We go from one form of place-based inequality in the last lab (segregation) to another form in this lab (gentrification). How do we determine whether a neighborhood is undergoing gentrification? Let’s find out using a quantitative approach. We’ll determine which neighborhoods are experiencing gentrification in the City of Fresno, an issue that has gotten a lot of local attention.

In this guide you will learn how to calculate the measure of gentrification used by Ding et al., (2016). This lab guide follows closely and supplements the material presented in class Handout 9.

Assignment 8 is due by 12:00 pm, May 28th on Canvas. See here for assignment guidelines. You must submit an .Rmd file and its associated .html file. Name the files: yourLastName_firstInitial_asgn08. For example: brazil_n_asgn08.

Open up a R Markdown file


Download the Lab template into an appropriate folder on your hard drive (preferably, a folder named ‘Lab 8’), open it in R Studio, and type and run your code there. The template is also located on Canvas under Files. The template is also located on Canvas under Files. Change the title (“Lab 8b”) and insert your name and date. Don’t change anything else inside the YAML (the stuff at the top in between the ---). Also keep the grey chunk after the YAML. For a rundown on the use of R Markdown, see the assignment guidelines.

Installing and loading packages


We will not be using any new packages in this lab. You’ll need to load the following packages. Unlike installing, you will always need to load packages whenever you start a new R session. As such, you’ll always need to use library() in your R Markdown file.

library(sf)
library(tidyverse)
library(tidycensus)
library(tigris)
library(tmap)
library(rmapshaper)
library(flextable)

Bringing in the data


We will use the Ding et al. (2016) method outlined in Handout 9 for measuring gentrification. The measure defines gentrification as follows

  • A neighborhood is eligible for gentrification if its median household income is less than the city’s median household income at the beginning of the observation period.
  • An eligible neighborhood gentrifies if its change from the beginning to the end of the observation period in median gross rent or median housing value is greater than the change in the city and the change in its percent of residents that have a college degree is more than the change in the city.

We will define the beginning of the period using 2006-2010 ACS data (this is the first ACS period using 2010 tract boundaries) and the end of the period using 2015-2019 ACS data. We need to bring in the appropriate variables from each of these years into R.

To demonstrate this measure, we will examine gentrification in the City of Fresno in California. We need to collect data at the tract and city levels.

Tract level data


Gentrification is a neighborhood socioeconomic change process. So we need to grab socioeconomic data from the beginning and end of the period for census tracts (the most common measure of neighborhood in the social sciences).

First, we will need to read in tract-level median gross rent, median housing value and percent with a college degree for Fresno city census tracts in the beginning of the period (2006-2010). We will use the get_acs() command to get these data. We won’t go through each line of code in detail because we’ve covered most of these operations and functions in prior labs. We’ve embedded comments within the code that briefly explains what each chunk is doing. Go back to prior guides (or RDS/GWR) if you need further help.

The one new function we use below is rename_with() which allows you to rename all the variables with names of a certain pattern. In this case, we rename variables with an “E” at the end, replacing “E” with nothing ““. Notice that I attached a 10 to the end of each variable that I will use in the gentrification measure to signify that it is for 2006-2010.

# Bring in 2006-2010 census tract data using the Census API 
ca.tracts10 <- get_acs(geography = "tract", 
              year = 2010,
              variables = c(medinc10 = "B19013_001", rent10  = "B25064_001",
                            houseval10 = "B25077_001", bachm = "B15002_015",
                            mastersm = "B15002_016", profm = "B15002_017",
                            phdm = "B15002_018", bachf = "B15002_032",
                            mastersf = "B15002_033", proff = "B15002_034",
                            phdf = "B15002_035", totcol = "B15002_001"),
              state = "CA",
              survey = "acs5",
              output = "wide",
              geometry = TRUE)

# Rename, calculate and keep essential vars. 
ca.tracts10 <- ca.tracts10 %>% 
  rename_with(~ sub("E$", "", .x), everything()) %>%
  mutate(pcol10 = 100*(bachm+mastersm+profm+phdm+bachf+mastersf+proff+phdf)/totcol) %>%
  select(c(GEOID,medinc10, rent10, houseval10, pcol10))

We then need to keep just the City of Fresno tracts. Here we use familiar friends that were introduced in Lab 6.

