Exploratory analysis

Homelessness Risk

Author

Laurie Platt & Olumide Ogunyanwo

Published

Last updated 20 June, 2025

Code 
import os
import locale
import datetime
import pandas as pd
from pandas.api.types import CategoricalDtype
import numpy as np
import hl_risk as hl
# https://github.com/ymrohit/ukpostcodeio
from ukpostcodeio.client import UKPostCodeIO
import geopandas as gpd
from plotnine import ggplot, aes, geom_line, facet_wrap, labs, theme_classic, geom_boxplot, geom_bar,geom_col,element_text, scale_y_continuous, theme, position_dodge, scale_fill_brewer
import warnings  
warnings.filterwarnings('ignore')

# Load the homelessness case & household membership dataframes
hl_case = hl.load_hl_case()
hl_case_hh = hl.load_hl_case_hh()

# Load the homelessness client and high level support client dataframes 
hl_client = hl.load_hl_client()
hl_high = hl.load_hl_high()

# Load the homeless households & people dataframes
hl_hh = hl.load_hl_hh()
hl_person = hl.load_hl_person()

# Number of rows in each dataframe
n_hl_case = hl_case.shape[0]
n_hl_case_hh = hl_case_hh.shape[0]

# Earliest and last case
case_earliest_date = hl_case["application_date"].min()
case_last_date = hl_case["application_date"].max()

We have data for 21,449 homelessness cases, and 32,344 records relating to the household members of homelessness cases. Our earliest homelessness case in the data is 01/04/2019 and the last is 31/03/2025.

1 Last settled postcode

What locations (last settled postcode) have the highest rates of homelessness?

Code 
# Cases where last_settled_postcode is not empty
hl_case_lsp = hl_case[hl_case["last_settled_postcode"].notnull()] 

n_hl_case_lsp = hl_case_lsp.shape[0]
pct_hl_case_lsp = round(n_hl_case_lsp/n_hl_case, 2)

Only 38% of homelessness cases have a last settled postcode value (we’ve not checked yet whether the postcode value is valid). Stephen mentioned this should be better in more recent years.

Code 
# New boolean field indicating if last settled postcode is empty
hl_case["lsp_null"] = hl_case["last_settled_postcode"].isnull()

# Annual count of cases where last_settled_postcode is empty and not empty
hl_case_lsp_yr = (
  hl_case[["hra_case_id", "application_year", "lsp_null"]]
    .groupby(by=["application_year", "lsp_null"])
    .size()
    .reset_index(name='cases')
)

# Reshape the table
hl_case_lsp_yr = (
  hl_case_lsp_yr
    .pivot(index="application_year", columns="lsp_null", values="cases")
    .reset_index()
    .fillna(0)
    .astype({
      True: "int", 
      False: "int"
    })
)

# Rename the columns
hl_case_lsp_yr = hl_case_lsp_yr.rename(columns = {
    "application_year": "year",
    False: "with_postcode",
    True: "no_postcode"
})

# Total number of cases
hl_case_lsp_yr["total_cases"] = (
  hl_case_lsp_yr["with_postcode"] + hl_case_lsp_yr["no_postcode"]
)

# Percentage of cases with a postcode  
hl_case_lsp_yr["pct_with_postcode"] = (
  round((hl_case_lsp_yr["with_postcode"] / hl_case_lsp_yr["total_cases"]), 2)*100
)
hl_case_lsp_yr = hl_case_lsp_yr.astype({"pct_with_postcode": "int"})
hl_case_lsp_yr = hl_case_lsp_yr.astype({"pct_with_postcode": "str"})
hl_case_lsp_yr["pct_with_postcode"] = hl_case_lsp_yr["pct_with_postcode"] + "%"

(
  hl_case_lsp_yr[["year", "pct_with_postcode"]]
    .style
      .hide(axis = "index")
      .set_properties(subset = ["pct_with_postcode"], **{"text-align": "right"})
)
Homelessness cases with last settled postcode
year pct_with_postcode
2019 0%
2020 0%
2021 0%
2022 31%
2023 74%
2024 87%
2025 74%
Code 
# List of postcodes 
# postcodes = hl_case_lsp["last_settled_postcode"].unique()[:230].tolist() # for testing
postcodes = hl_case_lsp["last_settled_postcode"].unique().tolist()

