Wednesday, February 24, 2021

Two views of Irish prices

Everyone loves a comparison.  Figures from the CSO today have been reported with the headline:

Prices in Republic 35% above EU average, CSO figures show

The cost of basic products and services in the Republic was 35.4 per cent higher than the EU average in 2019, and the second highest after Denmark, a new report by the Central Statistics Office (CSO) shows.


The CSO’s report shows that price levels here were 25.9 per cent above the EU average back in 2009. This dipped to 21.1 per cent above the EU average in 2011 as the economy entered into a recessionary period, before increasing again in recent years (to 35.4 per cent in 2019).

And here is a chart illustrating these figures.

PPP Price Levels Ireland EU27 2009-2019

The price level for the EU27 as a whole is set equal to 100 for each year with the Irish price level in 2019 estimated to be 35.4 per cent higher than this.  Denmark is also shown.   In 2019, Ireland had the second-highest price level for household final consumption expenditure in the EU, after Denmark.

So, this is one view of Irish prices.  In the last ten years Ireland’s price level has risen faster than the price level of the EU27 as a whole moving from being 20 per cent higher in 2010 to 35 per cent higher in 2019.  Irish prices started off higher and became even higher again.  Relative to EU’s highest price level, we can see that the Irish price level was significantly below the Danish price level in 2010 but has converged on the Danish level over the past decade.

Here’s a second view.  This is from the harmonised index of consumer prices (HICP).  This is what is used to measure inflation across the EU.  It does not give relative price levels across countries but gives an estimate of relative price changes within a country (i.e. the inflation rate).

HICP Ireland EU27 2009-2020

The indices for Ireland, Denmark and the EU27 are set equal to 100 for 2009 and the relative prices changes that occurred within each are shown by the changes from this point.  And it can be clearly seen that HICP inflation in Ireland has been much lower than for either Denmark or the EU27.

Since 2009, Danish consumer prices are up almost 12 per cent while for the EU27 as a whole there is an estimated price rise of over 15 per cent.  On the other hand, Irish prices in the HICP are up less than four per cent on 2009.

So there you have it.  Two different views.  The price level indices for household final consumption expenditure show that the Irish price level has risen by about 10 percentage points relative to both the price levels of Denmark and the EU27 (when the EU27 is set equal to 100).  On the other hand, the harmonised index for consumer prices shows that inflation in Ireland since 2009 has been eight percentage points lower than in Denmark and 11 points lower than the inflation for the EU27.

This is not to suggest that there is something wrong with either of these views of Irish prices.  The price level for household final consumption expenditure and the harmomised index of consumer prices are different things.  And as the Irish Times point out a lot of the action is linked to housing:

“The CSO’s Measuring Ireland’s Progress 2019 study suggests the primary reason for the higher cost of living here is housing.”

If you want to take, a very deep, trip into the weeds to see how housing can explain some of the divergence between the charts above see here.

Monday, February 22, 2021

Tiger phases of income inequality

A feature of the distribution of disposable income in Ireland over the past 20 years is how stable the standard measures of inequality have been, even as the economy has swung from boom to bust. Income inequality in Ireland is lower than it was in the 1980s and early 1990s (and likely much lower than it was in the 1970s) but measures such as the gini co-efficient appear to have been largely unchanged since 2000.

Gini Coefficient Ireland 1995-2019

Indeed, given the confidence intervals of the estimates it is likely that very few of the annual changes gave rise to a significant change in inequality.  In recent years, the confidence intervals have been of the range +/- 0.7 or 0.8.  This means there would have to be an annual change in the gini coefficient of 1.5 points, or more, in order for there to be a statistically significant change in inequality in a single year.

The lack of annual changes that are significant is probably no surprise.  Yes, inequality is impacted by conditions in a particular year but it is also the result of policies and outcomes going back several years, maybe even decades.

A sequence of changes in the same direction can result in a significant change in inequality.  For example, if the Irish gini coefficient was to start at a particular level and fall by one point in each of the following two years, then it is likely that neither of the annual changes would be considered significant (the confidence intervals for adjacent years would overlap) but the change over the two years would be significant (the confidence intervals for the starting and final years would not overlap).

With the above chart it is hard to set out any narrative for income inequality in Ireland apart from stating that inequality now is lower than it was in the 1990s.  But maybe we can extract a tale or two if we make a couple of adjustments:

  • truncate the vertical axis so that the changes are highlighted
  • calculate a three-year trailing average to smooth out some of the volatility in the single-year estimates.