# Bring in city boundaries.  We'll use the boundaries for the final year, 2019
pl <- places(state = "CA", year = 2019, cb = TRUE)

# Keep Fresno city
fresno <- pl %>%
          filter(NAME == "Fresno")

#Keep tracts in Fresno 
fresno.tracts <- ms_clip(target = ca.tracts10, clip = fresno, remove_slivers = TRUE)

Second, we need to get the same variables but for the end of the period, which we define as 2015-2019. I’ve appended 19 to the end of each variable name to signify these are 2019 data. This is done to distinguish these variables from the 2010 ACS data, which we’ll be downloading later when we measure gentrification. Also note that we don’t need any spatial data since ca.tracts10 is already an sf object, so we don’t include the argument geometry = TRUE.

# Bring in 2015-2019 census tract data using the Census API 
ca.tracts19 <- get_acs(geography = "tract", 
              year = 2019,
              variables = c(tpop = "B03002_001", 
                            white = "B03002_003", black = "B03002_004",
                            asian = "B03002_006", hisp = "B03002_012",
                            medinc19 = "B19013_001", rent19  = "B25064_001",
                            houseval19 = "B25077_001", bach = "B15003_022",
                            masters = "B15003_023", prof = "B15003_024",
                            phd = "B15003_025", totcol = "B15003_001"),
              state = "CA",
              survey = "acs5",
              output = "wide")

# Rename, calculate and keep essential vars. 
ca.tracts19 <- ca.tracts19 %>% 
  rename_with(~ sub("E$", "", .x), everything()) %>%
  mutate(pwhite19 = 100*(white/tpop), pasian19 = 100*(asian/tpop), 
        pblack19 = 100*(black/tpop), phisp19 = 100*(hisp/tpop), 
        pcol19 = 100*(bach+masters+prof+phd)/totcol) %>%
  select(c(GEOID,pwhite19, pasian19, pblack19, phisp19,
           medinc19, rent19, houseval19, pcol19))

Finally, let’s join the two tract level data objects. The sf object fresno.tracts contains 2006-2010 ACS tract-level data. We join the regular tibble ca.tracts19, which contains 2015-2019 ACS tract-level data, using left_join(). The GEOID is GEOID in both fresno.tracts and ca.tracts19

fresno.tracts <- fresno.tracts %>%
            left_join(ca.tracts19, by = "GEOID")

Make sure we’ve successfully merged all the variables we need by viewing the object fresno.tracts.

glimpse(fresno.tracts)
## Rows: 127
## Columns: 14
## $ GEOID      <chr> "06019000501", "06019000800", "06019001202", "06019001407",…
## $ houseval10 <dbl> 89100, 276900, 131500, 161400, 237100, 32700, 124000, 11930…
## $ medinc10   <dbl> 18154, 25658, 33649, 20425, 76429, 26436, 22675, 26316, 361…
## $ rent10     <dbl> 673, 806, 819, 721, 1393, 761, 790, 734, 961, 791, 774, 856…
## $ pcol10     <dbl> 3.483769, 1.055409, 3.276284, 8.578856, 17.285531, 8.157895…
## $ pwhite19   <dbl> 8.175355, 15.866873, 1.447051, 9.691538, 24.370594, 13.6855…
## $ pasian19   <dbl> 2.527646, 4.798762, 12.869985, 17.507295, 30.278617, 7.4952…
## $ pblack19   <dbl> 6.990521, 8.668731, 5.108529, 11.588162, 1.074186, 10.28007…
## $ phisp19    <dbl> 80.92417, 69.89164, 79.91668, 59.14965, 41.62471, 66.48631,…
## $ medinc19   <dbl> 23717, 25083, 39223, 27258, 79750, 27031, 32295, 24322, 368…
## $ rent19     <dbl> 798, 686, 1060, 834, 1605, 888, 940, 766, 1203, 973, 868, 9…
## $ houseval19 <dbl> 179000, 358800, 149700, 68400, 255000, 80500, 93500, 115000…
## $ pcol19     <dbl> 1.870504, 11.528822, 6.269592, 6.982068, 20.195972, 3.94554…
## $ geometry   <GEOMETRY [°]> POLYGON ((-119.7763 36.7503..., MULTIPOLYGON (((-1…