# Initialize the client with default settings
postcode_client = UKPostCodeIO()

# Initialize the client with custom timeout and retries
postcode_client = UKPostCodeIO(timeout=10, max_retries=5)

# postcodes.io has a maximum of 100 postcode bulk lookups at a time
# Loop through 100 postcodes at a time
n_bulk_max = 100
n_postcodes = len(postcodes)
bulk_loops = n_postcodes // n_bulk_max
bulk_mod = n_postcodes % n_bulk_max
if bulk_mod > 0: 
  bulk_loops += 1
else:
  bulk_mod = n_bulk_max

# Bulk lookup postcodes
for i in range(bulk_loops):
  j = i + 1
  batch_start = j * n_bulk_max - n_bulk_max 
  if j < bulk_loops:
    batch_end = j * n_bulk_max
  else:
    batch_end = batch_start + bulk_mod
  postcode_batch = postcodes[batch_start:batch_end]
  batch_results = postcode_client.bulk_lookup_postcodes(postcode_batch)
  if j == 1:
    results = batch_results
  else: 
    results += batch_results

# Good practice to close the client session when done
postcode_client.close()

# Flatten results and extract the ward  
results = pd.DataFrame(results)
results = results.result.apply(pd.Series)
results["admin_ward_code"] = results['codes'].str.get('admin_ward')
results = results[[
  "postcode",
  "admin_district",
  "pfa",
  "admin_ward",
  "admin_ward_code"
]]

# We searched on unique postcodes so need to rejoin
hl_case_lsp = hl_case_lsp.merge(
    right = results,
    how = "left", 
    left_on = "last_settled_postcode",
    right_on = "postcode"
)

# Identify invalid postcode and postcode outside of Sheffield
admin_district_ft = (hl_case_lsp[["admin_district", "pfa"]]
    .value_counts(dropna = False)
    .rename("cases")
    .reset_index()
    .fillna("Invalid postcode")
)

# Label neighbouring districts as South Yorkshire 
admin_district_ft.loc[
    (admin_district_ft['pfa'] == "South Yorkshire") &
    (admin_district_ft['admin_district'] != "Sheffield"), 
    'admin_district'] = "South Yorkshire"

# Label "other" districts
admin_district_ft.loc[
    ~admin_district_ft['admin_district'].isin([
        "Invalid postcode",
        "Sheffield",
        "South Yorkshire"
    ]),
    'admin_district'] = "Other"

# Group and re-sum
admin_district_ft = (
    admin_district_ft
        .groupby("admin_district")
        .sum("cases")
        .reset_index(names="admin_district")
)

# Change admin_district to a category data type
admin_district_ft["admin_district"] = (
    admin_district_ft["admin_district"].astype("category")
)

# Apply an order to our admin_district categories
admin_district_order = [
    "Sheffield", 
    "South Yorkshire", 
    "Other", 
    "Invalid postcode"
]

admin_district = CategoricalDtype(categories=admin_district_order, ordered=True)

admin_district_ft["admin_district"] = (
    admin_district_ft["admin_district"]
        .cat.reorder_categories(admin_district_order, ordered=True)
)

admin_district_ft = admin_district_ft.sort_values("admin_district")

(
  admin_district_ft
    .style
      .hide(axis = "index")
      .format(thousands = ",")
      .set_properties(subset = ["cases"], **{"text-align": "right"})
)
Postcode locations
admin_district cases
Sheffield 6,937
South Yorkshire 358
Other 645
Invalid postcode 428
Code 
# Summary table of homelessness cases by ward
ward_cases = (
    hl_case_lsp[hl_case_lsp["admin_district"] == "Sheffield"]
        .groupby("admin_ward")
        .size()
        .reset_index(name='cases')
).sort_values("cases", ascending = False).head()