By doing the above (and adding some annotation) we get the following:

Gini Coefficient Ireland 3YMA 1997-2019 Phases

These are the estimates available from Eurostat.  There is no estimate for 2002 so the figures in the above chart for 2002,2003 and 2004 are based on an average of two rather than three years.

The changes are now somewhat exaggerated (as the vertical axis is so narrow) but we can fit a narrative that links changes in inequality to the widely varying performance of the Irish economy over the past 25 years.  And given the confidence intervals of the individual-year estimates it is likely that the peak-to-trough changes shown in this chart are significant.

We start with the “good tiger” phase of the 1990s when the Irish economy underwent a transformation with a huge expansion in employment opportunities and strong income growth based on increased output and exports.  This period saw a relatively steep fall in the inequality of disposable income.

Then we enter the “bad tiger” phase of the early 2000s with a credit-fueled, construction dependent economy and significant net inward migration.  This period was accompanied by a rise income inequality though with around half of the gains of the late 1990s reversed.

It didn’t last and the “sick tiger” began to cough and splutter in 2008 as the credit binge came to an end.  Initially the crash led to fall in inequality perhaps as some of the high incomes of the boom evaporated but as the economy continued through recession and austerity inequality rose as we moved in the early 2010s.  Again, the rise saw about half of the reductions in equality of the preceding fall reversed.

Eventually, the “sick tiger” was replaced by the “recovering tiger”.  From 2014 on, there was a return to strong employment growth and even some modest wage inflation.  Without needing the stimulant of credit to act as a P.ED., this period seemed to mirror the performance of the economy in the 1990s though this time the reduction in inequality was at a slower rate.

Still, the outcome of the rollercoaster was that, at the end of the 25 years, disposable income inequality in Ireland was at its lowest ever measured level in 2019. 

Of course, the “recovering tiger” has been replaced by the “covid tiger”.  It is really difficult to tell what impact the pandemic will have on income inequality in Ireland.  Many households have seen their income largely unchanged (social welfare recipients, many employees, some self-employed). 

Other households have suffered very large income shocks – though state supports such as the PUP, TWSS and variants have meant that there continued to be strong growth in aggregate household disposable income in 2020.

It could be 2020 from an inequality perspective is much like most of preceding 20 years and will not have a statistically significant change in inequality.  We know that the hardest-hit sectors have been in the services part of the domestic economy.  These sectors have a higher share of low-paid workers.  But we also know, that:

  1. Low-paid workers are in households right across the distribution of disposable, and
  2. The income replacement rates for low-paid workers on the PUP and TWSS will be higher.   

It will be some time yet before we have updated inequality statistics for 2020 (and the CSO will also be working to ensure consistency with previous estimates) but, at this remove, as good a guess as any would that there will not be a statistically significant change in the gini coefficient in 2020.  Maybe.

Thursday, February 18, 2021

Distribution and Participation

Figures from the World Inequality Database show that the top one percent share of pre-tax fiscal income rose from around six per cent the mid-1980s to around 12 per cent in the mid-2010s.

Top One Percent Income Share 1983-2015

Figures from Eurostat show that the share of the population in non-agricultural employment rose from around 25 per cent in the mid-1980s to 45 per cent in 2019.

Non Agri Employment as Share of Population 1983-2019

The distribution of fiscal income is a useful indicator but the data is based on a sub-sample of the population that is always change: only those who have fiscal income, i.e. those with tax returns filed either by themselves or there employer.  The income shares across the population have to be estimated and it is assumed that those people who are not included in the data have no income – or, at least, have no pre-tax fiscal income.  

Changes in participation could have an effect on how the distribution of pre-tax fiscal income should be interpreted.  Changes in the distribution of pre-tax fiscal income among recipients may not necessarily reflect changes in the distribution of that income across the full population – even with the assumption of zero income for those with no pre-tax fiscal income.

Top One Percent Income Share and Employment 1983-2015 2

From the EU-SILC we do have a top one per cent share for disposable income, albeit for a shorter time period.  Of course, there may be issues with the use of a survey to measure incomes at the extreme of the distribution (super-high incomes may be under-represented in the sample) but to the extent that this is a consistent problem with income surveys the trends in incomes shares towards the top of the distribution may be relatively unaffected.

Top One Percent Share of Disposable Income 2005-2019

The top 1 per cent share of disposable income in the SILC fell from around seven per cent in 2005 and 2006 and dropped under five per cent after 2010. In recent years it has risen but at 5.5 per cent of so remains below the levels of the mid-2000s.

For what it’s worth, the bottom cut-off for the top one per cent in 2019 was an equivalised disposable income of just over €100,000 or €210,000 for a household of 2 adults and 2 children.  The SILC suggests there are close to 50,000 people living in such households.

