Why this study?
In its mission to promote the energy transition of households, Hello Watt offers a free application for monitoring its energy consumption via Linky meters for electricity and Gazpar for gas.
To complete its offer and following the announcement of the restrictions that would be applied to thermal colanders, Hello Watt sought to estimate the DPE of homes whose consumption was known.
However, against all odds, the data team concluded that knowing the consumption of a home did not make it possible to estimate the DPE reliably.
How is the ECD calculated?
Following this study, many diagnostic players have reacted with varying degrees of virulence, the main criticism being that ECD and consumption cannot be compared directly, since ECD is based on standard housing use.
The National Real Estate Federation (FNAIM) has published A press release in this sense, and an article in diagnosticeur-immobilier.fr rightly recalls that the Ministry of Energy Transition indicates: “this estimate cannot be compared to the actual bills of users”.
As the infodiag article in reaction to the study sums it up well, in the context of the DPE-3CL 2021 which is the method used for the diagnoses studied there: "we are only interested in the performance of the building and of its equipment, regardless of who occupies the property. Whether a person lives in his pavilion two months of the year, or whether he is present 365 days / 365, seven days a week, 24 hours a day, the estimate remains the same. The legislator wanted to do so to facilitate the comparison between goods for purchase or rental. »
So the DPE, even if it is expressed in kWh/year, deliberately ignores the factors that can influence consumption that are not related to the building itself. Some items of expenditure are even totally excluded, such as cooking or household appliances.
This degrades its reliability as an indicator of consumption, and in return makes it a better indicator to guide the renovation of buildings.
So does comparing DPE and consumption make sense?
Following the publication of the study, some players protested because comparing real consumption and an ECD would make no sense. However, this is exactly the method of "DPE on invoice" which was usable until July 1, 2021.
So certainly the new 3CL method is supposed to be more reliable and better represent the energy quality of the building, but Hello Watt defends that it is very relevant to compare it with the actual consumption both individually and on average.
After all, we renovate homes with the primary aim of reducing their energy consumption rather than to increase a rating, and there are other ways of reducing consumption such as raising awareness among individuals or setting up consumption monitoring tools.
What do we expect when comparing DPE and average consumption by class?
Even if, on the scale of a single dwelling, uses can vary consumption much more than the energy efficiency of the dwelling, by taking an average over a large number of dwellings, we would expect these variations to cancel each other out. .
For example, class C includes dwellings whose estimated consumption (for the uses taken into account by the DPE) is between 110 and 180 kWh/year/m² for standard use.
If the standard use that is taken into account by the 3CL method is a realistic average behavior, then we expect the average consumption of a class C dwelling for the uses taken into account by the DPE to be approximately 145 kWh, and we can add 50% for the uses not taken into account (household appliances, digital... according to ADEME, the uses taken into account for the DPE represent on average 66% of the consumption of the dwellings) which brings us to 217 kWh/year.
For class F, we expect an average consumption of 375 kWh/year/m², or 562 kWh/year/m² once all the uses have been integrated.
What do we observe in terms of average consumption by DPE class?
In the study presented here, 462 dwellings are considered, compared to 221 in the initial study. On the other hand, Hello Watt is limited to classes C to F because the other classes contain less than 20 accommodations, which could skew the averages.
Hello Watt expected that homes with higher energy efficiency ratings would consume less, and luckily they did! The graph below shows that low-rated homes consume more than high-rated homes.
On the other hand, the trend is ultimately quite weak:
- If we take Hello Watt's estimates, for class C the 195 kWh/year/m² are quite close to the estimated 217.
- On the other hand, for class F Hello Watt expected to have a consumption of 562 kWh/year/m², and we are very far from it since F housing only consumes 256 kWh/year/m² on average.
This is even more obvious if we consider the median:
- Half of the classified C dwellings consume more than 195 kWh/year/m², and the other half less.
- And for F-labeled housing, half consume more than 219 kWh/year/m² and half less.
In other words, almost half of F dwellings consume less than half of C dwellings!
What does this significant difference between mean and median mean for class F? The average consumption of this class is pulled up by a few very energy-intensive dwellings, but most of the dwellings in this class have consumption comparable to dwellings in other classes, which is reflected in the median.
