What can DECC’s NEED data tell us about levels and change in domestic electricity consumption?

Invited paper given at RSS2015 session on DECC's NEED data. Some animations may not work so well on SpeakerDeck.
Anderson, B., (2015) What can DECC’s NEED data tell us about levels and change in domestic electricity consumption? Invited paper, Royal Statistical Society International Conference 2015, Univesity of Exeter, 8/9/2015 .

Transcript

1. What can DECC’s NEED data tell us about levels and change in domestic electricity consumption? Ben Anderson b.anderson@soton.ac.uk (@dataknut) Engineering & Environment (Energy & Climate Change)
2. The menu § Levels & trends (& outliers) § Variation § What’s going on? § How can we tell? 2
3. The menu § Levels & trends (& outliers) § Variation § What’s going on? § How can we tell? 3
4. Levels & Trends 4 Source: DECC (2014) Energy Consumption in the UK (2014)
5. Levels & Trends: High Level 5 Source: DECC (2014) Energy Consumption in the UK (2014) Includes heat
6. 1 10 100 1000 10000 100000 2005 2006 2007 2008 2009 2010 2011 2012 Econs kWh Electricity Max Electricity Mean Electricity Median Electricity Min Levels & Trends: under NEED’s bonnet 6 Source: DECC (2014) NEED End User License Data, http://discover.ukdataservice.ac.uk/catalogue/?sn=7518 error bar = +/- 1 s.d •  NEED EULF •  204k dwellings •  2005-2012 •  1.6m records
7. Levels & Trends: under the bonnet 7 Source: DECC (2014) NEED End User License Data File
8. 1 10 100 1000 10000 100000 2005 2006 2007 2008 2009 2010 2011 2012 2013 Econs kWh Electricity Max Electricity Mean Electricity Median Electricity Min ECUK 2014 Non-corrected 'average' electricity Levels & Trends: under the bonnet 8 Source: DECC (2014) NEED End User License Data, error bar = +/- 1 s.d; DECC (2014) Energy Consumption in the UK (2014)
9. 0.00% 1.00% 2.00% 3.00% 4.00% 5.00% 6.00% 2005 2006 2007 2008 2009 2010 2011 2012 % of households M(issing) G (> 25,000) L (< 200) Levels & Trends: Exclusions 9 Source: DECC (2014) NEED End User License Data, includes ‘L’ and ‘G’ ‘Cleansing’: Some were ‘missing’ ‘Too high’ > 25,000 kWh -> ‘G’ ‘Too low’ < 200 kWh -> ‘L’
10. 97% 98% 98% 99% 99% 100% 100% 1 to 50 51-100 101-150 Over 151 Gas Other A or B C D E F G Floor area m2 Main fuel EE Band Valid Econs Excluded > 25,000 Exclusion analysis I 10 Source: DECC (2014) NEED End User License Data
11. -1 -0.5 0 0.5 1 1.5 2 2.5 2005 2006 2007 2008 2009 2010 2011 2012 b Floor area band (> 150m2) EE Band (G) Main heat fuel (not gas) Economy 7 Off gas Exclusion analysis II 11 Source: DECC (2014) NEED End User License Data, logistic regression predicting exclusion by year
12. -1 -0.5 0 0.5 1 1.5 2 2.5 2005 2006 2007 2008 2009 2010 2011 2012 b Floor area band (> 150m2) EE Band (G) Main heat fuel (not gas) Economy 7 Off gas Exclusion analysis II 12 Source: DECC (2014) NEED End User License Data, logistic regression predicting exclusion by year
13. The menu § Levels & trends (& outliers) § Variation § What’s going on? § How can we tell? 13
14. Is consumption constant? 14 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 Econs Decile at t+1 Econs decile at t 60.0-80.0 40.0-60.0 20.0-40.0 0.0-20.0 Source: DECC (2014) NEED End User License Data, transition analysis for deciles of yearly electricity consumption Transition probability
15. Is there much variation? 15 3206.802 2625.797 1881.947 -20000 -15000 -10000 -5000 0 5000 10000 15000 20000 25000 30000 overall between within Econs All valid Max Min sd Mean Source: DECC (2014) NEED End User License Data, Pattern for valid Econs
