This chapter provides a special study on the convergence of tax-to-GDP ratios and of tax structures in OECD countries from 1995-2016.
Revenue Statistics 2018
Revenue Statistics
Chapter 2. Special feature: Convergence of tax levels and tax structures in OECD countries
Abstract
2.1. Introduction
Tax levels, measured by the tax-to-GDP ratio, and tax structures, measured by tax categories as a share of total tax revenues, vary across OECD countries for a variety of historical, economic and political reasons, including choices on how to develop and fund the welfare state. However, since 1995, the trend of the OECD average1 tax-to-GDP ratio has shown similarity to that of many countries: dipping following the financial crisis before resuming a slow and long-running increase. Similarly, over this period a number of tax policy changes have impacted tax structures in OECD countries, including the introduction of value-added taxes in all but one OECD country, and the developing consensus around the impact of different taxes on growth and the consequent effort of many countries to implement these recommendations (OECD, 2010[1]).
This chapter explores the impact of these and other changes on the distribution of tax levels and tax structures in OECD countries, with particular attention to whether the OECD average is becoming more or less representative of the underlying distribution over time. The chapter considers sigma measures of convergence to examine the dispersion of the tax-to-GDP ratio in member countries in the years from 1995 to 2016 before examining intra-OECD groups at different levels of tax revenues. It then calculates an index of the similarity of all countries to the OECD average tax structure, assessing changes in tax structures across the period. The paper concludes by examining the combined changes in the dispersion of tax levels and structures around the OECD average.
2.2. OECD countries have higher and more similar tax levels in 2016 relative to 1995
Since 1995, the OECD average tax-to-GDP ratio has increased from 33.0% to 34.0% in 2016, with the lowest point of the period (32.2%) occurring in 2009 following the financial crisis. In this time, the highest tax-to-GDP ratio was seen in Sweden, at 49.0% in 20002, although for most of the period, the highest tax rate in each year was observed in Denmark (ranging from a low of 44.8% to a high of 48.6%). The lowest tax-to-GDP ratio was consistently seen in Mexico (ranging between 9.9% and 16.6% across the period).
Across the OECD, the dispersion of tax-to-GDP ratios decreased between 1995 and 2016, with a brief interruption in 2009, a period which coincided with the lowest average OECD tax-to-GDP ratio in the period. Figure 2.1 illustrates that the dispersion of tax-to-GDP ratios decreased around the OECD average tax-to-GDP ratio (standard deviation and coefficient of variation) as well as around the median (absolute deviation) and when country pairs are considered (Gini coefficient); with these measures of dispersion showing similar trends across the period. An analysis of year-on-year growth rates for these indicators supports this.
Figure 2.1. Dispersion of tax-to-GDP ratios in OECD countries, 1995-2016
Note: The figures on the horizontal axis (above the years) show the OECD average tax-to-GDP ratio in each year.
Source: Authors’ calculations based on OECD Revenue Statistics (2017) (OECD, 2017[2]).
The interquartile range shows a different trend over this period, due to changes in the tax levels of the top quartile of OECD countries. Conceptually, movements of the interquartile range and the other indicators in opposite directions can be explained if one of the outer quartiles is moving away from the other while converging among itself. This occurred, for example, between 2005 and 2008, where all other indicators showed decreased dispersion and the interquartile range increased. In these years, the threshold for the upper quartile increased, moving further away from the main distribution, but at the same time there was a decrease in the ratios of the four countries with the very highest tax levels (and a particularly sharp decrease in the two top countries). The combined impact of the increasing threshold for the top quartile, together with the reductions in the tax-to-GDP ratios of the highest countries, decreased dispersion around both the OECD average and median even as the interquartile range increased. A similar phenomenon occurred in 2000, where the threshold for the lower quartile decreased even as countries with the lowest tax-to-GDP ratios increased these.