The work of Kennedy et al (2019) gives income shares for both gross and disposable income in the fiscal data from tax returns.

Top One Percent Shares Gross and Disposable Income 2006-2015

This is different to measure of disposable income in the SILC but the pattern is roughly the same.  In the table we can see that the top 1 per cent share of disposable income fell from 2006 t0 2012 and rose thereafter but remained below the 2006 level.

Tuesday, February 16, 2021

Ireland to add 0.1% to 2020 US GDP

US MNCs causing distortions within Ireland’s national accounts is not unusual.  As a small economy it means the impact of these can be outsized (e.g. the 26 per cent GDP growth rate in 2015). 

It looks like something similar to happen to the US national accounts for 2020 – though the relative impact on the much larger US economy will be much smaller. Here we identify something that could add 0.1 per cent to US GDP.  This is small but non-trivial.

The issue in question is something we have looked at before: the changing nature of outbound royalties from Ireland.  The earlier post goes through more of the details but for our purposes here we note that there has been a very significant drop in outbound royalty payments from Ireland to The Netherlands and directly to Offshore Financial Centres.

Royalty Imports to EA and OFCs 2011-2020

It is easy to see that these collapsed in 2020.  Royalty payments to the Euro Area – almost all of which went to The Netherlands – plunged in Q1 2020, while the same happened for payments to Offshore Financial Centres – Bermuda, Cayman etc. – in Q3 2020.  This corresponds to announcements by companies such as Google and Facebook that they were unwinding their “double irish” (and “dutch sandwich”) structures.

Some companies have chosen to onshore their IP to Ireland thereby eliminating the royalties shown above.  On the other hand some companies have repatriated their IP to the US.  This means there are still outbound royalty payments from Ireland but now they are directed to the US which is where US GDP comes in.

Here are outbound royalty payments from Ireland to the US up to Q3 2020:

Royalty Imports to United States 2008-2020

Up, up and away!  For ten years up to 2019, these payments were little changed and typically were between €6 billion and €8 billion a year.  They surged to €9 billion in Q1 2020 alone, with the first three quarters of 2020 already totaling €28 billion.  If the Q4 figure is close to what was seen in Q3 then the annual figure will be something around €40 billion.

Previously these payments, for licenses to use technology developed in the US, went to the likes of Bermuda and the Cayman Islands, sometimes with a stopover in The Netherlands.

The profit accumulated in these jurisdictions did not contribute to US GDP but it did contribute to US GNP via factor flows in the balance of payments.  The profit is no longer being reported on small islands and the inbound factor flows to the US have been replaced by royalty income.

What are royalty imports in Ireland are royalty exports in the US.  The payments are now going to where the R&D that developed these platforms is undertaken.  Some payments would have gone from Bermuda and Cayman to the US to co-fund the R&D undertaken but the amount of these US exports would have been calculated as a share of the R&D cost. Now more of the value of the IP is flowing to the US.

So, in 2020, it looks like there be an additional €30 billion of royalties flowing from Ireland to the US.  There will be a partially offsetting reduction of around €10-15 billion in R&D exports from the US to Bermuda and the Cayman Islands etc.  So there could be a net increase in US GDP of €15-20 billion or so from these changes.

Of course, pretty much nothing will have changed in the US economy.  Just like little changed when Irish GDP recorded a 26 per cent increase.  Early estimates suggest that US GDP in 2020 was around €19 trillion, with 0.1 per cent of this being €19 billion. 

Balance of payments figures from the BEA do not seem to reflect the changes in royalty payments shown above in Ireland’s balance of payments data.  When this is incorporated it could could add something in the region of 0.1 per cent to GDP.

And, yes, the title here is a little bit disingenuous. There isn’t really anything that has happened in Ireland to change US GDP.  In overall terms the level of outbound royalties from Ireland is actually little changed.

Royalty Imports to All Countries 2008-2020

There is a reduction in the recent data and this is likely due to onshoring of some IP but outbound royalties from Ireland in 2020 are likely to have been around €70 billion, in line with the 2018 outturn. The dramatic changes of the earlier charts have to do with the destination of these royalties more than the quantum of them.

Now more of the payments are going to the US and making a larger contribution to US GDP. This better reflects the economic reality of these companies as the majority of the R&D behind their technology is undertaken in the US.

Wednesday, February 3, 2021

Further evidence of the end of the “double-irish”

A previous post looked at the changing nature of outbound royalties from Ireland.  The key point was a huge increase in 2020 in the flow of royalties from Ireland to the United States.