City level data


As described in Handout 9, we need to compare neighborhoods to the entire City of Fresno to determine if a neighborhood is (1) eligible to gentrify and (2) if they are, whether they gentrified. This means we need to bring in city data for the beginning (2006-10) and end (2015-19) periods. First, let’s bring in 2006-2010 ACS data for all places in California using get_acs(). We use the argument geography = "place". We specify these data as place level 2006-2010 by attaching a c10 to the end of each variable. We can’t usegeometry = TRUE to bring in spatial data for 2006-2010 city data because tidycensus does not provide spatial data for years before 2011.

# Bring in census tract data using the Census API 
ca.places10 <- get_acs(geography = "place", 
              year = 2010,
              variables = c(medincc10 = "B19013_001", rentc10  = "B25064_001",
                            housevalc10 = "B25077_001", bachm = "B15002_015",
                            mastersm = "B15002_016", profm = "B15002_017",
                            phdm = "B15002_018", bachf = "B15002_032",
                            mastersf = "B15002_033", proff = "B15002_034",
                            phdf = "B15002_035", totcol = "B15002_001"),
              state = "CA",
              survey = "acs5",
              output = "wide")

# Keep Fresno. Calculate and keep essential vars. Also take out zero population tracts
ca.places10 <- ca.places10 %>% 
    filter(NAME == "Fresno city, California") %>%
  rename_with(~ sub("E$", "", .x), everything()) %>%
  mutate(pcolc10 = 100*(bachm+mastersm+profm+phdm+bachf+mastersf+proff+phdf)/totcol) %>%
  select(c(GEOID,medincc10, rentc10, housevalc10, pcolc10))


We next need to bring in the last year (2015-2019) of city level data. We specify the data as place level 2015-2019 by attaching a c19 to the end of each variable. Here, we specify geometry = TRUE to get city spatial boundaries, and then filter to get just Fresno city.

# Bring in census tract data using the Census API 
ca.places19 <- get_acs(geography = "place", 
              year = 2019,
              variables = c(medincc19 = "B19013_001", rentc19  = "B25064_001",
                            housevalc19 = "B25077_001", bachc = "B15003_022",
                            mastersc = "B15003_023", profc = "B15003_024",
                            phdc = "B15003_025", totcolc = "B15003_001"),
              state = "CA",
              survey = "acs5",
              output = "wide",
              geometry = TRUE)

# Keep Fresno. Calculate and keep essential vars. Also keep just Fresno
fresno.city <- ca.places19 %>% 
  filter(NAME == "Fresno city, California") %>%
  rename_with(~ sub("E$", "", .x), everything()) %>%
  mutate(pcolc19 = 100*(bachc+mastersc+profc+phdc)/totcolc) %>%
  select(c(GEOID,medincc19, rentc19, housevalc19, pcolc19))

Finally, let’s join the two city level data objects. The sf object fresno.city contains 2015-2019 ACS data just for Fresno city. We join the regular tibble ca.places10, which contains 2006-2010 ACS city-level data, to fresno.city using left_join(). The GEOID is GEOID in both files.

fresno.city <- fresno.city %>%
            left_join(ca.places10, by = "GEOID")

Make sure to look at the data.

glimpse(fresno.city)
## Rows: 1
## Columns: 10
## $ GEOID       <chr> "0627000"
## $ medincc19   <dbl> 50432
## $ rentc19     <dbl> 1005
## $ housevalc19 <dbl> 242000
## $ pcolc19     <dbl> 21.93261
## $ medincc10   <dbl> 43124
## $ rentc10     <dbl> 832
## $ housevalc10 <dbl> 244200
## $ pcolc10     <dbl> 20.51262
## $ geometry    <MULTIPOLYGON [°]> MULTIPOLYGON (((-119.6798 3...

Joining the tract and city data


We have tract and city level data at the beginning and end of the period in separate data objects (fresno.tracts and fresno.city). We now need to join all of them together into a single file.

