy_YOY) %>%
  group_by(Majority_AIAN) %>%
  summarize(meanAIAN = mean(Unemploy_YOY, na.rm=T))

foo2 <- merged %>%
  st_drop_geometry() %>%
  select(Majority_White, Unemploy_YOY) %>%
  group_by(Majority_White) %>%
  summarize(meanWhite = mean(Unemploy_YOY, na.rm=T))

foo3 <- merged %>%
  st_drop_geometry() %>%
  select(Majority_Black, Unemploy_YOY) %>%
  group_by(Majority_Black) %>%
  summarize(meanBlack = mean(Unemploy_YOY, na.rm=T))

cbind(foo,foo2,foo3) %>%
  dplyr::select(-starts_with("Majority")) %>%
  round(., digits = 2) %>%
  set_rownames(c("Minority","Majority")) %>%
  set_colnames(c("AIAN", "White", "Black")) %>%
  kable(caption = "Increase in Unemployment Year-Over-Year (%)") %>%
  kable_styling("striped", full_width = F)

Increase in Unemployment Year-Over-Year (%)

	AIAN	White	Black
Minority	2.81	3.48	2.77
Majority	1.18	2.75	3.50

Modeling

At this point it’s evident that the economic fallout from the pandemic was not spread evenly across the country. This final analysis incorporates linear modeling to better understand whether this relationship was statistically significant.

For example, in 2019, for each additional percent of Native Americans in a county, the unemployment rate fell by only .04% - much less than other races. For each additional percent of Asian residents in a county in 2019, unemployment fell by 0.22%.

Model 3 shows that Asian communities were hit hardest economically by the pandemic. Followed by Black, then white, then Native communities. All models were statistically significant, proving that their is a relationship between race and county unemployment.

For each additional percent of Native Americans in a county, the unemployment rise year over year falls by .07% - implying that these communities were hit less hard by the pandemic job loss than other counties

#2019 county unemployment by Race
m1 <- lm(Unemploy_Rate ~ pctAIAN100 + pctWhite100 + pctBlack100 + pctAsian100, 
         data = merged %>%
           st_drop_geometry() %>%
           filter(Year == 2019))

summary(m1)

#2020 county unemployment by Race
m2 <- lm(Unemploy_Rate ~ pctAIAN100 + pctWhite100 + pctBlack100 + pctAsian100, 
         data = merged %>%
           st_drop_geometry() %>%
           filter(Year == 2020))

summary(m2)

#YOY County Unemploymnet Change by Race
m3 <- lm(Unemploy_YOY ~ pctAIAN100 + pctWhite100 + pctBlack100 + pctAsian100, 
         data = merged %>%
           st_drop_geometry() %>%
           filter(Year == 2020))

summary(m3)

stargazer(m1,m2,m3, type="html",align=TRUE, 
                                       single.row=TRUE,report="vc*",se=NULL,digits=2, dep.var.caption = "",
                                       title = "Relationship between Unemployment and County Race",
                                       column.labels=c("2019","2020","YOY Change"))

Relationship between Unemployment and County Race

	Unemploy_Rate		Unemploy_YOY
	2019	2020	YOY Change
	(1)	(2)	(3)
pctAIAN100	-0.04***	-0.03**	-0.07***
pctWhite100	-0.09***	-0.06***	-0.05***
pctBlack100	-0.06***	-0.02**	-0.04***
pctAsian100	-0.22***	0.01	0.08***
Constant	12.52***	12.18***	7.52***
Observations	3,219	3,141	3,141
R2	0.19	0.10	0.13
Adjusted R2	0.19	0.10	0.13
Residual Std. Error	1.61 (df = 3214)	2.16 (df = 3136)	1.49 (df = 3136)
F Statistic	187.34*** (df = 4; 3214)	91.57*** (df = 4; 3136)	114.67*** (df = 4; 3136)
Note:	p<0.1; p<0.05; p<0.01

The final three models examine the relationship between majority race in each county and the increase in unemployment from 2019 to 2020.

Majority AIAN counties had unemployment rate changes 1.63% less than other counties

Majority white counties had an unemployment increase 0.73% lower than minority white counties - the opposite is true for majority Black counties.

#Majority AIAN
m4 <- lm(Unemploy_YOY ~ Majority_AIAN,
         data = merged %>%
           st_drop_geometry() %>%
           filter(Year == 2020))
summary(m4)

#Majority White
m5 <- lm(Unemploy_YOY ~ Majority_White,
         data = merged %>%
           st_drop_geometry() %>%
           filter(Year == 2020))
summary(m5)

#Majority Black
m6 <- lm(Unemploy_YOY ~ Majority_Black,
         data = merged %>%
           st_drop_geometry() %>%
           filter(Year == 2020))
summary(m6)

stargazer(m4,m5,m6, type="html",align=TRUE, 
                                       single.row=TRUE,report="vc*",se=NULL,digits=2, dep.var.caption = "",
                                       title = "Relationship between Unemployment and Majority Race",
                                       column.labels=c("AIAN","White","Black"))

Relationship between Unemployment and Majority Race

	Unemploy_YOY
	AIAN	White	Black
	(1)	(2)	(3)
Majority_AIAN1	-1.63***
Majority_White1		-0.73***
Majority_Black1			0.72***
Constant	2.81***	3.48***	2.77***
Observations	3,141	3,141	3,141
R2	0.01	0.01	0.01
Adjusted R2	0.01	0.01	0.01
Residual Std. Error (df = 3139)	1.58	1.58	1.59
F Statistic (df = 1; 3139)	29.30***	37.58***	19.60***
Note:	p<0.1; p<0.05; p<0.01

Policy Implications

Majority Black counties had an unemployment rate more than 2% the national median in 2019. They also had the 2nd highest increase in job loss resulting from the pandemic. Policymakers must better understand the systemic barriers to unemployment for the African American community. This was a problem before the pandemic and only exacerbated by the recent economic recession.

It is also important to ensure that businesses employing Asian Americans are able to recover, since they represented the largest job loss by race.

Finally, Native Americans have the highest unemployment by race in this country. This was true before the pandemic and even moreso in 2020. Policymakers need to better understand the unemployment and working demands of this community if we want to forge a better, more equitable society.

I hope that these findings influence future policy action in targeting specific hiring and job creation for struggling counties.