Royalty Imports to United States 2008-2020

Under a “double irish” structure these royalties would have gone to a Caribbean island such as Bermuda with maybe a stop-off in The Netherlands (the “dutch sandwich”).  As set out in the post there has been a huge decline in royalties following that path with an increase in royalties paid to the US as shown above.

We have more evidence of this change in Google’s annual 10K report which was filed with the SEC.  Technically it’s the 10K report of Alphabet, the owner of Google, but they are pretty much the same in practical terms.

Anyway, here is the table that shows the impact of Google’s revised structure:

Google 10K 2020 Domestic Foreign Income

This shows the split of Google income before taxes into “domestic” and “foreign”.  And we can see that, somewhat oddly, up to 2019, Google had more of its income deemed as foreign than domestic.  This is odd because most of the risks, functions and assets that generate Google’s profits are based in the US. 

Included in those foreign profits would have been the profit from the royalty flows out of Ireland.  Google’s operations in Ireland pay a royalty for the right to sell advertising on Google’s platforms to customers outside the Americas. 

This is a payment for technology developed in the US but US tax law allowed the rights to this technology to leak out of the US very cheaply.  Companies were then able to locate these licenses in no-tax jurisdictions such as Bermuda.

As Google set out in their 2020 annual report:

As of December 31, 2019, we have simplified our corporate legal entity structure and now license intellectual property from the U.S. that was previously licensed from Bermuda resulting in an increase in the portion of our income earned in the U.S. 

So now instead of making a payment that ends up in Bermuda, Google’s Irish operation makes a payment that goes directly to the US.  This structure is much more in line with the economic reality and substance of Google’s activities.

For 2020, we can see that of Google’s $48 billion of profit before tax, $37.5 billion is now classed as “domestic” and $10.5 billion as foreign.  The “double-irish” is no more.  Google is now reporting most its profit where it is earned – in the US.

Tax Impact

Does this change have an impact on Google’s taxes? Not hugely.  We can assess the impact from the table reconciling Google’s effective tax rate with the statutory 21 per cent federal corporate income tax rate in the US (some notations added):

Google 10K 2020 Tax Recon

The two key provisions we are looking for the impact of are:

  1. GILTI – Global Intangible Low-Taxed Income
  2. FDII – Foreign-Derived Intangible Income

Both of these were introduced with the Tax Cuts and Jobs Act (TCJA) in 2017.  Under the “double irish”, a significant share of Google’s profits ended up in no-tax Bermuda (no income taxes anyway).  But that didn’t mean the profit was not subject to tax.

The US operates a worldwide corporate income tax regime so US companies owe US tax on their profits wherever earned.  Google might have been able to shift the highly-valuable license for its technology from the US to a no-tax jurisdiction but US tax was still due on those profits. 

Of course, pre-TCJA, companies could defer the payment of this tax until the profit was formally repatriated back to the US, if ever.  The TCJA introduced a “deemed repatriation tax” for these historical profits but at a rate much lower than the 35 per cent that would have applied without the deferral. 

Google, and other US MNCs, have been paying this US tax on these historical profits for the past few years though the bulk of the payment is not due until the mid-2020s.  In the annual report Google says:

As of December 31, 2020, we had long-term taxes payable of $6.5 billion related to a one-time transition tax payable incurred as a result of the U.S. Tax Cuts and Jobs Act ("Tax Act"). As permitted by the Tax Act, we will pay the transition tax in annual interest-free installments through 2025.


The quote refers to the US tax due on the profits shifted to Bermuda.  The TCJA abolished the deferral provisions that meant the tax was not paid at the time (and at a higher rate) and introduced the GILTI tax for these profits (with a tip of the hat to whoever came up with the acronym!). 

We don’t need to go into the details but at its essence GILTI imposes a 10.5 per cent US tax on foreign income that is untaxed with no additional US tax due if the foreign tax on these profits exceeds 13.125 per cent.

We know that in 2018 and 2019 a large share of Google’s foreign income ended up in Bermuda.  The profit that ended up in Bermuda would be taxed at 10.5 per cent by the US regardless of whether it is repatriated or not.  Of course, 10.5 per cent is significantly lower than the statutory 21 per cent rate so having a large share of its profit taxed at 10.5% would reduce Google’s effective rate.

We can see this in the table for the item “Foreign income taxed at different rates”.  The table is annotated to show that this includes the impact of GILTI.  In 2018, having this income taxed at 10.5 per cent instead of 21 per cent knocked up to 4.4 percentage points off Google’s effective tax rate while in 2019 the impact of income taxed at lower rates was 4.9 percentage points.  Under the new structure this fell to just 0.3 percentage points in 2020.


