Wednesday, May 9, 2018

Why is the Irish “average wage” in the OECD Taxing Wages so low?

A few weeks ago the OECD published the 2018 update of their Taxing Wages publication.  There is plenty in the report worth chewing over but one minor issue that arises is the level of the “average wage” benchmark used for Ireland.

OECD Taxing Wages Average Wage Ireland 2000-2017

It doesn’t really appear to be anything too noteworthy.  It rises fairly rapidly up to 2008 (the figures are nominal so no account is taken of inflation), drops in 2009, has been edging upwards for the past three or four years or so, and is put at €36,800 for 2017.

But what about when we look at it in a comparative context?  OK, an international comparison  isn’t as straightforward as looking at a single country through time but we can overcome some issues if we look at countries that share a common currency. 

Here are the nominal average wages used for 2016 for a group of euro-area countries.

  • Netherlands €50,120
  • Germany €48,300
  • Belgium €46,528
  • Austria €45,073
  • Finland €43,716
  • France €37,906
  • Ireland €35,430

So, we still haven’t adjusted for prices but some of the gaps here are very large.  Is the nominal average wage in Ireland 27 per cent lower than Germany and 29 per cent lower than The Netherlands?

The OECD also has an Annual Wages publication which uses aggregate data from the national accounts to work out an average wage.  As the notes say:

Average annual wages per full-time equivalent dependent employee are obtained by dividing the national-accounts-based total wage bill by the average number of employees in the total economy, which is then multiplied by the ratio of average usual weekly hours per full-time employee to average usually weekly hours for all employees.

So what does this tell us?  Here are the 2016 averages from this dataset.

  • Ireland €51,336
  • Netherlands €46,709
  • Belgium €43,097
  • Austria €41,421
  • Finland €41,209
  • Germany €38,302
  • France €36,809

Of the selected group Ireland has gone from last to first with a 44 per cent rise in the estimated annual average wage as measured using the national accounts versus that used in Taxing Wages.  For most of the other countries the difference is only a few per cent – bar Germany where the national accounts show an average that is over 20 per cent lower.  Still by far the greatest absolute difference is for Ireland.

So,what explains it?  Perhaps it is coverage.  The national accounts cover the entire economy where the Taxes Wages figure is derived from NACE sectors B to N, i.e. the business economy which excludes state-dominated sectors such as education and health. But this coverage is the same for all countries.

The answer lies in the annex of the Taxing Wages report

Table A4 shows the method used to calculate the average wage.  One piece is the type of workers covered.  There are columns which show whether supervisory or managerial workers are included.  There are 36 countries covered in Taxing Wages and 34 of them include both of these types of workers, many of which we can assume would be in the top half of the earnings distribution. 

Which countries do not have them included when the average is calculated? Turkey and Ireland.

Part-time workers are another category that get varying treatment across countries.  Some countries leave them out altogether and only include full-time employees when working out the average.  Other countries include them but do so on a full-time equivalent basis such that their earnings are included but are scaled up as if they were a full-time employee.  There are six countries who include part-time employees in the calculation but do so on the basis of their actual annual earnings as part-time workers.

In their notes the OECD say:

The worker is assumed to be full-time employed during the entire year without breaks for sickness or unemployment.  However, several countries are unable to separate and exclude part-time workers form the earnings figures (see Table A.4). Most of them report full-time equivalent wages in these cases.  In four countries (Chile, Ireland, Slovak Republic and Turkey), the wages of part-time workers can neither be excluded nor converted into full-time workers (for example, an OECD Secretariat analysis of available Eurostat earnings data for selected European countries has show that include part-time workers reduces average earnings by around 10%).

So, the Irish figures used in Taxing Wages exclude supervisory and managerial staff, and include part-time workers on an unadjusted basis. 

What would the Irish figure be if it was estimated in line with the practice for most other countries?  Figures on that basis are available from Eurostat’s Structure of Earnings Survey though it is only available every four years.  The averages (with part-time employees given on a full-time equivalent basis) for sectors B to N are given as:

  • 2002: €33,320
  • 2006: €40,761
  • 2010 : €42,111
  • 2014: €44,700

This would put Ireland pretty much in the middle of the set of countries used above. 

Would it make a difference if the 2017 average wage used for Ireland was €45,000 rather than €36,400?  It probably would.

