Underestimating housing consumption in the national accounts

It's not just household-level data such as the SILC where our approaches to the provision of social housing present measurement issues. In the national accounts, payments such as to HAP landlords are currently counted as a benefit-in-kind for households.  This is in contrast to the benefit-in-cash approach now applied in the SILC.

And, up to 2020, there was a further difference between the micro-data and macro-data approaches to measuring social housing.  In the national accounts an imputed social-transfer-in-kind was previously included for the 130,000 or so households who are local authorities tenants.  This has now been removed.

In the government accounts, on the revenue side, local authorities were treated as generating market output for sale.  For social housing this would be the rent contributions received from tenants plus an imputed amount to bring the value of the output in line with market values, which were based on rents in the private rental sector, using both unregulated and regulated rents.

On the expenditure side, this imputed output was used as a social-transfer-in-kind to households. There would also be expenditure incurred as compensation of employees, intermediate consumption and depreciation or fixed capital formation for the provision of local authority housing. 

This meant that the imputed value was on both sides of the accounts and netted out for the balance.  The  impact of local authority housing on the general government balance was the different between the rent contributions received from tenants and the other expenditure items incurred.

It also meant that the consumption of housing services by local authority tenants was based on the market value of the housing services they used as it is for private tenants (through actual rents paid) and for owner-occupiers(through imputed rents). 

The change introduced this year means that the consumption of housing services of local authority tenants is now treated as non-market output and the value of the consumption of that in the national accounts is based on the costs of providing it (mainly compensation of employees, intermediate consumption and depreciation) rather than an imputed market value. 

It should be recognised that most government-provided services (health, education, policing etc.) are included in national accounts on a cost rather than value basis.  The change to also do so with local authority housing had no net impact on the government’s accounts.  The imputed rents of local authority tenants were removed from both the revenue and expenditure  sides, as imputed market output for revenue and the social-transfer-in-kind for expenditure. 

Indicators like government spending to national income would have been reduced.  And so indeed would national income when the market output based on market value was replaced by non-market output based on costs. 

This change was introduced between the April 2021 Government Finance Statistics and the July 2021 Government Income and Expenditure Accounts.  The July publication included the following note:

Reclassification of local authority housing rent as non-market output

To date the provision of local authority housing was treated as a market output. This meant that the difference between the differential rent paid by the tenant and a market rent was calculated and included as P.11 (market output), with a corresponding imputed expenditure D.632 (social benefit in kind). However, Approved Housing Bodies (AHB) reclassified into the local government sector are considered as non-market producers, with no imputed rent calculation made.

On review, this approach was deemed not appropriate and thus a decision has been made to treat the local authority housing output as non-market. This ensures consistency with AHBs. This determination means that there is no longer an imputed D.632 expenditure related to local authority rent. Local authority rent payments are now recorded as P.131 (incidental sales and fees of non-market establishments). This methodology has been applied from 1995.

There can be lots of reasons for revisions between releases so attributing them solely to a methodological change may not always be correct.  Here are the figures for market and non-market output in the April and July releases.

For the three years shown (2018 to 2020), the market output of the general government sector was revised down by an average of €1.6 billion – and now takes a value of zero.  For the same years, non-market output was revised up by an average of €1.25 billion.  This suggests that around €350 million of housing consumption may have been “lost” as a result of the change in methodology.

In the greater scheme of things €350 million of consumption may not be that significant, though as a similar treatment is applied for housing by AHBs the underestimate may be slightly larger.  The underestimate is limited to the extent to which the costs of providing LA or AHB  housing is less than the market value of that housing.

In pre-COVID 2019, actual individual consumption was €133 billion, of which the consumption of housing services was €28.5 billion.  The consumption of housing services was made up of:

• Household Consumption Expenditure
  • Actual rentals for housing €5,393m
  • Imputed rentals for housing €16,459m
  • Maintenance and repair of the dwelling €289m
  • Water supply and miscellaneous services relating to the dwelling €357m
  • Electricity, gas and other fuels €3,091m
• Government Consumption Expenditure
  • Social transfers-in-kind via market producers €824m
  • Social transfers-in-kind via non-market production by Government €2,087m

A few hundred million extra in there isn’t going to make a huge difference but again maybe points to difficulties in measuring outcomes in relation to housing in Ireland.

