Leeds EPC / Property age Data Exploration

Imports:

import json, os, zipfile
import pandas as pd
import numpy as np
import geopandas as gpd

Load in England & Wales EPC data

Extract EPC data for Leeds only:

full_epc_data = zipfile.ZipFile('data/raw/England-and-Wales-EPC-All.zip', 'r')
lad_code = 'E08000035'
file_to_extract = [x for x in full_epc_data.namelist() if lad_code in x and 'certificates.csv' in x][0]
if os.path.exists(f'data/extracts/{file_to_extract}'):
    extracted_data_path = f'data/extracts/{file_to_extract}'
else:
    extracted_data_path = full_epc_data.extract(file_to_extract, path='data/extracts/')

Read in EPC data for Leeds:

epc_data = pd.read_csv(
    extracted_data_path,
    low_memory=False,
    keep_default_na=True,
    na_values=['INVALID!', 'NO DATA!', 'not recorded', 'blank']
)
epc_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 291914 entries, 0 to 291913
Data columns (total 90 columns):
 #   Column                         Non-Null Count   Dtype  
---  ------                         --------------   -----  
 0   LMK_KEY                        291914 non-null  object 
 1   ADDRESS1                       291914 non-null  object 
 2   ADDRESS2                       118730 non-null  object 
 3   ADDRESS3                       13970 non-null   object 
 4   POSTCODE                       291914 non-null  object 
 5   BUILDING_REFERENCE_NUMBER      291914 non-null  int64  
 6   CURRENT_ENERGY_RATING          291914 non-null  object 
 7   POTENTIAL_ENERGY_RATING        291914 non-null  object 
 8   CURRENT_ENERGY_EFFICIENCY      291914 non-null  int64  
 9   POTENTIAL_ENERGY_EFFICIENCY    291914 non-null  int64  
 10  PROPERTY_TYPE                  291914 non-null  object 
 11  BUILT_FORM                     284999 non-null  object 
 12  INSPECTION_DATE                291914 non-null  object 
 13  LOCAL_AUTHORITY                291914 non-null  object 
 14  CONSTITUENCY                   291914 non-null  object 
 15  COUNTY                         1 non-null       object 
 16  LODGEMENT_DATE                 291914 non-null  object 
 17  TRANSACTION_TYPE               291906 non-null  object 
 18  ENVIRONMENT_IMPACT_CURRENT     291914 non-null  int64  
 19  ENVIRONMENT_IMPACT_POTENTIAL   291914 non-null  int64  
 20  ENERGY_CONSUMPTION_CURRENT     291914 non-null  int64  
 21  ENERGY_CONSUMPTION_POTENTIAL   291914 non-null  float64
 22  CO2_EMISSIONS_CURRENT          291914 non-null  float64
 23  CO2_EMISS_CURR_PER_FLOOR_AREA  291914 non-null  float64
 24  CO2_EMISSIONS_POTENTIAL        291914 non-null  float64
 25  LIGHTING_COST_CURRENT          291914 non-null  float64
 26  LIGHTING_COST_POTENTIAL        291914 non-null  float64
 27  HEATING_COST_CURRENT           291914 non-null  float64
 28  HEATING_COST_POTENTIAL         291914 non-null  float64
 29  HOT_WATER_COST_CURRENT         291914 non-null  float64
 30  HOT_WATER_COST_POTENTIAL       291914 non-null  float64
 31  TOTAL_FLOOR_AREA               291914 non-null  float64
 32  ENERGY_TARIFF                  291403 non-null  object 
 33  MAINS_GAS_FLAG                 264109 non-null  object 
 34  FLOOR_LEVEL                    232517 non-null  object 
 35  FLAT_TOP_STOREY                74527 non-null   object 
 36  FLAT_STOREY_COUNT              26851 non-null   float64
 37  MAIN_HEATING_CONTROLS          266346 non-null  object 
 38  MULTI_GLAZE_PROPORTION         263774 non-null  float64
 39  GLAZED_TYPE                    264142 non-null  object 
 40  GLAZED_AREA                    265466 non-null  object 
 41  EXTENSION_COUNT                265466 non-null  float64
 42  NUMBER_HABITABLE_ROOMS         265466 non-null  float64
 43  NUMBER_HEATED_ROOMS            265466 non-null  float64
 44  LOW_ENERGY_LIGHTING            281490 non-null  float64
 45  NUMBER_OPEN_FIREPLACES         285720 non-null  float64
 46  HOTWATER_DESCRIPTION           291914 non-null  object 
 47  HOT_WATER_ENERGY_EFF           291198 non-null  object 
 48  HOT_WATER_ENV_EFF              291157 non-null  object 
 49  FLOOR_DESCRIPTION              291889 non-null  object 
 50  FLOOR_ENERGY_EFF               16789 non-null   object 
 51  FLOOR_ENV_EFF                  16789 non-null   object 
 52  WINDOWS_DESCRIPTION            291799 non-null  object 
 53  WINDOWS_ENERGY_EFF             291646 non-null  object 
 54  WINDOWS_ENV_EFF                291646 non-null  object 
 55  WALLS_DESCRIPTION              291912 non-null  object 
 56  WALLS_ENERGY_EFF               291643 non-null  object 
 57  WALLS_ENV_EFF                  291643 non-null  object 
 58  SECONDHEAT_DESCRIPTION         291914 non-null  object 
 59  SHEATING_ENERGY_EFF            0 non-null       float64
 60  SHEATING_ENV_EFF               0 non-null       float64
 61  ROOF_DESCRIPTION               291715 non-null  object 
 62  ROOF_ENERGY_EFF                234975 non-null  object 
 63  ROOF_ENV_EFF                   234975 non-null  object 
 64  MAINHEAT_DESCRIPTION           291914 non-null  object 
 65  MAINHEAT_ENERGY_EFF            291156 non-null  object 
 66  MAINHEAT_ENV_EFF               291159 non-null  object 
 67  MAINHEATCONT_DESCRIPTION       291914 non-null  object 
 68  MAINHEATC_ENERGY_EFF           291449 non-null  object 
 69  MAINHEATC_ENV_EFF              291449 non-null  object 
 70  LIGHTING_DESCRIPTION           291914 non-null  object 
 71  LIGHTING_ENERGY_EFF            291505 non-null  object 
 72  LIGHTING_ENV_EFF               291612 non-null  object 
 73  MAIN_FUEL                      289578 non-null  object 
 74  WIND_TURBINE_COUNT             272060 non-null  float64
 75  HEAT_LOSS_CORRIDOOR            74527 non-null   object 
 76  UNHEATED_CORRIDOR_LENGTH       32490 non-null   float64
 77  FLOOR_HEIGHT                   111261 non-null  float64
 78  PHOTO_SUPPLY                   161373 non-null  float64
 79  SOLAR_WATER_HEATING_FLAG       175187 non-null  object 
 80  MECHANICAL_VENTILATION         265466 non-null  object 
 81  ADDRESS                        291914 non-null  object 
 82  LOCAL_AUTHORITY_LABEL          291914 non-null  object 
 83  CONSTITUENCY_LABEL             291914 non-null  object 
 84  POSTTOWN                       291914 non-null  object 
 85  CONSTRUCTION_AGE_BAND          264008 non-null  object 
 86  LODGEMENT_DATETIME             291914 non-null  object 
 87  TENURE                         283010 non-null  object 
 88  FIXED_LIGHTING_OUTLETS_COUNT   114067 non-null  float64
 89  LOW_ENERGY_FIXED_LIGHT_COUNT   114066 non-null  float64
dtypes: float64(26), int64(6), object(58)
memory usage: 200.4+ MB