(
  ward_cases
    .style
      .hide(axis = "index")
      .format(thousands = ",")
      .set_properties(subset = ["cases"], **{"text-align": "right"})
)
Wards with the most homelessness cases
admin_ward cases
Burngreave 1,003
City 541
Firth Park 500
Manor Castle 464
Darnall 420
Code 
# Downloaded ward boundaries from ONS Ope Geography Portal as GeoPackage
# https://geoportal.statistics.gov.uk/datasets/bd816ce7fa384745851f942786671f87_0/explore?location=59.429324%2C28.266256%2C3.58

# Get Sheffield ward boundaries
ward_filename = "Wards_December_2024_Boundaries_UK_BFC_3254702094383695880.gpkg"
wards = gpd.read_file(os.path.join(hl.get_path_local_data(), ward_filename))
wards = wards[wards['LAD24NM'] == 'Sheffield']

# Join number of cases to ward boundaries
wards = wards.merge(
    right = ward_cases,
    how = "left", 
    left_on = "WD24NM",
    right_on = "admin_ward"   
)

wards[["cases"]] = wards[["cases"]].fillna(value=0)
wards = wards.astype({"cases": "int"})

ax = wards.plot(column='cases', legend=True)
ax.set_axis_off()
Homelessness cases by ward

Is there a better geography to analyse by than Wards?
TODO: Stephen has suggested the (100) neighbourhoods as an alternative set of boundaries to Wards. Housing Market Areas is another suggestion.
TODO: Nicola has suggested MSOAs. This can link with census data. As a smaller area it will enabled a more in depth look at say the top 5 Wards. Plus, it may highlight some pockets that are masked by wider areas e.g. Wynn Gardens.
TODO: Compare top 5 Wards with “all other Wards” .
TODO: Stephen has suggested adding “last settled accommodation type” to this exploratory analysis of last settled postcode.

2 Household type

2.1 Annual

We don’t have a complete year of data for 2019 or 2025.

Code 
# Number of annual cases by household type
hh_type_yr = (
  hl_case.loc[
    (hl_case["application_year"] > 2019) & (hl_case["application_year"] < 2025), 
    ["application_year", "household_type"]
  ]
  .value_counts(dropna = False)
  .rename("cases")
  .reset_index()
)

# Title for each facet
def titled(strip_title):
    return " ".join(s.title() if s != "and" else s for s in strip_title.split("_"))

# Facet plot
(
    ggplot(
      hh_type_yr, 
      aes(x="application_year", y="cases", colour="household_type")
    )
    + geom_line(size=.5, show_legend=False)
    + facet_wrap("household_type", labeller=titled)
    + labs(
      x="Year",
      y="Cases",
      colour="Household type",
      title="Annual homelessness cases by household type"
    )
    + theme_classic(base_size=9)
)

2.2 Seasonal

TODO

3 Demographic Profiling of Settled Home Loss Cases

Exploratory analysis building a profile of demographic, geographic and support need information for households presenting for specific Settled Home Loss Reasons.

3.1 Settled Home Loss Reason

What are the different settled homeloss reasons?

Code 
# Frequency table for the Settled Home Loss Reason column 

homeloss_reason_ft = (hl_case["settled_home_loss_reason"]
  .value_counts(dropna = True)   #missing records will be dropped or might want to keep as "Unknown"
  .rename("cases")
  .reset_index()
)