The OECD show that a single person on the average wage in Ireland has taxes and social insurance deductions of just under 20 per cent of their gross wage.  If this calculation was done using an average of €45,000 then the deductions would be 25 per cent of gross wage. 

This is part of an infographic the OECD used to promote the publication of Taxing Wages

OECD Taxing Wages Infographic

The OECD average is given as 25.5 per cent for 2017 and Ireland is included in the group of “low-tax” countries with tax rates for a single worker on the average wage of less than 20 per cent.  If an average wage was used that better reflected the actual outcomes in Ireland, in line with the methods used for other countries, then Ireland would be pretty much bang on the OECD average and would not be flagged (literally) on such infographics. 

Such a change would probably have limited impact on the estimated progressiveness of the Irish income tax system but it would change the range of values that are currently included in the analysis.  The OECD analysis extends up to earners on 167 per cent of the average wage.  This is around €60,000 with the average currently used but would be €75,000 if that was increased to the value proposed here.

If €45,000 was used for the Irish average wage (which the available data would support) then Ireland would not be down towards the bottom of charts like this.  This calculator puts the tax rate on a single person earning €45,000 in 2017 at 25 per cent.

OECD Taxing Wages Tax Rate on Average Wage

If a more realistic average wage was used, Ireland would pretty much be the same as the OECD average (and rise maybe ten places higher) in the above chart.  So why is such an unrepresentative average wage used for Ireland in this important report?

Why don’t the household sector accounts make sense?

Ireland’s national accounts have their foibles but they don’t seem to be limited to the corporate (MNCs) or financial (IFSC) sectors.  Their are some quirks in the household sector as well though maybe the heading above is a bit of an exaggeration.  The household accounts do make sense just maybe not in the way that we would expect.

The issue can be seen by looking at two outcomes:

  1. The flow of net lending or borrowing of the household sector.  This is essentially the outcome after all current and capital flows (income, expenditure, taxes and transfers) have been accounted for and gives a surplus available for use (net lending) or a deficit to be financed (net borrowing).
  2. The stock of loan liabilities of the household sector.  This is simply the sum of all loans (mortgages, car loans, short-term loans etc.) of the household sector taken at a point in time usually the year-end.

First the annual flow of net lending/net borrowing of the household sector:

Household Sector Net Lending

And the stock of loan liabilities:

Household Sector Loan Liabilities

OK, at first glance both of these are pretty much what would expect.  The household sector went deep into net borrowing by 2007 with loan liabilities rising steeply. The borrowing stopped abruptly in 2009 and the household sector became a net lender and this deleveraging is evident in the drop in the amount of loan liabilities from just north of €200 billion at the end of 2008 to around €140 billion at the of 2016.

So what’s the problem?

This issue is the difference between the amount of resources that the current and capital accounts tell us is available for net lending (such as repaying loans) and the drop in loan liabilities.  Here are the sum of the flows (net lending) versus the change in the stocks (loan liabilities) since 2008:

Household Sector Net Lending v Loan Liabilities

These don’t have to be the same but we would like to be able to explain the difference between them.  From 2009 to 2016 the loan liabilities of the household sector were reduced by around €60 billion.  Over the same period the household sector had cumulative net lending of around €15 billion.  The household sector has significantly reduced its loan liabilities but the resources to do so does not seem to have come from income.

And, as the first chart above shows, for the past few years we can see that the household sector has had a net lending outcome of close to zero (and even been a small net borrower) while at the same time continuing to reduce its stock of loan liabilities by around €8 billion a year. 

Where is the money coming from?

There are a number of ways the loan liabilities of the household sector can be reduced.  These include:

  • making repayments from income
  • selling assets (real or financial) to finance repayments
  • revaluations, write-offs or other stock adjustments.

Can these three factors explain the €60 billion reduction in loan liabilities that has occurred since 2008?  Even if all the available net lending generated from the first was devoted to repaying loans that would still leave €45 billion to be explained by the latter two.

We can get some insight into the contribution of revaluations and write-offs is we compare the flow of loan transactions (net drawdowns versus repayments) to the stock of liabilities.   Again we will take the end of 2008 as the starting point.

Household Sector Loan Transactions v Loan Liabilities

In the five years to the end of 2013 the €34 billion reduction in loan liabilities is closely matched by the minus €33 billion sum of loan transaction, i.e. the reduction was almost entirely achieved through repayments.  Since 2013 though, the reduction in the stock of household loan liabilities has exceeded the amount that can be attributed to loan transactions.  The gap reflects revaluations and write-offs with a €12 billion difference showing in the chart above by 2016.