We can see that there is 90 columns in the data, many of which contain lots of 'null' values. Let's limit the data to a subset of columns:

columns_to_keep = [
    'POSTCODE',
    'BUILDING_REFERENCE_NUMBER',
    'CURRENT_ENERGY_RATING',
    'POTENTIAL_ENERGY_RATING',
    'CURRENT_ENERGY_EFFICIENCY',
    'POTENTIAL_ENERGY_EFFICIENCY',
    'INSPECTION_DATE',
    'LODGEMENT_DATE',
    'CONSTRUCTION_AGE_BAND'
]
epc_data.drop(columns=epc_data.columns.difference(columns_to_keep), inplace=True)
epc_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 291914 entries, 0 to 291913
Data columns (total 9 columns):
 #   Column                       Non-Null Count   Dtype 
---  ------                       --------------   ----- 
 0   POSTCODE                     291914 non-null  object
 1   BUILDING_REFERENCE_NUMBER    291914 non-null  int64 
 2   CURRENT_ENERGY_RATING        291914 non-null  object
 3   POTENTIAL_ENERGY_RATING      291914 non-null  object
 4   CURRENT_ENERGY_EFFICIENCY    291914 non-null  int64 
 5   POTENTIAL_ENERGY_EFFICIENCY  291914 non-null  int64 
 6   INSPECTION_DATE              291914 non-null  object
 7   LODGEMENT_DATE               291914 non-null  object
 8   CONSTRUCTION_AGE_BAND        264008 non-null  object
dtypes: int64(3), object(6)
memory usage: 20.0+ MB