2.3. Groups within the OECD are emerging
In 2016, OECD countries fell into four groups with different tax levels: low (tax-to-GDP ratios less than 30%)3; low-mid (tax-to-GDP ratios between 30 and 35%); high-mid (tax-to-GDP ratios between 35 and 40%) and high (tax-to-GDP ratios greater than 40%) (Figure 2.2). 4
Box 2.1. Measuring convergence of tax levels
There is an extensive discussion of convergence of tax revenues in the literature, which has been born out of neoclassical frameworks of economic growth (Barro and Sala-I-Martin, 1992[3]) and (Mankiw, Romer and Weil D., 1992[4]). Convergence may be measured by beta convergence, which measures the rate at which poorer economies converge toward richer economies; sigma convergence, which measures the dispersion across a cross-section of economies using a variety of indicators; and gamma convergence, which measures changes in the ranking of economies over time. These approaches have been extended to estimate convergence of tax levels across countries.5
This special feature focuses on sigma convergence, measuring the dispersion of the tax-to-GDP ratio across OECD economies in the years from 1995 to 2016. Key indicators of sigma convergence include:
Standard deviation: a simple measure of dispersion from the mean, calculated as the square root of the variance (which averages the sum of squared residuals from the mean) and is measured in the same unit as the variable. It is affected by outliers.
Coefficient of variation: the standard deviation as a percentage of the mean. It has the advantage of standardising across data with different means, but is exposed to outliers as they affect both the standard deviation and the mean. It is also driven by the value of the mean in the denominator, e.g. very large values of the mean can lower the coefficient of variation while keeping the SD constant.
Absolute deviation from the average: the sum of the absolute distance of each country from the average (either mean or median), divided by the total number of countries. This indicator is measured in the same unit as the data. If the median is used as the central point, the indicator is outlier-resistant.
Gini coefficient: calculates the distance between all country pairs and therefore is a better measure for inter-country dispersion. It is normalised by dividing by twice the number of data points and the mean and is expressed as a number between 0 and 1, with 0 indicating no dispersion. It can also be displayed in a Lorenz curve. It is less affected by outliers due to the use of country pairs.
Range measures: the difference between two points in the distribution. The interquartile range is commonly used, which measures the difference between the upper (Q3) and the lower quartiles (Q1). Half of the data falls within this range. This measure is also resistant to outliers and avoids bias inherent in the mean.
Figure 2.2. Distribution of tax-to-GDP ratios in OECD countries, 2016
Note: Each + represents the tax-to-GDP ratio of an OECD country in 2016.
Source: Authors’ calculations based on OECD Revenue Statistics (2017).
If these groups are considered across the period from 1995 to 2016, the composition of the upper and lower groups was relatively stable. Among the countries with low tax-to-GDP ratios in 2016, only Ireland did not have a tax-to-GDP ratio under the threshold of 30% in 1995, but did from 2001 onwards with a brief exception in 2006 and 2007. In the countries with high tax-to-GDP ratios in 2016, all countries in that group also had tax-to-GDP ratios above 40 in 1995, and only one country fell below this threshold at any point during the period (Italy, in 2002 and 2004-5).
There was more change across the period for the two middle groups, with the composition of both groups becoming more stable after around 2000. For the low-mid tax-to-GDP ratio group (30-35% of GDP in 2016), all but three countries had tax-to-GDP ratios within the thresholds by 2000: each of these three countries saw an increasing tax-to-GDP ratio throughout the period, with Japan’s tax-to-GDP ratio increasing above the threshold in 2014 and Latvia’s in 2016; Lithuania in 2016 had a tax-to-GDP ratio of 29.9%. After 2000, a few of the other countries spent brief periods outside the thresholds: Estonia, Israel, Portugal and Spain had tax-to-GDP ratios below the threshold for this group in one year, and the Slovak Republic for seven; and Estonia and New Zealand briefly exceeded the upper threshold (from 2005-2007 in Estonia and 2005-2006 in New Zealand).
The high-mid group, comprised of countries with tax-to-GDP ratios of between 35% and 40% in 2016, also had all but two countries within this range from 19996: until 2012, Greece’s tax-to-GDP ratio was below 35% and prior to 1998, it was below 30%. Norway initially had a tax-to-GDP ratio over 40% but this dropped under the upper threshold for this group in 2013. After 1999, a few other countries in this group had tax-to-GDP ratios briefly outside the thresholds, including the Netherlands, Germany and Iceland, all prior to 2012.
