Chapter 2. Special feature: The taxation of median wage earners
Introduction
Taxing Wages compares labour costs and the tax and benefit position of different household types in OECD countries, using countryspecific average wages1 as a reference point to enable comparisons across countries. The mean wage, however, is significantly influenced by the fact that wage differentials are much larger in the upper part of the wage distribution (above the median wage). Because of this, the mean does not necessarily reflect the wage of the representative individual. This sensitivity is increasingly important in the light of growing wage inequalities in many OECD countries.
An alternative measure of the average wage is the median wage. This divides the labour income distribution into two parts, with 50% of workers at lower wage incomes and 50% of workers with wage incomes above the median wage. Unlike the mean, which puts disproportionately greater weight on high earning individuals and falls at a different point in the wage distribution of every country, the median wage is not influenced by wage differentials in the upper part of the wage distribution and is hence smaller than the mean wage in the presence of labour income inequality.
This Special Feature examines the taxation of single median wage earners in OECD countries. For simplicity, the term “average wage” in this Special Feature is a mean and corresponds to the average wage used in the rest of the Report. The Special Feature begins by setting out the conceptual differences between the mean and median wage and the methodology used to estimate median wages for the purposes of this analysis. It then shows the headline indicators (net personal tax rates and the tax wedge) for median wage earners in OECD countries, contrasting these indicators with the same indicators for the average wage earner. To examine the case of a lowincome worker, it then presents the same indicators for a worker at 67% of each of the median and average wage. Finally, it considers the composition of the tax wedge for the median and average worker at 67% and 100% of each wage, respectively.
Differences in the average and median wage in OECD countries
The difference between the mean and the median wage is influenced by the level of labour income inequality in each country: the more unequal the distribution of wages, typically, the higher the divergence of mean and median labour income. Since 1995, the mean and median wage in OECD countries have diverged in the presence of increasing inequality, as shown in Figure 2.1. For most countries, the growing difference reflects increased wage differentials in the upper part of the distribution in combination with constant inequality in the lower part. One popular explanation for this is technological change; see OECD (2017) for a discussion of changes in wage inequality and the link to technological change.
The median wage, therefore, may be considered a more comparable reference point for the distribution of wage levels across countries. While the mean wage reflects the total level of labour income and the number of workers in the country, the median wage reflects the labour income of the representative individual at the midpoint of the income distribution. It partly accounts for crosscountry differences in wage inequality, falling at the same point in the income distribution of each country. Using the median wage in the analysis results in a comparison of taxation at the same point of the income distribution in each country.
However, there are practical difficulties associated with the use of the median wage as a reference wage level for the Taxing Wages household types. The median wage requires comparable data on the wage distribution of a consistent group of fulltime workers in each country, making data on the median wage harder to obtain than the mean wage. Obtaining median wages on a comparable basis to the mean wages used in Taxing Wages is more difficult still (manual and nonmanual, sectors BN inclusive). Therefore, this chapter draws on several data sources to approximate the median wages based on mean wage growth rates, as described in Box 2.1.
The mean wage is the simple average of all wage income of fulltime employees and requires only information on the total wage income and the number of fulltime employees in a country. Calculating the median wage requires additional information on the income distribution, making it harder to obtain timely and comparable data.
There is no standardised OECD wide data set on median earnings, and only a few countries were able to provide median earnings that aligned with the methodology of the Taxing Wages average wage (TW AW) for 2017 (Belgium, Sweden and Switzerland). The data for median earnings of the remaining countries is drawn from two databases:

Median earnings for all EU OECD member states as well as Iceland, Norway and Turkey come from the Structure of Earnings Survey (EUROSTAT, 2014[1]), which is conducted every four years by Eurostat. The survey provides structural data on gross earnings, including data on median earnings. The median wage captures all economic activities with the exception of agriculture, public administration and defence (NACE Rev. 2 sections B to S (excluding O)) for companies with at least 10 employees. The latest year for which data is available is 2014.