There is no doubt that local authorities are not market establishments so deeming the goods and services they provide to be from non-market production has logic to it and imputing values is not an exact process.  However, for housing, a relatively close market comparator can be found, i.e. the private rental sector.  This can be used to give a market value for the output produced.  And up to this year that is what was done for local authority housing in Ireland.

For most EU countries, this isn’t an issue.  Almost all of the consumption of housing is the result of household expenditure.  Government spending does not go to provide housing, at least not directly.  The general government sectors don’t make payments to landlords or provide housing directly.

As the list above shows, in Ireland government consumption expenditure is responsible for around 10 per cent of total consumption of housing.  This figure is by far the largest in the EU.

For 21 of the EU27, the share of housing consumption due to government expenditure is less than one per cent (and is essentially zero for around half of those).  The closest country to Ireland is France and even then that is at a level that is less than half the outturn for Ireland. 

That is not to say there isn’t public or social housing in most of these countries. There is.  But it is happens in such a way that it is provided by entities that are outside the general government sector.  And they generate their revenues from rents paid by tenants rather than payments by government.  That’s not to say that either approach is right or wrong just that they are different.

The value of housing consumption may be underestimated in other countries to the extent that regulated or controlled rents are used in determining imputed rents for owner occupiers.  In Ireland, for around five per cent of households, the value of their housing consumption is based on the cost of providing those housing services rather than what the tenants might have to pay as private tenants. Again, neither is right or wrong. Just different.

Including HAP in Disposable Income

The CSO have published the 2020 results for the Survey of Income and Living Conditions (SILC).  There were a number of methodological changes that mean there is a series break in 2020 compared to earlier estimates.  These are set out in a useful Information Note.

The note also confirms that the Housing Assistance Payment (HAP) payments to landlords, as well as payments to landlords under the Rental Accommodation Scheme (RAS) are considered part of household income:

From 2020, social transfers in SILC are defined as the total income received from DEASP social welfare transfers (e.g. jobseekers related payments, state pension (contributory and non-contributory), family or children related allowances), income received as education related allowances (e.g. Student Universal Support Ireland [SUSI] grants) and housing related supports which include rent supplement and Local Authority contributions to landlords of SILC respondents who are Housing Assistance Payment (HAP) or Rental Accommodation Scheme (RAS) tenants.

This isn’t necessarily a change that was introduced this year but it is now clear that the rent payments made by local authorities to lands in the HAP and RAS are included in the household income of the tenants.  It does not appear that the rent paid by Local Authorities to Approved Housing Bodies (AHBs) is included in the income of AHB tenants.  These rents paid to AHBs, as with those paid under the HAP or RAS (or the smaller Mortgage-To-Rent programme) are based on market rents.  Nor is a similar social transfer to households included in the income of tenants of Local Authorities even though they benefit from subsidised rents in the same way as RAS or AHB tenants.

The rent contribution that recipients of HAP make to their local authority is calculated on the same basis as the above but the tenants may be required to make an additional payment to the landlord if the agreed rent exceeds the relevant HAP limit for that area.  For the first half of 2019 it was estimated that 28 per cent of HAP tenants were making a top-up payment to their landlord.

That the rent contributions tenants make to their local authorities under these programmes is part of those households housing costs is incontrovertible (as well as any top-up payments that may be made to private landlords in HAP). What is less clear-cut is whether the payments local authorities make landlords on behalf of tenants should be included in the tenants’ income.

In terms scaled up to national level, around €1.5 billion of housing supports are included in income in the 2020 SILC.  This includes €600 million for HAP,  €130 million for RAS and €25 million for MTR. As these are not taxable they are then also fully included in disposable income.  These payments make up close to one per cent of the aggregate disposable income of around €90 billion that is represented in the SILC.

This may not seem like a significant amount but the payments will certainly be significant for those households involved.  There are around 17,000 households in the Rental Accommodation Scheme and 68,000 in the Housing Assistance Payment programme.

The issue is whether payments that households don’t directly receive should be included in their disposable income?  There is no doubt that the payments benefit the living conditions of the households but they are not income that the households can choose how to spent. 

As of the 30th June 2019, the average monthly landlord payment was €830 per month and the average rent contribution from the tenant was €47.50 per month.  The treatment in the SILC is that the €830 is counted as household income.  It is not clear in the notes but it certainly should be the case that the €830 also be included in the household’s housing costs and not the €47.50 differential rent. 