# Format the table display
(
  homeloss_reason_ft.style
    .hide(axis = "index")
    .format(thousands = ",")
    .set_properties(subset = ["settled_home_loss_reason"], **{"text-align": "left"})
    .set_properties(subset = ["cases"], **{"text-align": "right"})
)
Frequency of Settled Home Loss Reason
settled_home_loss_reason cases
Family no longer willing or able to accommodate 4,664
Domestic abuse - victim 2,980
End of private rented tenancy - assured shorthold tenancy 2,423
Required to leave accommodation provided by Home Office as asylum support 1,541
Friends no longer willing or able to accommodate 1,259
Domestic abuse 1,254
Relationship with partner ended (non-violent breakdown) 1,137
Other 1,031
Non-racially motivated / other motivated violence or harassment 968
Eviction from supported housing 912
Departure from institution: Custody 736
End of social rented tenancy 564
End of private rented tenancy - not assured shorthold tenancy 407
Domestic abuse - alleged perpetrator excluded from property 187
Left institution with no accommodation available 177
Fire or flood / other emergency 158
Home no longer suitable due to disability / ill health 136
Property disrepair 134
Not known due to last settled accommodation Not known 97
Racially motivated violence or harassment 92
Voluntarily left accommodation to relocate 74
Mortgage repossession or sale of owner occupier property 73
Departure from institution: Hospital (psychiatric) 44
Departure from institution: Hospital (general) 44
Left HM Forces 9
Loss of tied accommodation 8
Departure from institution: Looked After Child Placement 8
Unsuccessful placement or exclusion from resettlement scheme/sponsorship 5

3.2 Family or Friends no longer willing or able to accommodate

Focusing on two specific reasons of settled home loss:

  • Family no longer willing or able to accommodate.
  • Friends no longer willing or able to accommodate.
Code 
# Define categories of interest

homeloss_reasons_fandf_categories_of_interest = ["Family no longer willing or able to accommodate", "Friends no longer willing or able to accommodate"]

# Filter the data for these 2 specific reasons

homeloss_reason_fandf_filtered = hl_case[hl_case["settled_home_loss_reason"].isin(homeloss_reasons_fandf_categories_of_interest)]

homeloss_reason_fandf_filtered_ft = (homeloss_reason_fandf_filtered     
  ["settled_home_loss_reason"]
  .value_counts(dropna = False)   
  .rename("cases")
  .reset_index()
)

# Format the table display
(
  homeloss_reason_fandf_filtered_ft.style
    .hide(axis = "index")
    .format(thousands = ",")
    .set_properties(subset = ["settled_home_loss_reason"], **{"text-align": "left"})
    .set_properties(subset = ["cases"], **{"text-align": "right"})
)
settled_home_loss_reason cases
Family no longer willing or able to accommodate 4,664
Friends no longer willing or able to accommodate 1,259

3.2.1 Breakdown by Age

Analysis of Settled Home Loss Reasons (Family or Friends no longer willing or able to accommodate) by applicant’s age at the time of application.

Code 
# Age Distribution for Applicants in the identified Home Loss Reasons categories of interest

homeloss_reason_fandf_filtered["age_on_application"].describe()


#Compare Age accross identified Home Loss Reasons categories of interest (Family or Friends no longer willing or able to accommodate) using a boxplot

(
  ggplot(homeloss_reason_fandf_filtered, aes(x="settled_home_loss_reason", y="age_on_application"))
  + geom_boxplot(fill="#4C72B0", color="black", outlier_alpha=0.3)
  + theme_classic()
  + labs(
      title="Age Distribution for Applicants in the identified Home Loss Reasons",
      x="Settled Home Loss Reason",
      y="Age on Application"
  )
  + scale_y_continuous(breaks=range(0, 101, 10))
  + theme(
      axis_text_x=element_text(rotation=45, ha="right", size=8),
      axis_text_y=element_text(size=10, margin={"r": 15}),
      axis_title_x=element_text(size=10, weight="bold"),
      plot_title=element_text(size=12, weight="bold", ha="center", margin={"b": 20}),
      #plot_margin={"t": 10, "r": 10, "b": 40, "l": 40}
  )

)

3.2.2 Breakdown by Age Group

Analysis of Settled Home Loss Reasons (Family or Friends no longer willing or able to accommodate) by applicant’s age group.