We see that (in the most common classes C to F) even on average, dwellings with a higher DPE do not consume much more than dwellings with a lower DPE.
Why do less well-rated accommodations not consume much more?
As stated in the original study, one possibility is that ECD is poorly designed or poorly implemented. Another hypothesis would be that the uses vary in such a way as to counter the impact of the energy performance of the dwelling.
Kezaco? For example, maybe F dwellings do not consume more than C dwellings because they sacrifice their comfort to heat less. This brings us back to the idea of fuel poverty.
Another formulation of this dependence between insulation and uses is the rebound effect, which starts from the observation that after renovation work, comfort increases but consumption does not decrease as much as hoped.
We can therefore see that these results are not necessarily in contradiction with the interests of DPEs! The DPE is an indicator of the energy performance of housing, which has an impact on consumption but also on the comfort of the inhabitants.
What do we observe in terms of dispersion?
In statistics, a distinction is made between central tendency indicators, such as the mean or median, and dispersion indicators such as the variance, the standard deviation or the interquartile range.
Where indicators of central tendency smooth out disparities to indicate trends, indicators of dispersion help quantify the closeness of measurements.
The following graph gives, for each DPE class, the median consumption but also the first and third quartiles, and the 5th and 95th percentiles.
We immediately see that the difference in median between the classes is insignificant compared to the very large variance within the same class. A possible explanation is that consumption habits from one person to another are so great that in the end ECD has little influence on consumption.
We also see that the worse the DPE class, the greater the uncertainty.
What can we say to a person who has an ECD C? We can tell it that it has a one in two chance that its consumption is between 137 and 234 kWh/year/m².
But what to say to a person who has an ECD F? It has a one in two chance that its consumption is between 107 and 387 kWh/year, which is a huge range, and there is still a one in two chance that it is outside the range!
In other words, a quarter of F dwellings consume less than 107 kWh/year/m², which is the threshold between classes B and C!
Should the DPE be modified so that it better reflects consumption?
We have established the relevance of comparing DPE and consumption, and we have established that DPE was a poor predictor of consumption, both in absolute terms (DPE is often far from consumption) and on average (misclassified dwellings do not consume much more than well-classified housing).
What conclusions can be drawn ? Until 2021, DPEs could use the invoice method, which by definition corresponded exactly to consumption. What would be the consequences of a return to the invoice method?
One of the main variables that is not taken into account by the DPE is intermittency: a poorly insulated secondary residence (for example a mountain chalet) will have a low average consumption, therefore a very good DPE on the bill, but a bad DPE 3CL. If we rely on the DPEs to prioritize energy renovations, this means:
- Based on the DPE 3CL, we give priority to renovating poorly insulated dwellings, regardless of whether they are used seasonally or not.
- Based on consumption or an ECD on invoice, priority is given to renovating the dwellings that consume the most.
The choice between these alternatives is political in nature, and it is presented here only in a simplified way and as a thought experiment. In particular, the DPE is certainly used in several areas and in particular the identification of thermal sieves, but renovation work is preceded by an energy audit which obeys different rules and is much more precise.
Many other variables are taken into account such as the comfort of the occupants, the possibilities of fraud, the quality of the DPE 3CL, the other residential uses of energy such as the heating of swimming pools or the recharging of electric vehicles.
En conclusion
Hello Watt, by its position, is fortunate to have access to consumption and DPE data for a large number of dwellings simultaneously, but their data set can have biases that are difficult to quantify. As noted by Liberation, their sample is not representative of the population, for example it only consists of homes that have recently completed an ECD, and we can assume that users of the Hello Watt application are more concerned about their energy consumption. . This database may also contain fewer second homes than expected, which could bias the results.
Even assuming that this sample is not biased, the link between housing performance and consumption is complex, so many phenomena could explain the disparity between ECD and consumption, both for individual dwellings and on average.
Various players analyze each component of this chain, from the properties of insulating materials to the influence of renovations on behavior (rebound effect) via the repeatability of diagnostics.
It is in the interest of our planet and in the interest of all of us, Hello Watt, individuals, diagnosticians and public authorities to act on the variables that are within our reach:
- Improve information to users on their actual and future consumption
- Encourage people to improve the comfort of housing and reduce their consumption
- Maximize diagnostic confidence
- Fight against fuel poverty