Data on median wages for the remaining OECD countries (Australia, Canada, Chile, Israel, Japan, Korea, Mexico, New Zealand and the United States) is taken from an internal earnings distribution database of the OECD Directorate for Employment, Labour and Social Affairs (ELS), which unites data from several different sources (such as national statistical institutions). Due to the variety of methodologies used, there is no unique definition of median wages and each of the data points is country dependent. Thus, there is a large variation in methodology, indicators (gross hourly, weekly, monthly or annual wage), type of worker (fulltime, fulltime equivalent, or fulltime and parttime worker) and company coverage (inclusion or exclusion of companies with fewer than 10 employees). Likewise, the latest year available varies across countries and data is available for either 2015, 2016 or 2017.
One of the biggest issues with the median wage data set used in this Special Feature is its timeliness. The limited availability of median wage for 2017 requires estimations for the majority of the countries for the analysis in this chapter. The methodologies also differ, e.g. differences in sector coverage (inclusion of the public sector), company size (exclusion of companies with fewer than 10 employees) or the type of worker (fulltime, fulltime equivalent, fulltime and parttime worker) relative to the TW AW. However, since both data sets (the SES and the ELS data set) contain median and average wage data for each of the countries, the ratio of the two indicators can be calculated and applied to the TW AW to estimate the median wage for the respective year in a manner that is consistent with the average wage used in the rest of Taxing Wages. To adjust the median wage to 2017, the growth rate of the TW AW between the base year and 2017 is applied.
The approach allows for methodological consistency for each country since the median wage has been estimated by reference to the same group of workers as the average wage to which it is compared. However, this estimation method implies that the ratio of the median and the average wage remains unchanged, regardless of the underlying methodology of the data set, and that the average and median wage have grown at the same rate. This assumption is unlikely to be correct and will also differ between countries.
The different steps of the estimation are illustrated below with the example of Finland:
This methodology is used to ensure comparability across countries and with the average wages used in this report. Due to differing methodologies and sectors or employees covered, country estimates may differ from the estimates used in this analysis.
Using the estimations described in Box 1, the median wage was on average 80.8% of the average wage in OECD countries in 2017, as shown in Table 2.1. In 23 OECD countries, the median wage was between 80.8% and 90% of the average wage; and the dispersion of median wages was greater at the lower end of the distribution: the remaining thirteen countries had median wages of less than 80% of the average wage in 2017, with the lowest median wage, measured as a share of the average wage, occurring in Turkey, at 63.2%.
Table 2.1 shows the average and median wages used in this Special Feature, as well as 67% of each wage level. The distribution of median wages across OECD countries in 2017 was similar to the distribution of the average wage. Both the median and average wage were highest in Switzerland and lowest in Mexico (measured in purchasingpower parity adjusted figures), and the top and bottom five countries remain the same. The OECD average median wage of 36 314 (USD, PPP adjusted) represented 80.8% of the average wage of 44 475. The largest differences between the average and median wage in currency terms were observed in Luxembourg (USD 15 956) and the United States (USD 14 931).
Tax burden indicators for workers on the median and average wages, 2017
This section presents results for the single earner at 100% of the median and average wage in 2017, referred to respectively as the “median worker” and the “average worker”.2 It considers the net personal tax rates and tax wedges that applied to median workers in OECD countries in 2017 and contrasts these with the rates that applied to the average worker.
Net personal tax rates of the median and average worker
Net personal tax rates include the combined impact of income taxes, social security contributions (SSCs) paid by employees and cash benefits, measured as a percentage of gross wages. They can be measured as marginal rates (the amount payable on the next unit of currency earned) or as average rates by dividing the total amount payable by the gross wage.
Figure 2.1 sets out the net personal tax rates for median and average workers in OECD countries in 2017. Net personal marginal tax rates (NPMTRs) are shown in the upper panel; and net personal average tax rates (NPATRs) are shown on the lower panel.
Figure 2.3 demonstrates that NPMTRs on the median wage in OECD countries in 2017 ranged from 55.6% in Belgium to 7.0% in Chile in 2017, with the distribution of rates being concentrated between 30% and 35% (14 countries). The OECD average NPMTR for the median worker is also within this range, at 34.1%.
The impact of measuring NPMTRs for the median rather than average worker will depend in part on the size of the differential between the median and average wage, in part on the progressivity of the personal tax system, and in part, due to the placement of both wage levels within the relevant bands. For example, the lower level of the median relative to average wage may mean that a lower tax bracket applies under the personal income tax system or may mean there is a change to the abatement rate of a credit or deduction.