Code 
# Age Group Distribution for Applicants in the identified Home Loss Reasons categories of interest

#Create Age Groups

#Age groups defined
age_bins = [0,25,35,45,55,65,75,85,100]
age_labels = ["24 and under", "25-34", "35-44", "45-54", "55-64", "65-74", "75-84", "85 and over"]      

#New Column for Age Group
homeloss_reason_fandf_filtered["age_group"] = pd.cut(homeloss_reason_fandf_filtered["age_on_application"], bins=age_bins, labels=age_labels, right=False) 

#Age Group distribution
homeloss_reason_fandf_filtered["age_group"].value_counts()


# Create table by Age groups and settled_home_loss_reason

homeloss_reason_fandf_age_group_ft = (
  homeloss_reason_fandf_filtered.groupby(     
  ["settled_home_loss_reason", "age_group"])
  .size() 
  .unstack(fill_value=0)
  .reset_index()
)

# Format the table display
(
  homeloss_reason_fandf_age_group_ft.style
    .hide(axis = "index")
    .format(thousands = ",")
    .set_properties(**{"text-align": "center"})
    .set_properties(subset = ["settled_home_loss_reason"], **{"text-align": "left"})
    
)
settled_home_loss_reason 24 and under 25-34 35-44 45-54 55-64 65-74 75-84 85 and over
Family no longer willing or able to accommodate 2,209 1,343 596 300 161 43 8 4
Friends no longer willing or able to accommodate 291 413 289 160 87 16 2 1

3.2.3 Breakdown by Ethnicity

Analysis of Settled Home Loss Reasons (Family or Friends no longer willing or able to accommodate) by applicant’s ethnicity.

Code 
# Ethnicity Distribution for Applicants in the identified Home Loss Reasons categories of interest

homeloss_reason_fandf_filtered["ethnicity"].value_counts()


# Create table by Ethnicity and settled_home_loss_reason

homeloss_reason_fandf_ethnicity_ft = (homeloss_reason_fandf_filtered
  .groupby(     
    ["ethnicity",
    "settled_home_loss_reason"])
    .size()
    .unstack(fill_value=0)
  .reset_index()
)

# Sort by Total cases in descending order
homeloss_reason_fandf_ethnicity_ft["Total"] = homeloss_reason_fandf_ethnicity_ft.iloc[:, 1:].sum(axis=1)
homeloss_reason_fandf_ethnicity_ft = homeloss_reason_fandf_ethnicity_ft.sort_values("Total", ascending=False).drop(columns="Total")


# Format the table display
(
  homeloss_reason_fandf_ethnicity_ft.style
    .hide(axis = "index")
    .format(thousands = ",")
    .set_properties(subset = ["ethnicity"], **{"text-align": "left"})
    .set_properties(subset = homeloss_reason_fandf_ethnicity_ft.columns[1:], **{"text-align": "right"})
)
ethnicity Family no longer willing or able to accommodate Friends no longer willing or able to accommodate
White: English/Welsh/Scottish/Northern Irish/British 2,543 483
Don't know / refused 423 122
Black/ African/Caribbean/Black British: African 273 149
Asian/Asian British: Pakistani 260 38
Any other ethnic group 156 114
Any other Asian background 168 92
Any other White background 166 77
Mixed/Multiple ethnic groups: White and Black Caribbean 155 16
Other ethnic group: Arab 117 53
Black/ African/Caribbean/Black British: Caribbean 131 28
Any other Mixed/Multiple ethnic background 65 18
White: Gypsy or Irish Traveller 20 4
Asian/Asian British: Bangladeshi 18 3
White: Irish 15 5
Any other Black/African/Caribbean background 9 9
Mixed/Multiple ethnic groups: White and Asian 13 4
Asian/Asian British: Indian 11 1
Mixed/Multiple ethnic groups: White and Black African 2 4
Code 
#Compare Ethnicity accross identified Home Loss Reasons categories of interest (Family or Friends no longer willing or able to accommodate)

# Melt to long format 
homeloss_reason_fandf_ethnicity_long = homeloss_reason_fandf_ethnicity_ft.melt(id_vars="ethnicity", var_name="settled_home_loss_reason", value_name="cases")