Load ONS Postcode lookup file and merge with the EPC data:

Grab ONS Postcode Lookup file (2020):

![ ! -f data/ONS-Postcode-Lookup.zip ] && wget https://www.arcgis.com/sharing/rest/content/items/b982ad8161dd4fc29bbe76038ae32c10/data -O 'data/ONS-Postcode-Lookup.zip'

Read in the ONS Postcode Lookup file & drop unnecessary columns:

pcd_lookup = pd.read_csv('data/ONS-Postcode-Lookup.zip', compression='zip', usecols=['pcds', 'lsoa11cd', 'ladcd'], low_memory=False)
pcd_lookup.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 2647046 entries, 0 to 2647045
Data columns (total 3 columns):
 #   Column    Dtype 
---  ------    ----- 
 0   pcds      object
 1   lsoa11cd  object
 2   ladcd     object
dtypes: object(3)
memory usage: 60.6+ MB

Cleanup the EPC Data

Merge the EPC data with the ONS postcode lookup (removing any rows with invalid postcodes):

epc_data['POSTCODE'] = epc_data['POSTCODE'].str.replace('\s*', '').str.upper()
pcd_lookup['POSTCODE'] = pcd_lookup['pcds'].str.replace('\s*', '').str.upper()
count = len(epc_data)
epc_data = epc_data.merge(pcd_lookup, on='POSTCODE', how='inner')
print(f'Dropped {count - len(epc_data)} rows with invalid postcodes')

Dropped 125 rows with invalid postcodes

Remove entries where the LSOA isn't in Leeds

leeds_lsoa_codes = pcd_lookup[pcd_lookup['ladcd'] == lad_code]['lsoa11cd'].drop_duplicates()
epc_data = epc_data[epc_data['lsoa11cd'].isin(leeds_lsoa_codes)]

Sort by EPC date and remove outdated EPC's (ie. if a property has multiple EPC's)

epc_data['LODGEMENT_DATE'] = pd.to_datetime(epc_data['LODGEMENT_DATE'], infer_datetime_format=True, errors='coerce')
epc_data['INSPECTION_DATE'] = pd.to_datetime(epc_data['INSPECTION_DATE'], infer_datetime_format=True, errors='coerce')
epc_data.sort_values(by=['INSPECTION_DATE', 'LODGEMENT_DATE'], ascending=False, inplace=True)

count = len(epc_data)
epc_data.drop_duplicates(subset=['POSTCODE', 'BUILDING_REFERENCE_NUMBER'], keep='first', ignore_index=True, inplace=True)
print(f"Dropped {count - len(epc_data)} outdated EPC's")

Dropped 67987 outdated EPC's

Create a table to store summarised data by LSOA:

lsoa_summary = pd.DataFrame(index=leeds_lsoa_codes)

Calculate average (mode) current & potential EPC ratings by LSOA:

mode = lambda x: pd.Series.mode(x)[0]
lsoa_summary['CURRENT_EPC_RATING_MODE'] = epc_data.groupby('lsoa11cd')['CURRENT_ENERGY_RATING'].agg(mode)
lsoa_summary['POTENTIAL_EPC_RATING_MODE'] = epc_data.groupby('lsoa11cd')['POTENTIAL_ENERGY_RATING'].agg(mode)
lsoa_summary.info()

<class 'pandas.core.frame.DataFrame'>
Index: 482 entries, E01011527 to E01011500
Data columns (total 2 columns):
 #   Column                     Non-Null Count  Dtype 
---  ------                     --------------  ----- 
 0   CURRENT_EPC_RATING_MODE    482 non-null    object
 1   POTENTIAL_EPC_RATING_MODE  482 non-null    object
dtypes: object(2)
memory usage: 11.3+ KB

Calculate average mode property age bands by LSOA from the EPC Data:

Replace blank age band values with 'UNKNOWN':

epc_data['CONSTRUCTION_AGE_BAND'].fillna('UNKNOWN', inplace=True)