So does this mean that the profit is now subject to tax in the US at 21 per cent.  Not quite.  We now turn to another TCJA provision, FDII –Foreign Derived Intangible Income. This is even more complex than GILTI but essentially is the tax that is applied when companies in the US sell property for a non-US person to use.

Google in the US is licensing the right to use its technology in markets around the world to Google in Ireland.  The fee that Google Ireland pays (the royalty) is taxed under the FDII provisions. This imposes a 13.1 per cent US tax on these profits (though the calculation of the taxable income is far from straightforward).

Anyway, we have moved from a situation where the profit Google earned selling to customers via Ireland was taxed at 10.5 per cent under the GILTI to this profit being taxed at 13.1 under FDII.  The impact of this can again be seen in the above table and is easy to identify with the item labelled “Foreign-derived intangible income deduction”.

The table shows that this increased in 2020 – with the increased payments from Ireland to the US – and that having this profit taxed under the FDII provisions rather than at the statutory headline rate of 21 per cent reduced Google’s effective tax rate by 3.0 percentage points.

This is lower than what was achieved under the GILTI and is part of the reason for the rise in Google’s effective tax in 2020, but we are still talking about something around one percentage point so it is not huge.

Why? And were there alternatives?

There are many reasons why Google has done this including changes in tax law in Ireland and the US.  An overarching reason is the OECD’s BEPS project.  One of the aims of this project is to try to ensure that company profits are better aligned with substance.  Google has no substance in Bermuda so declaring a large share of its profit there did not align with its substance.

Transfer pricing rules have rightfully changed to reduce firms’ ability to do this.  Of course, getting the license out of the US and into the no-tax jurisdiction in the first place is the ultimate reason such profits ended up there but the OECD has no jurisdiction over the US transfer pricing rules which allowed it.

There is a reason why profits from US firms dominate in places like Bermuda as shown in the OECD’s publication of data from country-by-country reports.

OECD CbCR data for Bermuda 2016 Chart

This is from a limited sample of parent jurisdictions but the overall picture is unlikely to change as the OECD publishes later updates of this data for subsequent years.

Google didn’t have to relocate the license from Bermuda to the US. In theory they could have moved it to any country in the world.  In practice, they were limited to locations they have substance.  One country where Google has substance is Ireland and they could have transferred the license to Ireland.

We know that some MNCs have done this with a subsidiary in Ireland buying the license.  The outlay under this transaction becomes eligible as a tax deduction under Section 291A of the Consolidated Tax Acts.  And there has been a huge increase in the amount of capital allowances claimed under S291A in recent years, most notably in 2015.

Google could have done this.  With the 80 per cent cap on the amount of capital allowances that can be claimed in any single year it would have faced an effective tax on these profits of 2.5 per cent in Ireland with additional tax due to the US under the GILTI provisions bringing it up to around 11 per cent.

Google chose not to do this. Google chose to relocate the license to the US and have the profit taxed under the FDII provisions of the TCJA. As noted above this results in a slightly higher effective tax on these profits.

Why not use S291A in Ireland and GILTI in the US to cut its tax bill?  It’s possible that Google determined it did not have sufficient substance in Ireland to warrant locating the profit linked to its licenses in Ireland. 

Google could have moved the license to Ireland but may have faced difficulties, including from tax authorities in market countries, if it could not show that it has sufficient substance in Ireland to justify the profits being reported there.  By moving the license to the US, Google is very unlikely to face such difficulties.

If we look at the onshoring of IP to Ireland it does appear to be more prevalent among manufacturing companies.  Pharmaceutical companies have multi-billion euro plants in Ireland.  It is much easier for them to show they have substance in Ireland given the role manufacturing plays for pharmaceutical and other industrial companies. 

Can Google really say that its profit is generated in Ireland? Pointing to a couple of empty office buildings in Dublin is unlikely to pass muster.

The end of the “double-irish”

Anyway there we have it. The end of the double irish.  And we can see the same if we look at other companies.  Here is the domestic/foreign split from Facebook’s 10K report published last week.

Facebook 10K 2020 Domestic Foreign Income

Again we see the significant switch from foreign to domestic. This is because Facebook transferred the license to sell advertising on its platforms from the Cayman Islands back to the US.  This was incorrectly reported as Facebook transferring assets out of Ireland as we discussed here

We could go through the same GILTI/FDII rigmarole for Facebook as we did for Google, but we won’t.  The conclusion is simple: the “double-irish” is dead.  Its demise has not seen tax bills soar.  But I doubt we’ll see many headlines about that.