# Create a Facet plot
(
  ggplot(homeloss_reason_fandf_ethnicity_long, aes(x="settled_home_loss_reason", y= "cases"))
  + geom_col(fill="#4c6c88", width=0.6)
  + facet_wrap("~ethnicity", labeller="label_value", scales="free_y", ncol=3)
  + theme_classic(base_size=8)
  + labs(
      title="Ethicity Distribution for Applicants in the identified Home Loss Reasons",
      x="Settled Home Loss Reason",
      y="Cases"
  )
  + theme(
      axis_text_x=element_text(rotation=45, ha="right", size=7),
      figure_size=(12, 10),
      strip_text=element_text(weight="bold", size=8),
      plot_title=element_text(size=10, weight="bold")
  )
  
)

3.3 End of private rented tenancy

Focusing on two specific reasons of settled home loss:

  • End of private rented tenancy - assured shorthold tenancy.
  • End of private rented tenancy - not assured shorthold tenancy.
Code 
# Define categories of interest

homeloss_reasons_ept_categories_of_interest = ["End of private rented tenancy - assured shorthold tenancy", "End of private rented tenancy - not assured shorthold tenancy"]

# Filter the data for these 2 specific reasons

homeloss_reason_ept_filtered = hl_case[hl_case["settled_home_loss_reason"].isin(homeloss_reasons_ept_categories_of_interest)]

homeloss_reason_ept_filtered_ft = (homeloss_reason_ept_filtered     
  ["settled_home_loss_reason"]
  .value_counts(dropna = False)   
  .rename("cases")
  .reset_index()
)

# Format the table display
(
  homeloss_reason_ept_filtered_ft.style
    .hide(axis = "index")
    .format(thousands = ",")
    .set_properties(subset = ["settled_home_loss_reason"], **{"text-align": "left"})
    .set_properties(subset = ["cases"], **{"text-align": "right"})
)
settled_home_loss_reason cases
End of private rented tenancy - assured shorthold tenancy 2,423
End of private rented tenancy - not assured shorthold tenancy 407

3.3.1 Breakdown by Age

Analysis of Settled Home Loss Reasons (End of private rented tenancy) by applicant’s age at the time of application.

Code 
# Age Distribution for Applicants in the identified Home Loss Reasons categories of interest

homeloss_reason_ept_filtered["age_on_application"].describe()

#Compare Age accross identified Home Loss Reasons categories of interest (End of private rented tenancy) using a boxplot

(
  ggplot(homeloss_reason_ept_filtered, aes(x="settled_home_loss_reason", y="age_on_application"))
  + geom_boxplot(fill="#4C72B0", color="black", outlier_alpha=0.3)
  + theme_classic()
  + labs(
      title="Age Distribution for Applicants in the identified Home Loss Reasons",
      x="Settled Home Loss Reason",
      y="Age on Application"
  )
  + scale_y_continuous(breaks=range(0, 101, 10))
  + theme(
      axis_text_x=element_text(rotation=45, ha="right", size=8),
      axis_text_y=element_text(size=10, margin={"r": 15}),
      axis_title_x=element_text(size=10, weight="bold"),
      plot_title=element_text(size=12, weight="bold", ha="center", margin={"b": 20}),
      #plot_margin={"t": 10, "r": 10, "b": 40, "l": 40}
  )

)

3.3.2 Breakdown by Age Group

Analysis of Settled Home Loss Reasons (End of private rented tenancy) by applicant’s age group

Code 
# Age Group Distribution for Applicants in the identified Home Loss Reasons categories of interest

#Create Age Groups

#Age groups defined
age_bins = [0,25,35,45,55,65,75,85,100]
age_labels = ["24 and under", "25-34", "35-44", "45-54", "55-64", "65-74", "75-84", "85 and over"]      

#New Column for Age Group
homeloss_reason_ept_filtered["age_group"] = pd.cut(homeloss_reason_ept_filtered["age_on_application"], bins=age_bins, labels=age_labels, right=False) 

#Age Group distribution
homeloss_reason_ept_filtered["age_group"].value_counts()

# Create table by Age groups and settled_home_loss_reason

homeloss_reason_ept_age_group_ft = (
  homeloss_reason_ept_filtered.groupby(     
  ["settled_home_loss_reason", "age_group"])
  .size() 
  .unstack(fill_value=0)
  .reset_index()
)