When you view fresno.tracts, you’ll notice that we don’t have any variable indicating the city identifier to merge in city-level data. Instead of left_join(), we instead use the st_join() function to join the city data to the tract data, which is a part of the sf package. Here, rather than using an ID like GEOID, the joining is based on geographic location. We already used st_join() to join two spatial data object in Lab 8a. Let’s use it again.

fresno.tracts <- fresno.tracts %>%
                st_join(fresno.city, left=FALSE)

This function joins the variables from fresno.city to the data frame fresno.tracts.

Look at the variable names to see if the join was successful.

names(fresno.tracts)
##  [1] "GEOID.x"     "houseval10"  "medinc10"    "rent10"      "pcol10"     
##  [6] "pwhite19"    "pasian19"    "pblack19"    "phisp19"     "medinc19"   
## [11] "rent19"      "houseval19"  "pcol19"      "GEOID.y"     "medincc19"  
## [16] "rentc19"     "housevalc19" "pcolc19"     "medincc10"   "rentc10"    
## [21] "housevalc10" "pcolc10"     "geometry"

We can also look at the data.

glimpse(fresno.tracts)
## Rows: 127
## Columns: 23
## $ GEOID.x     <chr> "06019000501", "06019000800", "06019001202", "06019001407"…
## $ houseval10  <dbl> 89100, 276900, 131500, 161400, 237100, 32700, 124000, 1193…
## $ medinc10    <dbl> 18154, 25658, 33649, 20425, 76429, 26436, 22675, 26316, 36…
## $ rent10      <dbl> 673, 806, 819, 721, 1393, 761, 790, 734, 961, 791, 774, 85…
## $ pcol10      <dbl> 3.483769, 1.055409, 3.276284, 8.578856, 17.285531, 8.15789…
## $ pwhite19    <dbl> 8.175355, 15.866873, 1.447051, 9.691538, 24.370594, 13.685…
## $ pasian19    <dbl> 2.527646, 4.798762, 12.869985, 17.507295, 30.278617, 7.495…
## $ pblack19    <dbl> 6.990521, 8.668731, 5.108529, 11.588162, 1.074186, 10.2800…
## $ phisp19     <dbl> 80.92417, 69.89164, 79.91668, 59.14965, 41.62471, 66.48631…
## $ medinc19    <dbl> 23717, 25083, 39223, 27258, 79750, 27031, 32295, 24322, 36…
## $ rent19      <dbl> 798, 686, 1060, 834, 1605, 888, 940, 766, 1203, 973, 868, …
## $ houseval19  <dbl> 179000, 358800, 149700, 68400, 255000, 80500, 93500, 11500…
## $ pcol19      <dbl> 1.870504, 11.528822, 6.269592, 6.982068, 20.195972, 3.9455…
## $ GEOID.y     <chr> "0627000", "0627000", "0627000", "0627000", "0627000", "06…
## $ medincc19   <dbl> 50432, 50432, 50432, 50432, 50432, 50432, 50432, 50432, 50…
## $ rentc19     <dbl> 1005, 1005, 1005, 1005, 1005, 1005, 1005, 1005, 1005, 1005…
## $ housevalc19 <dbl> 242000, 242000, 242000, 242000, 242000, 242000, 242000, 24…
## $ pcolc19     <dbl> 21.93261, 21.93261, 21.93261, 21.93261, 21.93261, 21.93261…
## $ medincc10   <dbl> 43124, 43124, 43124, 43124, 43124, 43124, 43124, 43124, 43…
## $ rentc10     <dbl> 832, 832, 832, 832, 832, 832, 832, 832, 832, 832, 832, 832…
## $ housevalc10 <dbl> 244200, 244200, 244200, 244200, 244200, 244200, 244200, 24…
## $ pcolc10     <dbl> 20.51262, 20.51262, 20.51262, 20.51262, 20.51262, 20.51262…
## $ geometry    <GEOMETRY [°]> POLYGON ((-119.7763 36.7503..., MULTIPOLYGON (((-…

We’re done!

Measuring Gentrification


We’ve got all the data we need in one data object (fresno.tracts). Now we can start constructing our gentrification measure.