The OECD average NPMTR in 2017 was 1.4 percentage points lower for the median worker relative to the average worker (34.1% compared to 35.5%). However, it is notable that in the case of a large number of OECD countries the NPMTRs remain the same at both these wage levels (e.g., the NPMTR on the average wage is the same as the NPMTR on the median wage in 25 OECD countries). Relatively small changes (less than four percentage points) between the NPMTR for the median and the average wage earners are observed in Canada  the only country where the NPMTR is higher on the median wage than on the average wage  as well as in Germany, Spain, Korea and Mexico. Concerning Canada, the NPMTR is higher for the worker at the median wage, who contributes to the unemployment insurance since their contributions have not reached their maximum amount. In contrast, at the average wage level, the maximum unemployment contributions have been reached and no unemployment contribution is paid on the additional dollar. Decreases of greater than four percentage points are observed in the remaining six countries (France, United States, Luxembourg, Turkey, Austria and Switzerland); in each case due to a decrease in the applicable marginal tax rate of personal income taxes.
Figure 2.3, in 2017 the NPATR for the single worker in OECD countries at the median wage varied from 38.5% in Belgium to 7% in Chile, with an average NPATR of 23.4%. At the median wage, NPATRs for the single worker are concentrated between 20% and 25% of gross wages (this is the case for 16 countries).
Given that average tax rates take into account all personal income taxes and employee SSCs paid, as well as cash benefits, in a progressive tax system they will be affected by a change in wage level even if the tax settings applied (e.g., the applicable tax bracket) do not change. For example, the average wage and median wage earner might be placed within the same tax bracket but with the average wage at the higher end and the median wage at the lower end; resulting in a decrease in the average tax rate but no change in the marginal tax rate.
As median wages are lower than average wages in all OECD countries, and most OECD countries tax labour income progressively, NPATRs are lower for all but two OECD countries for the median worker relative to the average worker, as shown in the lower panel of Figure 2.3. No change is observed between the NPATR for the median and average worker in Chile and Hungary, as both countries apply flat taxation at these levels of labour income.
The distribution of NPATRs on the median wage across countries also differs from the distribution of NPATRs on the average wage, although the distribution is unchanged at the highest and lowest ends of the range: the top six countries and the bottom nine countries are the same for both levels of wage earnings. The average reduction in NPATRs across the OECD is from 25.6% for the average worker to 23.4% for the median worker, a decrease of 2.2 percentage points.
Tax wedge
The tax wedge takes into account personal income taxes, employee and employer SSCs and net cash benefits, measured as a percentage of total labour costs (i.e., gross wages plus employer SSCs). The key difference between net personal tax rates and the tax wedge is, therefore, the inclusion of employer SSCs in both the numerator and the denominator. The size of the employer SSCs relative to both NPATRs and gross wage determines the difference between the NPATR and tax wedge of each country.
Figure 2.4 shows the tax wedge for both the median and average workers in OECD countries, showing the marginal tax wedge in the upper panel and the average tax wedge in the lower panel. For the median worker, the marginal tax wedge in 2017 ranged from 66.4% in Belgium to 7% in Chile, with an OECD average of 43.5%; median workers in sixteen OECD countries had tax wedges of between 40% and 50%. Similarly, the average tax wedge on median workers in 2017 ranged from 52.0% in Belgium to 7% in Chile, with an OECD average of 34.3%. The median worker in 21 OECD countries faced a tax wedge of between 30% and 45%.
When the tax wedge on the median worker is compared to the tax wedge on the average worker, differences will result from the impact of the different levels of wage income in the progressive nature of each individual tax system. Consequently, as the difference between the net personal tax rate indicator and the tax wedge are driven by the inclusion of employer SSCs in the latter, changes in the tax wedge at the median rather than the average wage will follow those observed in the NPATR analysis, except where the change in wage level affects the amount of employer SSCs paid.
In almost all OECD countries in 2017, there was no difference between the amount of employer SSCs paid at the average versus the median wage. This is shown in Figure 2.5, which decomposes the change in the average tax wedge. For the OECD average tax wedge in 2017, 94.0% of the difference between the tax wedge at the median and the average wage is derived from personal income taxes, with most of the remainder being derived from employer SSCs. However, employer SSCs change between the median and the average wage in only six countries and the level of this change is small, with the only change of greater than one percentage point seen in Turkey (3.4 percentage points) and changes of less than 0.7 percentage points seen in all other countries.