16. 2113.822 1801.831 1280.581 -6000 -4000 -2000 0 2000 4000 6000 8000 10000 12000 14000 overall between within Econs 90% Max Min sd Mean Two types of variation: • Differences in • socio-demographics • appliances & infrastructures • habits Between household: • Changes in • socio-demographics • appliances & infrastructures • habits Within household: 16
17. The menu § Levels & trends (& outliers) § Variation § What’s going on? § How can we tell? 17
18. Differences between households 18 Source: DECC (2014) NEED End User License Data, log Econs for 2012 only, multivariate OLS regression using STATA Variables: region, economy 7, fuel, property age, type, floor area, EE band, loft insulation, wall type -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 North East (contrast) North West Yorkshire & The Humber East Midlands West Midlands East of England London South East South West Wales Economy 7 (No) Economy 7 (Yes) Main heat fuel (Gas) Main heat fuel (Other) before 1930 (contrast) 1930-1949 1950-1966 1967-1982 1983-1995 1996 onwards Detached house (contrast) Semi-detached house End terrace house Mid terrace house Bungalow Flat (inc. maisonette) 1 to 50 m2 (contrast) 51-100 m2 101-150 m2 Over 151 m2 Band A or B (contrast) Band C Band D Band E Band F Band G Loft: Less than 150mm Loft: Greater than or equal to Cavity wall Other wall b (coefficient)
19. Differences between households 19 -0.05 0 0.05 0.1 0.15 0.2 0.25 North East (contrast) North West Yorkshire & The Humber East Midlands West Midlands East of England London South East South West Wales b (coefficient) -0.1 -0.05 0 0.05 0.1 0.15 0.2 Band A or B (contrast) Band C Band D Band E Band F Band G Loft: Less than 150mm Loft: Greater than or equal to 150mm Cavity wall Other wall b (coefficient) Source: DECC (2014) NEED End User License Data, log Econs for 2012 only, multivariate OLS regression using STATA Variables: region, economy 7, fuel, property age, type, floor area, EE band, loft insulation, wall type
20. Differences between households 20 0 0.1 0.2 0.3 0.4 0.5 0.6 Economy 7 (No) Economy 7 (Yes) Main heat fuel (Gas) Main heat fuel (Other) b (coefficient) 0 0.2 0.4 0.6 0.8 1 1 to 50 m2 (contrast) 51-100 m2 101-150 m2 Over 151 m2 b (coefficient) -0.3 -0.2 -0.1 0 0.1 0.2 before 1930 (contrast) 1930-1949 1950-1966 1967-1982 1983-1995 1996 onwards Detached house (contrast) Semi-detached house End terrace house Mid terrace house Bungalow Flat (inc. maisonette) b (coefficient) Source: DECC (2014) NEED End User License Data, log Econs for 2012 only, multivariate OLS regression using STATA Variables: region, economy 7, fuel, property age, type, floor area, EE band, loft insulation, wall type
21. Adjusted r sq = 0.15 => 85% of the variance is unexplained What’s missing? But 21 Source: DECC (2014) NEED End User License Data, log Econs for 2012 only, multivariate OLS regression using STATA Variables: region, economy 7, fuel, property age, type, floor area, EE band, loft insulation, wall type
22. People! 22 Source: DECC (2014) NEED End User License Data, Econs for 2012
23. How can we add people? 23 Source: Living Costs and Food Survey 2002-2012, multivariate OLS regression on reported electricity bill -2 -1.5 -1 -0.5 0 0.5 1 1.5 Semi (Detached) Terrace Flat/maisonette Other 2 rooms (1 room) 3 rooms 4 rooms 5+ rooms Rent from council (Own) Social rent Private rent/other -2 -1.5 -1 -0.5 0 0.5 1 1.5 HRP NS-SEC 2 (NS-SEC 1) HRP NS-SEC 3 HRP Inactive HRP Retired HRP single person HRP other N Children < 14 N Children 14-16 HRP 25-34 (16-24) 35-44 45-54 55-64 65-74 75+