Gentrification eligible tracts


First, let’s determine whether a tract is eligible to gentrify in the first place. A neighborhood is eligible for gentrification if its median household income is less than the city’s median household income in 2006-2010. We create a variable named eligible that specifies whether a neighborhood is eligible or not using the ifelse() function within mutate().

fresno.tracts <- fresno.tracts %>%
            mutate(eligible = ifelse(medinc10 < medincc10, 
                                     "Eligible", 
                                     "Not Eligible"))

Here, the ifelse() function tells R that if a tract’s median household income medinc10 is less than the city level median household income medincc10, label that neighborhood “Eligible”. Otherwise, label the neighborhood as “Not Eligible”.

What is the percentage of neighborhoods that are eligible to gentrify? Here, we are summarizing a categorical variable, which we covered in Lab 4. Let’s put the results in a nice looking table by dropping the geometry using the function st_drop_geometry() and creating a nicely formatted table using the function flextable().

fresno.tracts %>%
  group_by(eligible) %>%
  summarize(n = n()) %>%
  mutate(Percent = 100*(n / sum(n))) %>%
  ungroup() %>%
  st_drop_geometry() %>%
  flextable() %>%
  colformat_double(digits = 2)

eligible

n

Percent

Eligible

71

55.91

Not Eligible

55

43.31

1

0.79

There is one tract missing a designation because it is missing values on median household income and median housing value (use summary(fresno.tracts) to detect missingness).

Gentrifying tracts


Next, we identify the eligible tracts that are gentrifying and the ones that are not gentrifying. An eligible neighborhood gentrifies if its change between 2006-2010 and 2015-2019 in median gross rent or median housing value is greater than the change in the city and the change in its percent of residents that have a college degree is more than the change in the city.

Let’s first calculate the tract-level 2010 to 2019 differences in rent rentch, housing values housech, and percent with a college degree pcolch.

fresno.tracts <- fresno.tracts %>%
            mutate(rentch = rent19-rent10, 
                   housech = houseval19-houseval10,
                   pcolch = pcol19-pcol10)

Next, we calculate the city-level differences in rent rentchc, housing values housechc, and percent with a college degree pcolchc.

fresno.tracts <- fresno.tracts %>%
            mutate(rentchc = rentc19-rentc10, 
                   housechc = housevalc19-housevalc10,
                   pcolchc = pcolc19-pcolc10)

We now have all the pieces to construct a variable gent that labels tracts as “Not eligible”, “Gentrifying”, and “Not Gentrifying”. We do this using a set of ifelse() statements within mutate().

fresno.tracts <- fresno.tracts %>%
            mutate(gent = ifelse(eligible == "Not Eligible", "Not Eligible",
                          ifelse(eligible == "Eligible" & pcolch > pcolchc &
                                (rentch > rentchc | housech > housechc), "Gentrifying",
                                "Not Gentrifying")))

Examining Gentrification


What is the percent of neighborhoods in Fresno that experienced gentrification? Let’s put it in a nice looking table by dropping the geometry using the function st_drop_geometry() and using the function flextable().

fresno.tracts %>%
  group_by(gent) %>%
  summarize(n = n()) %>%
  mutate(Percent = 100*(n / sum(n))) %>%
  ungroup() %>%
  st_drop_geometry() %>%
  flextable() %>%
  colformat_double(digits = 2)

gent

n

Percent

Gentrifying

22

17.32

Not Eligible

55

43.31

Not Gentrifying

49

38.58

1

0.79

Approximately 17 percent of neighborhoods in Fresno experienced gentrification between 2019 and 2019. Where are these neighborhoods located? Let’s make a nice color patch map using our best buddy tm_shape().

tm_shape(fresno.tracts, unit = "mi") +
  tm_polygons(fill = "gent", 
              fill.scale = tm_scale(style = "cat",
                                    values = "paired"),
              fill.legend = tm_legend(title = "", frame = FALSE),
              col_alpha = 0) +
  tm_scalebar(breaks = c(0, 1, 2), text.size  = 0.75, 
              position = tm_pos_in("right", "bottom"))  +
  tm_compass(type = "4star", position = tm_pos_in("right", "top")) +
  tm_title("2010-2019 Gentrification Status in Fresno") +
  tm_layout(frame = FALSE, scale = 0.7)