The movement of countries’ tax-to-GDP ratios across the period is indicated in Figure 2.3, which shows tax-to-GDP ratios in 1995 and 2016, as well as the minimum and maximum tax-to-GDP ratio for each country, regardless of year (in the shaded area).
Figure 2.3. Tax-to-GDP ratios in OECD countries: 1995, 2016; minimum and maximums between these dates
Note: Figures shown for Iceland in 2016 exclude the one-off revenues from the stability contributions. If included, the tax-to-GDP ratio for Iceland in 2016 would be 51.6%, as shown in the first chapter.
Source: Authors’ calculations based on OECD Revenue Statistics (2017).
Throughout the period, divergence within these four groups was highest in the group with the lowest tax-to-GDP ratios, although dispersion in this group narrowed considerably across the period. Another strong decrease in dispersion was seen in the high-mid tax-to-GDP ratio countries from 2010 to 2015, concurrently with an increase in its mean tax-to-GDP ratio (Figure 2.4). This increase was partially explained by increases in the tax-to-GDP ratios in Greece and Iceland, which initially had lower tax-to-GDP ratios, and smaller increases in most other countries in the group, with the exception of Norway.
With one exception, tax-to-GDP ratios in each of the four groups became less dispersed around a higher mean tax-to-GDP ratio in 2016 than in 1995. The exception was the group with low-mid tax-to-GDP ratios, where although divergence decreased, the mean tax-to-GDP ratio has not yet recovered from its drop during the financial crisis and remains lower than in 1995 (Figure 2.4), despite the fact that the two countries initially below the threshold had by 2016 increased their tax-to-GDP ratios past 30%. Of the thirteen countries in this group, seven had lower tax-to-GDP ratios in 2016 than in 1995; by contrast, only nine countries in the rest of the OECD experienced falls in this timeframe.
Figure 2.4. Dispersion and mean tax-to-GDP ratios by intra-OECD groups, 1995-2016
Note: Top-left: countries with a tax-to-GDP ratio of less than 30% in 2016 (8 countries); top-right, countries with a tax-to-GDP ratio of between 30% and 35% in 2016 (12 countries); bottom- left, 35-40% (8 countries) and bottom-right, greater than 40% (7 countries). Standard deviation and Gini coefficients are shown on the left hand axis (the Gini is scaled by a factor of 40 for comparability) and the mean tax-to-GDP ratio is on the right-hand axis.
Source: Authors’ calculations based on OECD Revenue Statistics (2017).
Dividing the OECD into these groups also demonstrates why dispersion for the OECD as a whole increased in 2005. This was due to an increase in dispersion among countries with high tax-to-GDP ratios: in 2005, the ratios of the two highest countries (Sweden and Denmark) increased sharply relative to 2004, before falling back in 2006. At the same time, Austria’s tax-to-GDP ratio, one of four situated in the middle of this range, fell in 2005 before being caught up to in 2006 by Italy, the country with the lowest tax-to-GDP ratio in this group.
The combination of these trends meant that by 2016, countries with high and high-mid tax levels had converged towards each other and away from low and low-mid countries, which also became less divergent and moved away from the group of countries with higher tax-to-GDP ratios. Most of the movement in tax levels occurred between 1995 and 2009, when the gap between countries with high and high-mid ratios and countries with low and low-mid ratios widened. The two lower groups were also more impacted by falls in the tax-to-GDP ratio during the financial crisis. The difference between the group means has since stayed relatively constant: since 2009, all groups have increased their mean tax-to-GDP ratios at similar rates except for the high ratio group, where it has been decreasing slowly since 2014 (Figure 2.5).
Figure 2.5. Mean tax-to-GDP ratios across countries with low, low-mid, high-mid and high levels, 1995-2016
Source: Authors’ calculations based on OECD Revenue Statistics (2017).
2.4. Tax structures converge toward higher shares of VAT, SSCs and CIT in total taxes
Across the same period there has been a large amount of change in the tax structures of OECD countries and the OECD average, measured as the decomposition of total tax revenue by eight major tax categories (Box 2.2). Across the period, there has been a shift away from personal income taxes and non-VAT consumption taxes, toward higher social security contributions and VAT.