24. -2 -1.5 -1 -0.5 0 0.5 1 1.5 Total raindays 3 year anomaly low (previous quarter) Mean rainfall 3 year anomaly high (previous quarter) Mean sunshine 3 year anomaly high (previous quarter) And we can add context 24 Source: Living Costs and Food Survey 2002-2012, multivariate OLS regression on reported electricity bill Drier than usual Sunnier than usual
25. -2 -1.5 -1 -0.5 0 0.5 1 1.5 Total raindays 3 year anomaly low (previous quarter) Mean rainfall 3 year anomaly high (previous quarter) Mean sunshine 3 year anomaly high (previous quarter) And we can add context 25 Source: Living Costs and Food Survey 2002-2012, multivariate OLS regression on reported electricity bill Drier than usual Sunnier than usual
26. This is reported expenditure Unknown error Link NEED & LCFS households? But 26 Source: Living Costs and Food Survey 2002-2012 Source: SPRG/ARCC-Water Survey, 2011 www.sprg.ac.uk
27. • Differences in • socio-demographics • appliances & infrastructures • habits Between household: • Changes in • socio-demographics • appliances & infrastructures • habits Within household: Two types of variation: 27
28. Differences within households Consumption at time t+1 Consumption at time t 28 Source: DECC (2014) NEED End User License Data, log Econs, multivariate lagged regression using STATA Variables: region, economy 7, fuel, property age, type, floor area, EE band, cavity wall/loft insulation/boiler install date Explains c. 40% 3206.802 2625.797 1881.947 -20000 -10000 0 10000 20000 30000 overall between within Econs All valid Max Min sd Mean
29. Differences within households 29 Source: DECC (2014) NEED End User License Data, log Econs, multivariate xt regression (re/fe) using STATA Variables: region, economy 7, fuel, property age, type, floor area, EE band, cavity wall/loft insulation/boiler install date Time constant variables give similar results to cross-sectional models -0.08 -0.07 -0.06 -0.05 -0.04 -0.03 -0.02 -0.01 0.00 Have boiler installed Have loft insulation installed Have cavity wall insulation installed on scheme coefficient •  ~ 6 % Econs change •  ~ 0.3 % within variance
30. So: 30 Source: DECC (2014) NEED End User License Data, log Econs, multivariate xt regression using STATA Variables: region, economy 7, fuel, property age, type, floor area, EE band, cavity wall/loft insulation/boiler install date • ~ 0.3% of within variance Change variables • ~ 29% of between variance Time constant variables
31. What’s missing? Appliance & infrastructure change Social change Changes in ‘what people do’ 31
32. What to do? 32 Source: Longhi, S (2014) Residential energy use and the relevance of changes in household circumstances Energy Economics, 49 , 440 -450, http://dx.doi.org/10.1016/j.eneco.2015.03.018
33. Source: SPRG/ARCC-Water Survey, 2011 www.sprg.ac.uk But: 33 This is also reported expenditure Unknown error Link NEED & USOC households?
34. Electricity consumption: to summarise • Between is slightly better than within NEED: we can’t explain much variance • Cross-sectional • Check for reported consumption errors • Difference between LCFS <-> NEED linkage • Longitudinal • Check for reported consumption errors • Difference within over time USOC<-> NEED linkage 34
35. Questions? §  b.anderson@soton.ac.uk §  @dataknut §  Original 2014 'End User License' version of the data: –  available from: UK DATA ARCHIVE: Study Number 7518 - National Energy Efficiency Data- Framework, 2014 http://discover.ukdataservice.ac.uk/catalogue/?sn=7518 –  Detailed documentation: https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/332169/ need_anonymised_dataset_accompanying_documentation.pdf –  Full coding details of variables at: https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/315189/ need_dataset_look_ups.xlsx §  STATA code: https://github.com/dataknut/DECC-data 35