What are the mean percentages of residents that are Hispanic and non-Hispanic white, Black, and Asian in 2010 in each of the gentrification categories?

fresno.tracts %>%
  st_drop_geometry() %>%
  filter(is.na(gent) == FALSE) %>%
  group_by(gent) %>%
  summarize("% White" = mean(pwhite19),
            "% Black" = mean(pblack19),
            "% Hispanic" = mean(phisp19),
            "% Asian" = mean(pasian19)) %>%
  flextable() %>%
  colformat_double(digits = 2)

gent

% White

% Black

% Hispanic

% Asian

Gentrifying

15.86

7.26

63.41

11.45

Not Eligible

42.92

4.56

36.21

12.98

Not Gentrifying

17.33

8.19

60.81

10.81

Exploring the association between gentrification and racial/ethnic composition, demographic and socioeconomic characteristics, and other conditions is a great idea for your final project.

Assignment 8


Download and open the Assignment 8 R Markdown Script. The script can also be found on Canvas (Files - Week 8 - Assignment). Any response requiring a data analysis task must be supported by code you generate to produce your result. (Just examining your various objects in the “Environment” section of R Studio is insufficient—you must use scripted commands.).


  1. In Assignment 6, we descriptively examined the claim that Houston is the most racially integrated city in the United States. Let’s employ the segregation tools we learned this week to explore this claim even further. Let’s also compare Houston to Sacramento, a city that has also been proclaimed as among most racially diverse in the nation. Read in the Houston and Sacramento shapefiles houstondems.shp and sacdems.shp, which can be found on Canvas in the zipped folder assignment8q1.zip (Files - Week 8 - Assignment). These files contain data for census tract in the Houston and Sacramento metropolitan areas. Consider the datasets to be cleaned and ready for analysis. A record layout of the data can be found here.
  1. Calculate the Black/White, Hispanic/White and Asian/White Dissimilarity Indices for Houston and Sacramento. Present these values in a presentation-ready table(s). (3 points)
  2. Calculate the Black/White, Hispanic/White and Asian/White Interaction Indices for Houston and Sacramento. Present these values in a presentation-ready table(s). (3 points)
  3. Based on your answers to questions (a) and (b), which city is most segregated? Why? (2 points)
  4. Instead of examining segregation at the city level, let’s find where it exists at the neighborhood level. Calculate the Location Quotient for Racial Residential Segregation (LQRSS) for the Asian population for each city. (1 point)
  5. Show presentation-ready maps of the Asian LQRSS for each city. (2 points)
  6. Let’s examine the socioeconomic variables that may be correlated with neighborhood Asian concentration in each city. Calculate the correlation between the Asian LQRSS and percent of residents under 18 years old, percent of residents between 22 and 34, and percent foreign born in Houston. Do the same for Sacramento. Summarize the results in your own words, noting differences and similarities between the two cities. (2 points)


  1. We created a gentrification measure for the City of Fresno in Lab 8b. This question asks you to use the same method from Lab (Ding et al., 2016) to create a gentrification measure for the City of Sacramento, which has received a lot of local media attention for the rising presence of gentrification in certain neighborhoods. Download the file assignment8q2.zip from Canvas (Files - Week 8 - Assignment) into the appropriate folder on your hard drive. Unzip the file and read in the file saccitygent.shp, which is a shapefile of Sacramento census tract boundaries and demographic and socioeconomic variables. Consider the dataset to be cleaned and ready for analysis. A record layout of the data can be found here.

  1. Create a variable that determines whether a neighborhood is not eligible for gentrification, is eligible and is not gentrifying, and is eligible and is gentrifying using the Ding et al., measure. (3 points)

  2. Show a presentation-ready table of the percentages of tracts in Sacramento that were not eligible to gentrify, experienced gentrification and did not experience gentrification between 2010 and 2019. (2 points)

  3. Produce a presentation-ready map of the gentrification variable you created in (a). From the map, summarize the spatial patterns of gentrification in the city. (2 points)

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Website created and maintained by Noli Brazil