In 1995, the predominant shares of the OECD tax structure were social security contributions and personal income taxes (26.2% and 25.0% of total tax revenues, respectively) with corporate income taxes at 8.1% of total tax revenue. Value-added taxes were already the predominant share of consumption tax revenues (at 18.1% of total tax revenues) but taxes on imports and exports, and other consumption taxes continued to play a significant role (15.8% of total tax revenue, combined).
Box 2.2. Measuring convergence of tax structures
Tax structures are measured by a second key indicator in Revenue Statistics: a tax category as a share of total revenue. The OECD Interpretative Guide provides a highly-detailed breakdown of tax revenues into different tax categories. This special feature divides tax revenues into eight categories as set out in Table 2.1.
Table 2.1. Disaggregation of tax structures used to construct D-index
Source: Authors’ calculations based on OECD Revenue Statistics (2018).
Capturing the similarity of tax structures comprised of a number of parts in a single indicator is difficult. However, composite indices can be used to aggregate the different tax categories that make up total tax revenue and provide a single indication of similarity. These include:
The D-index (Delgado, 2013[5]): this calculates the absolute difference for the share of each tax category in a country from its share in the OECD and sums it, providing an indicator of the difference of that country’s tax structure from the OECD average tax structure. Consequently, a value of 0 indicates that the country’s tax structure is the same as the OECD average structure.
The structural similarity index (Becker and Elsayyad, 2009[6]): this calculates the minimum share of each tax category for every country pair in the OECD and, to generate the indicator for an individual country, averages it for the number of pairs. Consequently, a value of 0 indicates that there are no similarities between that country and any other country.
This special feature focuses on the D-index, which measures the distance of each country’s tax structure from the OECD average.
By 2016, personal income taxes and social security contributions continued to be the two most significant sources of revenue, on average, but the importance of social security contributions had grown and personal income taxes fallen (to 27.2% and 23.8%, respectively). The share of value-added taxes (VAT) also increased (to 20.2%) whereas other consumption taxes, and particularly taxes on imports and exports, fell. The fall in non-VAT goods and services taxes was not quite offset by the increase in VAT.
The direction of these changes is not sensitive to the start- or end-year within a five-year period in either direction, although the magnitudes differ depending on the start and end-point. However, although the share of corporate income taxes was higher in 2016 than in 1995, the trend across the period was less clear as corporate income taxes were more volatile across the period than the other categories: starting at 8.1% of total revenues in 1995, and increasing to between 8.5% and 8.7% from 1997 to 2003, before peaking at 11.1% in 2007 and falling back to between 8.6 and 9.0% from 2009 to 2016.
Tax structures across individual OECD countries became closer to the OECD average structure across the period 1995 to 2016, as measured by the D-index (Box 2.2). The mean country distance from the OECD average tax structure declined constantly between 1995 and 2004, with a particularly steep decrease in the first few years, partially impacted by the introduction of VAT in Slovenia (1999) and Australia (2000). Tax structures diverged from the OECD average before and immediately after the financial crisis (Figure 2.6) before resuming (in 2011) a descent towards their lowest level of divergence across the period in 2016.
Figure 2.6. Mean difference of country tax structures to OECD average structure (D-index), 1995-2016
Source: Authors’ calculations based on OECD Revenue Statistics (2017).