Determine what the current Age Band values are in the EPC data:

current_age_band_values = epc_data['CONSTRUCTION_AGE_BAND'].unique().tolist()
current_age_band_values.sort()
print('Current age band values: \n')
print('\n'.join(current_age_band_values))

Current age band values: 

England and Wales: 1900-1929
England and Wales: 1930-1949
England and Wales: 1950-1966
England and Wales: 1967-1975
England and Wales: 1976-1982
England and Wales: 1983-1990
England and Wales: 1991-1995
England and Wales: 1996-2002
England and Wales: 2003-2006
England and Wales: 2007 onwards
England and Wales: before 1900
UNKNOWN

Standardise the naming of these age bands:

epc_data['CONSTRUCTION_AGE_BAND'] = epc_data['CONSTRUCTION_AGE_BAND'].str.lstrip('England and Wales: ') \
                                                                     .str.replace('before ', '0-') \
                                                                     .str.replace(' onwards', '-2020')
standardised_epc_age_bands = epc_data['CONSTRUCTION_AGE_BAND'].unique().tolist()
standardised_epc_age_bands.sort()
print(f'New age band values: \n')
print('\n'.join(standardised_epc_age_bands))

New age band values: 

0-1900
1900-1929
1930-1949
1950-1966
1967-1975
1976-1982
1983-1990
1991-1995
1996-2002
2003-2006
2007-2020
UNKNOWN

Let's look at the total number of properties in each age band:

epc_data['CONSTRUCTION_AGE_BAND'].value_counts().sort_index().plot.bar(xlabel='Age Band', ylabel='Number of properties')

<AxesSubplot:xlabel='Age Band', ylabel='Number of properties'>

Now let's calculate the number of properties in each age band by LSOA (from the EPC data):

epc_age_bands_by_lsoa = epc_data[['lsoa11cd', 'CONSTRUCTION_AGE_BAND']].fillna('UNKNOWN').pivot_table(index=['lsoa11cd'], columns='CONSTRUCTION_AGE_BAND', aggfunc='size', fill_value=0)
epc_age_bands_by_lsoa

CONSTRUCTION_AGE_BAND	0-1900	1900-1929	1930-1949	1950-1966	1967-1975	1976-1982	1983-1990	1991-1995	1996-2002	2003-2006	2007-2020	UNKNOWN
lsoa11cd
E01011264	16	3	69	116	41	3	0	1	0	0	0	4
E01011265	163	110	38	14	47	17	34	9	9	56	10	168
E01011266	13	34	83	45	36	11	6	5	8	93	134	79
E01011267	32	22	45	158	44	11	5	10	14	13	12	94
E01011268	0	0	70	203	27	41	29	3	3	1	0	2
...	...	...	...	...	...	...	...	...	...	...	...	...
E01033031	8	24	9	88	21	1	0	1	15	62	327	332
E01033032	2	7	8	65	87	5	21	4	44	255	267	858
E01033033	0	0	1	75	7	1	0	1	10	198	252	1
E01033034	18	5	0	0	0	0	6	9	154	287	325	221
E01033035	0	2	2	41	51	0	7	2	20	0	0	587

482 rows × 12 columns

Calculate total number of properties and mode age band by LSOA:

lsoa_summary[['EPC_PROPERTY_COUNT', 'EPC_PROPERTY_AGE_BAND_MODE']] = epc_age_bands_by_lsoa.agg(['sum', lambda row: row.idxmax(axis=1)], axis=1)
lsoa_summary

	CURRENT_EPC_RATING_MODE	POTENTIAL_EPC_RATING_MODE	EPC_PROPERTY_COUNT	EPC_PROPERTY_AGE_BAND_MODE
lsoa11cd
E01011527	D	B	352	1967-1975
E01011589	D	B	461	0-1900
E01011457	D	C	337	0-1900
E01011521	D	B	589	2003-2006
E01011529	D	C	285	1950-1966
...	...	...	...	...
E01011537	D	B	479	1930-1949
E01011551	D	B	421	1900-1929
E01011499	C	B	940	UNKNOWN
E01032496	C	B	601	2003-2006
E01011500	D	B	355	1950-1966

482 rows × 4 columns

Now let's look at the number of LSOA's per average age band:

lsoa_count_by_mode_age_band = lsoa_summary.groupby('EPC_PROPERTY_AGE_BAND_MODE').agg('size')
lsoa_count_by_mode_age_band.sort_index().plot.bar(xlabel='Average Property Age Band', ylabel="Number of LSOA's")