# Format the table display
(
  homeloss_reason_ept_age_group_ft.style
    .hide(axis = "index")
    .format(thousands = ",")
    .set_properties(**{"text-align": "center"})
    .set_properties(subset = ["settled_home_loss_reason"], **{"text-align": "left"})
    
)
settled_home_loss_reason 24 and under 25-34 35-44 45-54 55-64 65-74 75-84 85 and over
End of private rented tenancy - assured shorthold tenancy 242 818 726 401 187 39 10 0
End of private rented tenancy - not assured shorthold tenancy 52 124 134 64 28 4 1 0

3.3.3 Breakdown by Ethnicity

Analysis of Settled Home Loss Reasons (End of private rented tenancy) by applicant’s ethnicity.

Code 
# Ethnicity Distribution for Applicants in the identified Home Loss Reasons categories of interest

homeloss_reason_ept_filtered["ethnicity"].value_counts()

# Create table by Ethnicity and settled_home_loss_reason

homeloss_reason_ept_ethnicity_ft = (homeloss_reason_ept_filtered
  .groupby(     
    ["ethnicity",
    "settled_home_loss_reason"])
    .size()
    .unstack(fill_value=0)
  .reset_index()
)

# Sort by Total cases in descending order
homeloss_reason_ept_ethnicity_ft["Total"] = homeloss_reason_ept_ethnicity_ft.iloc[:, 1:].sum(axis=1)
homeloss_reason_ept_ethnicity_ft = homeloss_reason_ept_ethnicity_ft.sort_values("Total", ascending=False).drop(columns="Total")

# Format the table display
(
  homeloss_reason_ept_ethnicity_ft.style
    .hide(axis = "index")
    .format(thousands = ",")
    .set_properties(subset = ["ethnicity"], **{"text-align": "left"})
    .set_properties(subset = homeloss_reason_ept_ethnicity_ft.columns[1:], **{"text-align": "right"})
)
ethnicity End of private rented tenancy - assured shorthold tenancy End of private rented tenancy - not assured shorthold tenancy
White: English/Welsh/Scottish/Northern Irish/British 1,175 169
Don't know / refused 236 43
Black/ African/Caribbean/Black British: African 215 50
Any other White background 166 14
Any other ethnic group 135 32
Any other Asian background 104 25
Asian/Asian British: Pakistani 98 24
Other ethnic group: Arab 57 10
Black/ African/Caribbean/Black British: Caribbean 51 4
Mixed/Multiple ethnic groups: White and Black Caribbean 39 6
Any other Mixed/Multiple ethnic background 29 7
Asian/Asian British: Bangladeshi 22 2
Asian/Asian British: Indian 8 3
White: Gypsy or Irish Traveller 9 0
Any other Black/African/Caribbean background 7 0
White: Irish 7 0
Asian/Asian British: Chinese 4 1
Mixed/Multiple ethnic groups: White and Black African 2 3
Mixed/Multiple ethnic groups: White and Asian 2 1
Code 
#Compare Ethnicity accross identified Home Loss Reasons categories of interest (Family or Friends no longer willing or able to accommodate)

# Melt to long format 
homeloss_reason_ept_ethnicity_long = homeloss_reason_ept_ethnicity_ft.melt(id_vars="ethnicity", var_name="settled_home_loss_reason", value_name="cases")

# Create a Facet plot
(
  ggplot(homeloss_reason_ept_ethnicity_long, aes(x="settled_home_loss_reason", y= "cases"))
  + geom_col(fill="#4c6c88", width=0.6)
  + facet_wrap("~ethnicity", labeller="label_value", scales="free_y", ncol=3)
  + theme_classic(base_size=8)
  + labs(
      title="Ethicity Distribution for Applicants in the identified Home Loss Reasons",
      x="Settled Home Loss Reason",
      y="Cases"
  )
  + theme(
      axis_text_x=element_text(rotation=45, ha="right", size=7),
      figure_size=(12, 10),
      strip_text=element_text(weight="bold", size=8),
      plot_title=element_text(size=10, weight="bold")
  )

)

3.3.4 Geographic Profile

TODO