The small changes in employer SSCs as a result of a switch from the average to the median wage mean that the differences observed between the average and the median wage in relation to net personal tax rates are also seen in relation to the tax wedge at both income levels. When calculated for median compared to average workers, the scale and direction of change in the marginal tax wedge for all countries are the same as for the NPMTR in each country. In 2017, the OECD average marginal tax wedge for the median worker is 1.2 percentage points lower than for the average worker (44.7% compared to 43.5%), as shown in Figure 2.4 (upper panel).
Figure 2.4 also shows the average tax wedge for median relative to average earners, with the intracountry differences in the tax wedge for these two wage levels also being relatively consistent with the changes observed in the NPATRs, although the overall country positions differ depending on the underlying level of SSCs in these countries. The OECD average tax wedge for the median worker in 2017 was 2.0 percentage points lower than for the average worker (34.3% versus 36.2%, with the residual due to rounding).
Table 2.2 sets out the key indicators for median and average wage earners in 2017.
Tax burden indicators for lowincome workers: contrasting 67% of median and average wages, 2017
A second household type for which Taxing Wages calculates tax burden indicators are lowincome earners, represented by the single individual earning 67% of the average wage. This section calculates the tax burden indicators for workers at 67% of the median wage, indicating the net personal average tax rates and tax wedges that apply at this level of earnings, and contrasting it to the worker earning 67% of the average wage.
On average across the OECD, 67% of the median wage in 2017 was 54.1% of the average wage. Across all OECD countries in 2017, 67% of the median wage ranged from 59.6% of the average wage in Canada to 42.4% of the average wage in Turkey. Table 2.1 above, shows the level of 67% of both median and average wages in all OECD countries in PPPadjusted USD.
Given that the median wage is more representative of the distribution of wages in OECD countries, and provides a more consistent basis for comparison across the distribution of each country, the worker at 67% of median wage is also more representative of lowincome earners across the OECD than the similar household type measured on the basis of the average wage. Contrasting the level of minimum wages in OECD countries with the level of 67% of median wage earnings, both measured as a percentage of average wage Figure 2.6, demonstrates that 67% of median earnings is closer to the level of minimum wage observed in all OECD countries in which a minimum wage applies, although in the case of Chile and Turkey, this earnings level falls below the minimum wage.
Net personal tax rates at 67% of the median and average wage
Net personal marginal tax rates on workers at 67% of the median wage ranged from 73.5% in Belgium to 7.0% in Chile in 2017, with an OECD average of 31.7%. The distribution of rates is more concentrated at the lower end of the spectrum, with 16 OECD countries having NPMTRs at this wage level between 25% and 35%. The two outliers are Belgium and France which have comparatively high NPMTRs at over 65% of gross wage earnings, more than 20 percentage points higher than the third highest NPMTR. The high marginal tax rates in Belgium and France are due to the lack of a reduction of employee SSCs in Belgium at 67% of median wage earnings and decreasing inwork benefits for the 67% median wage earner in France. These are two examples of a wellknown issue: the targeting of in work tax credits and similar provisions at low wages results in spikes in marginal tax rates. The same phenomenon could occur in other countries at other wage levels, due the wide use of in work tax credits. The results for Belgium and France should therefore be interpreted as examples of a more general issue.
Partially for these reasons, Belgium and France are also two of the four countries in which a single earner at 67% of the median wage faces a higher NPMTR than a worker at 67% of the average wage; the other two countries being Spain, which also has comparatively high NPMTRs, and the Slovak Republic. The high NPMTRs for the median worker in these countries are primarily related to SSCs or a reduction in allowances:

in Belgium, the SSC reduction is not applied at 67% of the AW;

in France, no inwork benefit is paid at 67% of the AW;

in Spain the withdrawal effect of the tax allowance against increasing earnings is captured in the marginal rate at 67% of median wage;

in the Slovak Republic where a health insurance allowance reduces SSCs at 67% of the median wage but not at 67% of the AW, and the withdrawal effect of the SSC allowance against increasing earnings is captured in the marginal rate for 67% of the median wage.