But we still don’t know where the money came for the loan transactions shown above.  All told they are responsible for €47 billion of the €60 billion reduction in household loan liabilities since 2008.  If every cent of net lending the households generated from their income this could only provide €15 billion over the same period.  Where did the other €32 billion come from?

Maybe it came from deposits. This could be borrowers themselves using deposits. It would also show if an asset (such as a house) was sold with the seller using the proceeds to repay a loan while the purchaser funded all or a large part of the transaction with deposits.  For the household sector in aggregate give rise to a reduction in loans and a reduction in deposits. 

But household deposits have actually increased over the period, from €120 billion at the end of 2008 to €135 billion at the end of 2016.  Of course, this only adds to the mystery.  If, as we would expect, some of the household net lending went to increase deposits then even less would have been available for loan reductions so the gap to be explained is even larger than €32 billion.

How do the accounts explain the gap?

The financial accounts do make sense, i.e. add up, so we can see where it is implied the money came from. And this is the sale of assets – the second of the three possibilities listed above. shares.  What did the household sector sell? The answer apparently is unlisted shares, that is shares in private or family companies. 

However looking at the balance sheet doesn’t reveal that this is the place to look.  Here is the stock of unlisted shares held by the household sector.

Household Sector Unlisted Shares

There has been some reduction since 2009 but only in the order of a couple of a billion.  The stock position was €39.5 billion at the start of 2009 and had declined to €32.75 billion at the end of 2016.  It is not clear from this that the household sector has generated substantial resources from the sale of unlisted shares to fund the level of debt reduction we have seen.  But if we look at the transactions and revaluations behind these end-of-year positions this is what we see since 2009.

Household Sector Unlisted Shares Transactions Table

The stock position might have declined by only €7 billion or so between 2009 and 2016 but the cumulative impact of transactions was almost minus €53 billion, i.e. the household sector had net sales of unlisted shares of €53 billion!  The reason this had such a small impact on the stock position is that there were offsetting positive revaluations and other adjustments of plus €46 billion.  These are big numbers.

But wouldn’t we see some evidence of these transactions elsewhere. For example, selling tens of billions of assets would probably have some Capital Gains Tax implications but CGT receipts for the past few years have been in the order of hundreds of millions and, of course, reflect realised gains on a broad range of assets not just shares in private companies.  There is no evidence of tens of billions worth of transactions in unlisted shares. 

And it is interesting to note when they were added to the accounts.  This table shows the figures as published in different vintages of the Financial Accounts going back to 2012.

Household Sector Vintages of Unlisted Shares Transactions Table

In the 2013 release, the reported transactions for the years 2009 to 2012 summed to –€7.4 billion.  The latest estimates for those years sum to –€26.3 billion.  And it can be seen that the changes have been made in relatively neat amounts.  The 2009 figure has been revised by €5,000 million compared to what was in the accounts originally.  Similarly even figures can be seen for the other changes.

The absurdity of the reported Irish outcomes can be seen if we look at transactions in unlisted shares by the household sectors across the EU15.

EU15 Household Sector Unlisted Shares Transactions Table

No country comes close to having the scale of household sector transactions in unlisted shares to that of Ireland.  The median sum for the period 2009 to 2016 across the EU15 is +0.3 percentage points of GDP.  As can be see the outcome for Ireland is –26.6 percentage points of GDP (and that’s even with an inappropriately large denominator).

So Ireland’s household sector accounts do make sense. The net lending/borrowing shown in the Capital Account is a close match for the net financial transactions shown in the Financial Account. 

Household Sector Net Outcomes CSO 2002-2016

The discrepancy between net financial transactions (B.9F) and net lending/net borrowing (B.9)this is included by Eurostat (as variable B.9FX9) in their sector accounts.  But the relative consistency shown in the chart above, particularly since 2010, is only achieved through the inclusion of massive transactions in unlisted shares which do not make sense. 

As we went through before the CSO and Central Bank views of the financial transactions that are very different.  The Central Bank do not have have tens of billions of net sales transactions in unlisted shares.

So, where do we stand?

Here are the net lending/net borrowing outcomes for the household sectors of the EU15 in nominal per capita terms since 1995.