Different definitions abound but disposable income would seem like something a household's decisions should be able to have a bearing on how it is used - even if that can be spending on necessities or long-term commitments.  The household can, to a certain extent, choose how to do these. 

If a private tenant moves to a property that has a lower rent, then the use of their disposable income will change.  With a lower rent, the household’s income after housing costs are deducted will rise. 

If a household receiving HAP moves to an area that has lower HAP limits, then the inclusion of HAP as income, would see their income fall after the move.  The decision to move to an area with lower rents hasn't changed their spending (at least for the 72 per cent not making a top-up payment).  They continue to pay the appropriate differentiated rent. But in the SILC the move has changed their income. 

It seems a little incongruent that a spending/consumption decision would change income.  But then, the HAP aids consumption and does contribute to living standards so could be viewed as contributing to income.  But should the disposable income of a household with a medical car be increased every time they attend a GP?  No.  This is a benefit-in-kind rather than a cash transfer.

Why should HAP be counted as a cash transfer rather than a benefit-in-kind?  The State is paying for housing services.  Though that is with the intention of maintaining income after housing costs for beneficiaries.

However, it is not clear why the rent paid by a local authority paid to a private landlord should be counted in the household’s income but not the rent local authorities pay to AHBs.  Perhaps it could be justified on the basis that the household under HAP can choose they property the rents are to be paid for but in either case their income after housing costs will be determined by the differential rent not the rent paid to the landlord (whether that is a private landlord or an AHB).

Consider the contrived example of two households of similar composition, similar income and living in similar properties. Both are eligible for social housing. 

One of the households is an AHB tenant and the household’s income is under the at-risk-of-poverty threshold.  The household’s income after housing costs is calculated after the differential rent, and other costs, are deducted.  Logically, the household’s income after housing costs are deducted from it will also be below the AROP threshold

The second household is a HAP recipient and rents a similar property to the first household but does so from a private landlord.  The inclusion of the rent paid to the landlord in this household’s disposable income puts it above the at-risk-of-poverty threshold.

These households are similar in almost all respects. But one is considered below the at-risk-of-poverty threshold and one is above it.  The only difference is that one is a AHB tenant and one is a HAP recipient.  Assuming no top-up payment to the HAP landlord, both households have similar income remaining after housing costs are deducted. Their living standards are likely to be similar but their income in the SILC will differ significantly.

So, it could be that the headline at-risk-of-poverty rate is being under-estimated because a cash transfer for housing that households cannot spend is included in the income for HAP and RAS tenants.  Would it be better classed as a benefit-in-kind? That is how HAP and RAS are treated in the national accounts?  Or if an income transfer is to be included in household income should it be included for all social housing tenants and not just those in HAP or RAS?

The level and distribution of income in Ireland in the 2020 SILC

Eurostat have 2020 figures for Ireland to the EU-SILC, the EU’s Statistics on Income and Living Conditions.  Ireland is one of the last countries to have provided figures to Eurostat and the national version won’t be published by the CSO for another few weeks.

Those figures won’t be much different to what is now available on Eurostat but will come with much more detailed background notes.  One such item to be explained is that all the 2020 figures Eurostat have for Ireland are marked with a “b” – for series break.  It is not yet clear what this is. It could be that was a change during 2020 with in-person interviews shelved as COVID hit.

And at the outset it is probably worth noting that although this is the 2020 SILC, the year represents when the data was collected rather than the period for which it applies to.  The CSO carry out the survey across the full year, and respondents are asked for details of their income in the 12 months prior to the survey.

Thus, if someone was interviewed in January 2020 for the 2020 SILC, the reference period for their income would almost wholly encompass 2019.  This would move one month forward for people interviewed in February and so on.  Indeed, when this data goes through the OECD’s methodology it will be assigned to 2019 when published on the OECD’s income inequality database.

Another difference worth noting is the equivalent scale used to compare households of different sizes.  The CSO apply a national equivalence scale with applies a weight of 1.00 to the first adult, 0.66 to all subsequent adults and 0.33 to all children under 14.  These are added together with the household’s income divided by the result to get income in terms of equivalent people.

Eurostat uses the OECD-modified scale which gives a weight of 1.00 to the first adult, but 0.5 to all subsequent adults and 0.3 to all children under 14.