<AxesSubplot:xlabel='Average Property Age Band', ylabel="Number of LSOA's">

Now let's compare the EPC property age data with data from the VOA Housing Stock 2020 data

Get VOA 2020 Housing Stock data to compare:

voa_housing_stock_data = pd.read_csv('https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/920382/Table_CTSOP4_1_2020.csv', encoding='cp1252', low_memory=False)
voa_housing_stock_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 383404 entries, 0 to 383403
Data columns (total 22 columns):
 #   Column          Non-Null Count   Dtype  
---  ------          --------------   -----  
 0   GEOGRAPHY       383404 non-null  object 
 1   BA_CODE         3053 non-null    float64
 2   ECODE           383404 non-null  object 
 3   AREA_NAME       383404 non-null  object 
 4   band            383368 non-null  object 
 5   BP_PRE_1900     383368 non-null  object 
 6   BP_1900_1918    383368 non-null  object 
 7   BP_1919_1929    383368 non-null  object 
 8   BP_1930_1939    383368 non-null  object 
 9   BP_1945_1954    383368 non-null  object 
 10  BP_1955_1964    383368 non-null  object 
 11  BP_1965_1972    383368 non-null  object 
 12  BP_1973_1982    383368 non-null  object 
 13  BP_1983_1992    383368 non-null  object 
 14  BP_1993_1999    383368 non-null  object 
 15  BP_2000_2008    383368 non-null  object 
 16  BP_2009_2011    383368 non-null  object 
 17  BP_2012_2014    383368 non-null  object 
 18  BP_2015_2017    383368 non-null  object 
 19  BP_2018_2020    383368 non-null  object 
 20  BP_UNKNOWN      383368 non-null  object 
 21  ALL_PROPERTIES  383368 non-null  object 
dtypes: float64(1), object(21)
memory usage: 64.4+ MB

Filter to Leeds LSOA's only:

voa_housing_stock_data = voa_housing_stock_data[(voa_housing_stock_data['band'] == 'All') & (voa_housing_stock_data['ECODE'].isin(leeds_lsoa_codes))]
voa_housing_stock_data.set_index('ECODE', inplace=True)
voa_housing_stock_data.index.name = 'lsoa11cd'

Drop unnecessary columns and standardise age band naming:

columns_to_keep = [x for x in voa_housing_stock_data.columns if x.startswith('BP')]
voa_housing_stock_data.drop(columns=voa_housing_stock_data.columns.difference(columns_to_keep), inplace=True)
voa_housing_stock_data.rename(columns=lambda x: x.replace('BP_', '').replace('_', '-'), inplace=True)

Convert column datatypes to numerical:

voa_housing_stock_data = voa_housing_stock_data.apply(pd.to_numeric, errors='coerce').fillna(0)

Calculate total properties and mode age band by LSOA:

lsoa_summary[['VOA_PROPERTY_COUNT', 'VOA_PROPERTY_AGE_BAND_MODE']] = voa_housing_stock_data.agg(['sum', lambda row: row.idxmax(axis=1)], axis=1)
lsoa_summary

	CURRENT_EPC_RATING_MODE	POTENTIAL_EPC_RATING_MODE	EPC_PROPERTY_COUNT	EPC_PROPERTY_AGE_BAND_MODE	VOA_PROPERTY_COUNT	VOA_PROPERTY_AGE_BAND_MODE
lsoa11cd
E01011527	D	B	352	1967-1975	780	1965-1972
E01011589	D	B	461	0-1900	730	PRE-1900
E01011457	D	C	337	0-1900	650	PRE-1900
E01011521	D	B	589	2003-2006	1040	2000-2008
E01011529	D	C	285	1950-1966	590	1973-1982
...	...	...	...	...	...	...
E01011537	D	B	479	1930-1949	1010	1965-1972
E01011551	D	B	421	1900-1929	750	PRE-1900
E01011499	C	B	940	UNKNOWN	1290	2000-2008
E01032496	C	B	601	2003-2006	970	2000-2008
E01011500	D	B	355	1950-1966	720	1955-1964

482 rows × 6 columns

lsoa_summary.info()