There was no difference between the NPMTRs at 67% of the median and the average wage in 21 countries and the remaining 11 countries had lower NPMTRs for the worker at 67% of the median wage relative to the average wage. The largest reductions in NPMTRs for the worker at 67% of the median, compared to 67% of the average, wage were observed in Portugal, at 28.5 percentage points, and Korea at 10.4 percentage points. In all other countries the reduction in NPMTR for a worker at 67% of the median wage relative to the average wage was less than eight percentage points. The OECD average NPMTR for the wage earner at 67% of the median wage was 3.8 percentage points lower than for the worker at 67% of the average wage (31.7% compared to 35.5%).
Tax wedge at 67% of the median and average wage
When the total amount of tax paid at the employee level is considered, the picture is quite different. The NPATR was lower or the same in all countries for the wage earner at 67% of the median wage relative to the wage earner at 67% of the average wage. Net personal average tax rates at 67% of median wage earnings ranged from 32.3% in Hungary to 1.7% in Mexico, with an OECD average of 18.6%. The average NPATR on the earner at 67% of the median wage was 3.2 percentage points lower than for an earner at 67% of the average wage (21.3%). Across the OECD, no difference was observed in two countries: in Hungary, which also had the highest NPATR on both levels of income earner; and in Chile, which had the third lowest NPATR. The largest difference in the NPATR between 67% of the median and average wage was in Portugal, which was driven by the fact that, unlike at 67% of the AW, no personal income tax is paid at 67% of the median wage.
By including the employer SSCs to calculate the tax wedge, Figure 1.8 shows that the difference for marginal and average tax wedges between these two income levels is generally similar, for each country, to the difference between NPMTRs and NPATRs, although the magnitude of the difference in the tax wedge indicator changes depending on the size of employer SSCs in each jurisdiction.
As for NPMTRs, in most countries the marginal tax wedge is the same for median and average earners at 67% of each wage level. The NPMTR on the worker at 67% of the median wage is higher than the NPMTR at 67% of the average wage in four countries (Belgium, France, Slovak Republic and Spain, as discussed earlier) and lower in 11 countries, with the most pronounced decrease seen again in Portugal, as discussed above. The difference between the marginal tax wedge between 67% of the median and 67% of the average wage in the Slovak Republic is due to the health insurance allowance that reduces employee and employer SSCs at 67% of the median wage but does not apply at 67% of the AW .
The average tax wedge on the single worker at 67% of the median wage is highest in Hungary (46.2% of gross wage earnings) and lowest in Chile (7.0%) with an OECD average of 29.8%. The OECD average tax wedge on a worker at 67% of the median wage is 3.2 percentage points lower than the tax wedge applying at 67% of the average wage (32.4%). The strongest reductions in the tax wedge between 67% of the median and 67% of the average wage are seen in Portugal (where no personal income tax applies at 67% of the median wage), France (where the inwork benefit is paid at 67% of the median wage) and Turkey (where a tax credit is paid on earnings below the minimum wage, which is the case at 67% of median wage).
Table 2.3 sets out the key indicators for median and average wage earners at 67% of each wage level in 2017.
Comparing the composition of the average tax wedge for 100% and 67% of median and average wage earnings
Differences in the labour taxation of median and average wage earnings arise due to the interaction of lower wage levels at median earnings with the progressive nature of most labour taxation systems in the OECD. This progressivity is primarily seen in the personal income tax system and in the availability of deductions and allowances for personal income taxpayers. Both employer and employee SSCs, are less progressive and contribute less to the change in net personal tax rates and tax wedges as the wage level decreases.
Figure 2.9 shows the decomposition of the OECD average tax wedge for each of the four income levels shown in this Special Feature. Within both pairs of income levels (100% of the median and the average wage; and 67% of each), personal income taxes change most when median rather than average wage levels are considered. This is due to the progressive tax schedules that apply in almost all OECD countries and due to deductions and allowances that apply up to a maximum level of income. The amount of personal income taxes, measured as a percentage of total labour costs, increases roughly two percentage points on average between 67% of the median wage and 67% of the average wage and just slightly less than two percentage points between 100% of the median wage and the average wage. Personal income taxes also increase in almost every OECD country when moving from the median wage to the comparable level of average wage.