This means that a household of 2 adults and 2 young children would have an equivalising factor of 2.32 with the CSO’s approach and a factor of 2.1 with Eurostat’s approach.  This changes the level of equivalised income within each dataset but should not have a hugely significant impact on growth rates or other relative comparisons.  So, the figures the CSO itself publishes in a few weeks could be slightly different but the overall trends will be the same.

The Level of Income

We will start with median equivalised income in nominal terms.

In and of itself the actual level is not that informative.  Eurostat’s figure for 2020 is €26,250 but income per equivalent person is not a concept we can readily relate to.  What matters are the growth rate and relative differences, both within the income distribution in Ireland and with other countries.

Eurostat’s figures show that for the SILC data collected in 2020, the change in median equivalised income was +2.8 per cent.  That matches the average nominal growth rate of the previous 15 years though the significant volatility in the outcomes that gave rise to that average is evident from the chart.

Relative to the rest of the EU15, Ireland had the fourth-highest median equivalised disposable income  - in nominal terms.  While some exchange rate conversions are made no adjustment is made for different price levels in the below chart.

The Distribution of Income

Three of the commonly used inequality measures that are applied to the SILC data are the gini coefficient, the quintile share ratio and the at-risk-of-poverty rate.  Here is the gini coefficient fir Ireland since 2004 with a higher figure representing higher inequality.

Ireland’s estimated gini co-efficient has been trending downward over the last few decades – but it should be noted that the changes are exaggerated by the truncated vertical axis used in the above chart.  The changes are small.  The Eurostat figure for the 2020 SILC is not significantly different from what it was the previous year, going from 0.283 to 0.287.

Within the EU15, Ireland’s gini coefficient for disposable income is in the middle of the pack.  Ireland stands out more for the change in the gini coefficient over the last 15 years.  This following chart shows how the average for each country from 2018 to 2020 differs from its average for 2004 to 2006. 

For most of the EU15, the gini was either unchanged or increasing over the period.  For those countries showing a reduction in their gini coefficient the fall in Ireland was the second largest, with only Portugal showing a larger fall.

The gini coefficient is a useful indicator but as an measure which condenses a population-wide distribution of income into a single number looking at other measures can also be useful.  The quintile share ratio compares the income share of the top 20 per cent of the income distribution to the income share of the bottom 20 per cent.

The pattern here corresponds to what is shown by the gini coefficient.  Over the last 15 years, Ireland’s quintile share ratio has fallen from around five in the mid-2000s to around four now.  This indicates that incomes at the bottom of the income distribution have grown faster than those at the top.

Ireland’s quintile share ratio is the sixth lowest in the EU15.

The at-risk-of-poverty rate focuses in the lower end of the income distribution and looks at how many people live in households with an equivalised income that is less than 60 per cent of the national median. 

With Eurostat putting Ireland’s median equivalised income at €26,250 in 2020, this gives an at-risk-of-poverty threshold of €15,750 for a single-person household.  If we multiple this by 2.1 we get the threshold for a 2 adult plus 2 young children household: €33,075.

Here is the share of people in Ireland who live in households with an equivalised income below the 60 per cent threshold.

This shows a similar pattern to the quintile share ratio and Ireland’s position in the EU15 is also the same (sixth lowest).

As it is a measure of inequality, the at-risk-of-poverty rate is not always a good indicator of changes in living standards at the lower end of the income distribution.  The post-2008 period in Ireland is a good illustration of this.  We know that there were very significant falls in income but this is not reflected in any noticeable increase in the at-risk-of-poverty rate in the period from 2009 to 2012.

This is because the at-risk-of-poverty rate is a relative measure.  As incomes in the economy fell, the threshold for been assessed as at-risk-of-poverty also fell.  One way to get an insight into absolute changes in living standards is to use a fixed threshold (with changes only made for inflation rather than the general trend of income in the economy).  Eurostat provide an anchored at-risk-of-poverty rate with the 2005 threshold as the anchor.

The at-risk-of-poverty rate was 20 per cent in 2005.  In the chart above, the 2005 threshold is rolled forward (adjusted for inflation) and the share of the population below that threshold is reported.  It is the changes rather than the levels that are informative here.  We can see that this anchored at-risk-of-poverty measure rose significantly after 2008.  It had been falling consistently since 2014, but was unchanged in 2020.

There’s much more to the SILC than income figures but that’s probably enough for now.