All but eight OECD countries were more similar to the OECD average tax structure in 1995 than in 2016 (Figure 2.7). These countries were Lithuania, Poland, the United States, Italy, Mexico, New Zealand, Hungary and Estonia:
in Lithuania, where the largest change in the D-index was observed, this was caused by a sharp increase in the share of social security contributions between 1995 and 2016 and a fall in personal income taxes, which in both cases were significantly greater than the corresponding change in the OECD average;
in Poland this was due to an increase in VAT (measured either as a share of tax revenue or of GDP) that was higher than the change for the OECD on average. This also mechanically decreased the share of personal income taxes away from the OECD average share;
in Italy, where corporate income taxes fell as a share of total tax revenue and GDP, countering the increase seen for the OECD average;
in Mexico, where increases in personal and corporate income taxes as a percentage of GDP meant that the share of social security contributions in total tax revenues fell away from the average OECD share; the increase in corporate income taxes also moved further away from the OECD average, whereas the increase in personal income taxes approached the OECD average;
in the United States, where personal income taxes increased as a share of total tax revenues, and further away from the OECD average; and the increase in VAT in the OECD average structure also widened the difference;
in Estonia, where personal income taxes fell more quickly than the OECD average share between 1995 and 2016, and where the share of corporate income taxes decreased (as opposed to the increase in the OECD average share);
in New Zealand, where a sharp fall in the share of personal income taxes toward the OECD average was more than offset by an increase in corporate income taxes and VAT, away from the OECD average share; and
in Switzerland, where the D-index increased slightly as a result of higher growth of the share of corporate income taxes in Switzerland than in the OECD average, but no change in the share of VAT in Switzerland when it increased for the average.
Figure 2.7. Distance from the OECD average tax structure (D-index), 1995 and 2016
Source: Authors’ calculations based on OECD Revenue Statistics (2017).
Countries that increased their similarity to the OECD average the most were those that introduced VAT (Slovenia and Australia) or that made large changes in social security contributions (e.g. in Latvia where they fell as a percentage of GDP and total taxes, or in Korea, where they increased strongly in both) that drew the share of these revenues toward the OECD average.
Similarly, of the five countries where the distance from the OECD average is the highest (Figure 2.8), four do not levy taxes in one of the major categories: Australia and New Zealand do not levy social security contributions and in Denmark, revenues from social security contributions are less than 1% of total tax revenue; the United States does not have a VAT. However, the country least similar to the mean in 2016 was Chile.7 The differences were due to higher shares of corporate income tax and VAT revenues (20.9% and 41.2% of total revenues, respectively, although the differences as a percentage of GDP were more comparable) and a corresponding low share of personal income taxes and social security contributions (8.8% and 7.2%, respectively, with both also lower than the OECD average as a percentage of GDP).
Across the OECD, there are six countries in 2016 with tax structures that differ by less than 20% from the OECD average structure (Figure 2.7and Figure 2.8). These are all EU countries with the exception of Norway.
Figure 2.8. Tax structures in the OECD ordered by proximity to the OECD average, 2016
Note: Countries are ordered by increasing proximity to the OECD average structure (right). Numbers underneath the country names show the D-index for each country in 2016.
Source: Authors’ calculations based on OECD Revenue Statistics (2017).
2.5. OECD average tax level and structure are more representative
The combined impact of these changes means that the OECD average tax-to-GDP ratio and the OECD average tax structure are more representative in 2016 than they were in 1995. Dispersions of tax-to-GDP ratios from the OECD average (whether measured by standard deviations, coefficient of variation, or absolute deviations from the mean), are currently at a comparatively low level and near their lowest level across the period. Similarly, the mean distance of OECD countries from the OECD average tax structure is at its lowest level since 1995 (Figure 2.6).
This is illustrated in Figure 2.9, which shows the distribution of all OECD countries by reference to the OECD average. Each country is placed along the horizontal axis by their tax-to-GDP ratios in 1995 (left-hand panel) and 2016 (right-hand panel). Along the vertical axis, the graphs show the position of each country on the D-index in each year, with proximity to zero indicating proximity to the OECD average. The OECD average tax-to-GDP ratio is shown on the horizontal axis with a D-index of zero. The inner squares shown on each graph include countries that are located within five percentage points of the OECD average tax-to-GDP ratio in each year as well as having a D-index of less than 20. The outer square is set at the boundary of plus or minus ten percentage points of the OECD tax-to-GDP ratio and a D-index of less than 40.
Although these boundaries are arbitrary, they serve to highlight the increasing convergence of OECD countries towards the OECD average tax level and tax structure. In 1995, only one country fell within the inner box; whereas seventeen fell outside both boxes. In 2016, five countries fell within the inner box; and only ten fell outside both.