Employer SSCs also increase, on average, when moving from the median wage to the comparable level of the average wage. When moving between 67% of both wage levels, employer SSCs increase by 0.4 percentage points and the increase is smaller between the median and average wage, at 0.1 percentage points. The small size of this increase reflects the fact that across OECD countries, differences in employer SSCs are seen in only nine OECD countries between 67% of each wage level and in only six countries between 100% of each wage level.
Employee SSCs are the same, on average, for the worker at 67% of the average wage and the wage earner at both the median and the average wage, and slightly lower for the worker at 67% of median wage, the lowest income level of the group. This again reflects that differences in employee SSCs, even at 67% of both wage levels, are only seen in six OECD countries. Finally, the impact of cash benefits is very limited for these family types, showing a reduction on average of 0.1% for the worker at 67% of the median wage and no reduction for the other family types, but would be more significant if compared against the lowincome earner with children where cash benefits are more prominent (as shown in the special feature in the 2018 edition of Taxing Wages).
Conclusions
Compared to the average wage, the median wage can be considered a better measure of the impact of taxation across the distribution of wage earnings, because it is less influenced by outliers. This is particularly important in the presence of increasing inequality in most OECD countries. The median wage also permits a comparison of labour taxation across OECD countries at a consistent point in the wage distribution of each country, which is not the case when using the average wage as a reference point. However, the main drawback in using the median wage is that calculating the median wage is more difficult than calculating the average wage. Due to a lack of comparable data, a number of estimations have had to be used in this Special Feature to ensure that the median wage is both timely and consistent with the methodology for the average wage used in the rest of the Report.
Median wages are consistently lower than average wages in OECD countries, at 80.8% of the average wage in 2017 (i.e., the OECD average) and in 24 OECD countries, the median wage is closer to 67% than to 100% of the average wage. The interaction of the lower wage levels and the progressive labour taxation applied in OECD countries means that the average tax burden, calculated either as net personal tax rates or the tax wedge, is lower (or in rare cases the same) for the worker on the median wage relative to the worker on the average wage. This is the case in all OECD countries except Chile and Hungary, although the difference is not significant in most countries. The findings suggest that taxes at the mean wage are – for most countries – a helpful approximation of taxes at the median wage.
To assess the impact of labour taxation on lowincome workers, a reference point of 67% of median wage can also offer a better representation of lowincome workers than 67% of average wage. Sixtyseven percent of the median wage is closer to the minimum wage in almost all countries for which data is available and only below the level of minimum wage in two countries (Turkey and Portugal). Net personal average tax rates and the average tax wedge are, therefore, again lower in all OECD countries for the worker at 67% of the median wage. Marginal tax rates for the worker on 67% of the median wage are the same as for the worker on 67% of the average wage in 24 countries and lower in 11 countries. However, in four countries, the NPMTR and the marginal tax wedge are higher for the worker on 67% of the median wage compared to those on 67% of the average wage. These are examples of spikes in the marginal tax rate that result from the targeting of in work tax credits and other provisions at low wages. Similar results could appear in other countries at other wage levels and the issue is not specific to these four countries.
The lower tax levels applying to the median rather than average wage levels are primarily driven by personal income taxation, which is the most progressive of the major components of the labour tax system in most OECD countries. Employee SSCs play a negligible role and while employer SSCs are higher on average, these generally only have an impact in a handful of countries. Cash benefits do not play a large role in the tax rates on median earnings, even at 67% of median earnings, but would likely play a larger role if families with children were included in the analysis.
References
[1] EUROSTAT (2014), Structure of Earnings Survey (SES), https://ec.europa.eu/eurostat/web/microdata/structureofearningssurvey.
[2] OECD (2018), OECD Economic Outlook, Volume 2018 Issue 2, OECD Publishing, Paris, https://doi.org/10.1787/16097408.
[3] OECD (2017), OECD Employment Outlook 2017, OECD Publishing, Paris, https://doi.org/10.1787/19991266.
Notes
← 1. The average wage used in this publication is the mean of all wage incomes of fulltime employees (including manual and nonmanual) in sectors BN inclusive, as explained in the Annex.
← 2. The average worker corresponds to the single wage earner at 100% of average wage discussed in the rest of the Report.