EU15 ISA Household Sector Net Lending Per Capita

There isn’t much of a surprise in seeing Ireland as the stand-out series up to 2006 and 2007.  But after getting back into the middle of the pack by 2009 we have dropped down through the ranks again in recent years.

And if we go back to the Current Account we can get the gross savings rate which is the share of gross disposable income (with an adjustment for pension funds) that is not used for consumption expenditure.

EU15 ISA Household Sector Gross Davings Rate

Just as with the outcome for net lending/net borrowing which comes after capital spending has been accounted for the gross savings rates show the recent outturns for Ireland to be “low”.

Is any of this important?

It is if trying to gauge the impact on the economy of future changes in the household sector is important.  A "deleveraging hypothesis" would suggest that the household sector has been repaying significant amounts of debt built up during the credit boom and that as these debts are repaid the household sector will have more resources available for consumption and investment as elevated savings to make debt repayments will no longer be required. 

This hypothesis is in line with the huge reduction in debt liabilities we have seen on the household sector's financial balance sheet since 2009.  However, the capital account shows that the household sector is already a net borrower and that there is limited scope to increase consumption and capital formation from existing resources.  This would suggest a stable path for the gross savings ratio over the next few years and forecasts of consumption growth in line with income growth would probably be appropriate. 

But what if household consumption expenditure and/or capital formation grows faster than household income? Should we be concerned particularly if the national accounts indicate that this could only be possible through increased borrowing? 

All this feeds through to expected changes in the current account of the Balance of Payments and informs one's view of whether the level of economic activity is sustainable or not. If activity is dependent on being financed by deficits that activity requires the lending and/or borrowing behavior to continue which, history tells us, is something that can change very quickly.

If the increased activity is being funded through a reduction in gross savings, net lending and/or current account surpluses then reasons for concern should be limited but if it is being funded by increased borrowing and/or current account deficits then at a minimum it should at least trigger an amber warning which may lead to concern being warranted. 

We don't have visible sight of the underlying current account (because of an issue with R&D services imports) and until we do the household gross savings rate and the outcome for net lending/net borrowing take up increased significance.  These would suggest that the financial capacity of the household sector is “tight” and that any rapid expansion should be a cause for concern.

So what to do?

Taking the gross saving rate and net borrowing numbers at face value offers some support for an “overheating hypothesis” and whatever policy prescriptions would follow from that diagnosis.  As the net lending/borrowing chart (the first chart above) shows it is a long way from the dizzying days of 2004 to 2007 but the Irish economy is one in which conditions can change rapidly.  Housing can be a driver of such swings.

But should we take the numbers at face value?  We think there is an underlying current account surplus in the Balance of Payments and that this possibly increased in 2017.   We know that the underlying net international investment position of the economy has been rapidly improving.  We know that the financial position of the household sector has improved and that debt reduction has made a significant contribution to that.  These sit counter to an “overheating hypothesis”.

So maybe just have contradictory signals and none can be taken as evidence of anything. But shouldn’t the pieces of the jigsaw fit together? Shouldn’t we have a coherent story?  In this instance the contradictory signals are in the current, capital and financial accounts of the household sector.

Putting together the national accounts requires a certain amount of estimation and interpolation. Hard evidence will only get you so far and jumps that join the dots are sometimes required.  But putting in tens of billions of sales transactions does not seem credible.

When ranking the hard evidence used for the household sector it is likely that the financial balance sheet would be towards the top of the reliability stakes.  Getting a handle on the loan and deposit positions of the household sector is possible using data from a fairly small number of financial institutions and while there might be some uncertainty around the levels, the direction and magnitude of changes should have greater confidence associated with them. 

These changes point to a €60 billion reduction in household loan liabilities since 2009 and a €15 billion increase in household deposits over the same period.  These figures might be wrong but to be out by tens of billions seems unlikely.

If these balance-sheet changes are considered reliable shouldn’t the rest of the accounts be consistent with them – and in a manner that makes more sense than using absurd transactions in unlisted shares?  It is possible the inconsistency is due to unreported income, in the informal or black economy, not showing up in the current account but one would imagine there is plenty of consumption spending that is not picked up either.

If the accounts are to be more consistent it would require income to be higher, spending to the lower, or some combination of the two.  However it was achieved, the outcome would be a higher gross savings rate and an outcome for net lending more in line with the observed deleveraging.

Of course we actually had that this time last year but such coherence was revised away. Can we have it back please?