Figure 2.9. Country distances from OECD average tax levels and structures, 1995 and 2016
Note: The boxes show an arbitrary indication of proximity to the OECD average in each year: +/- 5 percentage points and a D-index of less than 20; and +/-10 percentage points and a D-index of less than 40.
Source: Authors’ calculations based on OECD Revenue Statistics 2017.
2.6. Conclusions
From 1995 to 2016, the distribution of countries around the OECD average tax-to-GDP ratio and average tax structure has become more similar and in 2016, the OECD average was more representative of the underlying distribution than at any point in the preceding 20 years. OECD countries have converged around a higher level of taxation: as the dispersion of tax-to-GDP levels in the OECD has decreased, the levels of the tax-to-GDP ratio have also increased and were one percentage point higher in 2016 than in 1995.
Among OECD countries, however, there remains considerable heterogeneity of tax levels, and in 2016, four main groups of OECD countries were evident: those with low, low-mid, high-mid, and high tax-to-GDP ratios. Within these four groups, countries with low tax-to-GDP ratios have the greatest divergence in tax-to-GDP ratios, but their differences have narrowed considerably and their tax-to-GDP ratios increased quickly across the period. In the high-mid group, by contrast, the mean tax-to-GDP ratio has increased slowly but converged quickly since 2010. All of the four intra-OECD groups have become more similar across the period and have higher tax-to-GDP ratios than in 1995, with the exception of the low-mid group, where the tax-to-GDP ratio has not yet recovered to pre-crisis peaks and in 2016, remains lower than it was in 1995.
In 2016, the gap between countries with high and high-mid, and with low and low-mid, tax-to-GDP ratios was higher than it was in 1995, in part because tax levels in countries with lower tax-to-GDP ratios were more impacted by the financial crisis in 2008. Between 1995 and 2007, the gap in the mean tax levels of countries with high & high-mid ratios and countries with low and low-mid ratios widened, and has since stayed relatively constant. Since 2009, all groups have increased their mean tax-to-GDP ratios at similar rates except for the high ratio group, where it has been decreasing slowly since 2014, although it remains above 1995 levels.
OECD countries have also converged towards a tax structure that places more emphasis on value-added taxes, social security contributions, and, to a lesser extent, on corporate income taxes; while moving away from personal income taxes and non-VAT forms of consumption taxes. In 1995, the mean D-index score across OECD countries was 37.7, compared to 32.8 in 2016, indicating a higher degree of convergence of OECD countries with the OECD average. Only eight countries have moved further away from the OECD average structure over the period (Lithuania, Poland, the United States, Italy, Mexico, New Zealand, Hungary and Estonia). For almost all countries, and on average, the OECD average tax level and tax structure is therefore now more representative than it has been at any point in the last 21 years.
Individual countries differ in their distance from the OECD average, with six countries having a D-index of under 20, indicating proximity to the OECD average tax structure: Portugal, Norway, Finland, Luxembourg, Spain and Latvia. By contrast, the countries with the most different tax structures to the OECD average were Chile, New Zealand, Denmark, Australia and the United States, each with a D-index greater than 50.
Future analysis could explore links between tax structures and tax levels and whether the intra-OECD groups seen by tax level also have characteristic structures or have been converging internally towards more similar tax structures. This analysis could also seek to understand the drivers of different tax levels and structures in OECD countries: for example, the link between social benefit systems and the level of social security contributions, or the impact of the privatisation of infrastructure spending on revenue needs and structure. Similarly, exploring the structural similarity index could allow relationships between the tax structures of country pairs to be examined and could lead to the development of country groups by reference to the composition of tax revenues.
References
[3] Barro, R. and X. Sala-I-Martin (1992), “Convergence”, Journal of Political Economy, Vol. 100, pp. 223-251, http://dx.doi.org/10.1086/261816.
[6] Becker, J. and M. Elsayyad (2009), “The evolution and convergence of OECD tax systems”, Intereconomics, Vol. 44, pp. 105-113, http://dx.doi.org/10.1007/s10272-009-0284-0.
[5] Delgado, F. (2013), “Are Taxes Converging in Europe? Trends and Some Insights into the Effect of Economic Crisis”, Journal of Global Economics, Vol. 1, pp. 1-3, http://dx.doi.org/10.4172/2375-4389.1000102.