<class 'pandas.core.frame.DataFrame'>
Index: 482 entries, E01011527 to E01011500
Data columns (total 6 columns):
 #   Column                      Non-Null Count  Dtype 
---  ------                      --------------  ----- 
 0   CURRENT_EPC_RATING_MODE     482 non-null    object
 1   POTENTIAL_EPC_RATING_MODE   482 non-null    object
 2   EPC_PROPERTY_COUNT          482 non-null    object
 3   EPC_PROPERTY_AGE_BAND_MODE  482 non-null    object
 4   VOA_PROPERTY_COUNT          482 non-null    object
 5   VOA_PROPERTY_AGE_BAND_MODE  482 non-null    object
dtypes: object(6)
memory usage: 26.4+ KB

lsoa_summary['EPC_PROPERTY_COUNT'] = pd.to_numeric(lsoa_summary['EPC_PROPERTY_COUNT'], downcast='integer')
lsoa_summary['VOA_PROPERTY_COUNT'] = pd.to_numeric(lsoa_summary['VOA_PROPERTY_COUNT'], downcast='integer')

Calculate percentage of total properties with a published EPC:

lsoa_summary['PROPERTIES_WITH_EPC_PERC'] = (lsoa_summary['EPC_PROPERTY_COUNT'] / lsoa_summary['VOA_PROPERTY_COUNT']).round(2)
lsoa_summary

	CURRENT_EPC_RATING_MODE	POTENTIAL_EPC_RATING_MODE	EPC_PROPERTY_COUNT	EPC_PROPERTY_AGE_BAND_MODE	VOA_PROPERTY_COUNT	VOA_PROPERTY_AGE_BAND_MODE	PROPERTIES_WITH_EPC_PERC
lsoa11cd
E01011527	D	B	352	1967-1975	780	1965-1972	0.45
E01011589	D	B	461	0-1900	730	PRE-1900	0.63
E01011457	D	C	337	0-1900	650	PRE-1900	0.52
E01011521	D	B	589	2003-2006	1040	2000-2008	0.57
E01011529	D	C	285	1950-1966	590	1973-1982	0.48
...	...	...	...	...	...	...	...
E01011537	D	B	479	1930-1949	1010	1965-1972	0.47
E01011551	D	B	421	1900-1929	750	PRE-1900	0.56
E01011499	C	B	940	UNKNOWN	1290	2000-2008	0.73
E01032496	C	B	601	2003-2006	970	2000-2008	0.62
E01011500	D	B	355	1950-1966	720	1955-1964	0.49

482 rows × 7 columns

Sort by LSOA and save as CSV file:

lsoa_summary.sort_index().to_csv('output/leeds-lsoa-summary.csv', index_label='lsoa11cd')

Grab the LSOA 2011 Boundaries Shapefile from the ONS and filter to Leeds only:

lsoa_geo = gpd.read_file('https://opendata.arcgis.com/datasets/c892586698ad4d268f9288f1df20ab77_0.zip?outSR=%7B%22latestWkid%22%3A27700%2C%22wkid%22%3A27700%7D')
lsoa_geo = lsoa_geo[lsoa_geo['LSOA11CD'].isin(leeds_lsoa_codes)]
lsoa_geo.drop(columns=lsoa_geo.columns.difference(['LSOA11CD', 'LSOA11NM', 'geometry']), inplace=True)
lsoa_geo.to_crs(epsg=4326, inplace=True)

Merge the LSOA boundaries table with our LSOA summary table to create a geographic summary:

lsoa_geo = lsoa_geo.merge(lsoa_summary, left_on='LSOA11CD', right_on='lsoa11cd', how='inner')
lsoa_geo.info()

<class 'geopandas.geodataframe.GeoDataFrame'>
Int64Index: 482 entries, 0 to 481
Data columns (total 10 columns):
 #   Column                      Non-Null Count  Dtype   
---  ------                      --------------  -----   
 0   LSOA11CD                    482 non-null    object  
 1   LSOA11NM                    482 non-null    object  
 2   geometry                    482 non-null    geometry
 3   CURRENT_EPC_RATING_MODE     482 non-null    object  
 4   POTENTIAL_EPC_RATING_MODE   482 non-null    object  
 5   EPC_PROPERTY_COUNT          482 non-null    int16   
 6   EPC_PROPERTY_AGE_BAND_MODE  482 non-null    object  
 7   VOA_PROPERTY_COUNT          482 non-null    int16   
 8   VOA_PROPERTY_AGE_BAND_MODE  482 non-null    object  
 9   PROPERTIES_WITH_EPC_PERC    482 non-null    float64 
dtypes: float64(1), geometry(1), int16(2), object(6)
memory usage: 35.8+ KB

Save as GeoJSON:

lsoa_geo.to_file('output/leeds-lsoa-summary.geojson', driver='GeoJSON')