[7] Delgado, F. and M. Presno (2011), “Convergence of fiscal pressure in the EU: a time series approach”, Applied Economics, Vol. 43, pp. 4257-4267, http://dx.doi.org/10.1080/00036846.2010.491449.
[11] Dvorokova, K. (2014), “Sigma Versus Beta-convergence in EU28: do they lead to different results?”, Mathematical Methods in Finance and Business Administration, pp. 88-94, http://www.wseas.us/e-library/conferences/2014/Tenerife/ECONMATH/ECONMATH-13.pdf.
[9] Gemmell, N. and R. Kneller (2003), “Fiscal Policy, Growth and Convergence in Europe”, New Zealand Treasury Working Paper, Vol. 14, pp. 1-28, https://treasury.govt.nz/sites/default/files/2007-09/twp03-14.pdf.
[4] Mankiw, G., D. Romer and Weil D. (1992), “A contribution to the empirics of economic growth”, The Quarterly Journal of Economics, Vol. 107, pp. 407-37, https://eml.berkeley.edu/~dromer/papers/MRW_QJE1992.pdf.
[2] OECD (2017), Revenue Statistics 1965-2016, OECD Publishing, Paris, http://dx.doi.org/10.1787/10.1787/9789264283183-en.
[1] OECD (2010), Tax Policy Reform and Economic Growth, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264091085-en.
[8] Tibulca, I. (2015), “Is there Evidence of Tax Convergence in the European Union?”, Procedia Economics and Finance, Vol. 32, pp. 194-199, https://doi.org/10.1016/S2212-5671(15)01382-9.
[10] Young, A., M. Higgins and D. Levy (2008), “Sigma Convergence versus Beta Convergence: Evidence from U.S. County-Level Data”, Journal of Money, Credit and Banking, Vol. 40, pp. 1083-1093, https://www.biu.ac.il/soc/ec/d_levy/wp/jmcb5.pdf.
Notes
← 1. In this special feature, as in the rest of the publication, the term “OECD average” refers only to the unweighted mean of OECD countries and specifically to the two key indicators used throughout the publication: i.e. the OECD average tax-to-GDP ratio or the OECD average tax structure. Otherwise, the term “average” is avoided and reference is instead made to the mean or median for clarity. All means presented are unweighted.
← 2. In 2016, Iceland received revenues from one-off stability contributions that increased its tax-to-GDP ratio by 15.7 percentage points, to 51.6% of GDP. As the special feature considers trends over time, these one-off revenues have not been included in the analysis. For the purposes of this special feature, the tax level and structure in Iceland has been calculated exclusive of the revenues from these one-off contributions.
← 3. Lithuania, with a tax-to-GDP ratio of 29.9% in 2016, has been included in the low-mid group for this analysis. The country with the next highest ratio is Switzerland, with a ratio of 27.8% in 2016, which has been included in the group of countries with low tax-to-GDP ratios in 2016. Future analysis could consider whether it would be more appropriate to further divide the lowest group, particularly if additional countries were to be included.
← 4. Had the groups been constructed on the basis of data for 2015 they would have been the same, with two exceptions. Both Latvia and Lithuania would have been included in the low group rather than the low-mid group, as their tax-to-GDP ratios were 28.9% and 29.2% in 2015.
← 5. See for example (Delgado and Presno, 2011[7]); (Dvorokova, 2014[11]); (Young, Higgins and Levy, 2008[10]); (Gemmell and Kneller, 2003[9]); (Tibulca, 2015[8]).
← 6. Iceland was below the thresholds for this group from 2001-2002 and 2009-2011 and above them in 2006; and Germany had a tax-to-GDP ratio of below 35% from 2002 to 2007.
← 7. Although Chile, and to a lesser extent New Zealand, Denmark and Australia are outliers, they do not affect the overall trend of the indicator, although do increase the average distance. When Chile is excluded from the calculation, the OECD-35 average D-index in 1995 was 36.4 and in 2016 it was 31.6 (compared to 37.7 and 32.7 with Chile included). The trend over time is also similar.