Is the Last Mile the Longest? Economic Gains from Gender Equality in Nordic Countries

Annex B.  Methods and data

This annex provides details of the methods and data used in the empirical exercises presented in Sections 3 and 4. It starts with a description of the methods and data used in the growth accounting exercises in Sections 3.1 and 3.2, followed by a similar description for the development accounting exercise in Section 3.3 and, lastly, an overview of the methods used for the forward-looking projections in Section 4.

Growth accounting estimates in Sections 3.1 and 3.2

The analyses presented in Sections 3.1 and 3.2 are based primarily on growth accounting – a procedure used to identify and decompose the sources of economic growth. Using macro-economic time-series, the basic principle of growth accounting is to split economic growth into its main components parts under standard macroeconomic theory: labour, capital, and total factor productivity (Solow, 1956[86]). Data permitting, the contribution of each of these main components can then be further decomposed into sub-components, as required.

Theoretical model

The starting point for the growth accounting exercise is a standard Cobb-Douglas production function with constant returns to scale. Total economic output, measured by Gross Domestic Product (GDP), can be expressed using the following multiplicative function:

$Y=K^{1-\alpha }.{\left(A.Q.E.\frac{H}{E}\right)}^{\alpha }$ $\left(1\right)$

where Y is GDP, K is physical capital, A is technological progress, Q is human capital per person employed, E is headcount employment, H is total hours worked (and therefore, H/E hours worked per person employed), and α is the labour share of output.

After some manipulations, this can be re-written as:

$Y={\left(\frac{K}{Y}\right)}^{(1-\alpha )/\alpha }.A.Q.E.\frac{H}{E}$ $\left(2\right)$

and output per person, measured by GDP per capita, can be expressed as:

$\frac{Y}{P}={\left(\frac{K}{Y}\right)}^{(1-\alpha )/\alpha }.A.Q.\frac{WP}{P}.\frac{E}{WP}.\frac{H}{E}$ $\left(3\right)$

where P stands for the population and WP for the working-age population, Y/P is GDP per capita, K/Y is the capital-to-output ratio, WP/P is the working-age share of the population, and E/WP is the employment-to-working-age-population ratio, or employment rate. Note that, in this particular model, employment is limited to those of working-age (15-64 year-olds) only. This is to allow for better modelling of how shifts in the working-age share of the population contribute to output.

A useful manipulation here is to split GDP per capita into the product of labour utilisation (covering the working-age share of the population, the employment-to-working-age-population ratio, and hours worked per person employed), and labour productivity or output per hour worked, covering all other factors (physical capital, technological progress, and human capital):

$\frac{Y}{P}=\left(\frac{Y}{E.H}\right).\left(\frac{WP}{P}.\frac{E}{WP}.\frac{H}{E}\right)$ $\left(4\right)$

where:

$\frac{Y}{E.H}={\left(\frac{K}{Y}\right)}^{(1-\alpha )/\alpha }.A.Q$ $\left(5\right)$

Labour utilisation can also itself be split into its component parts. The focus here is on disaggregating the employment rate (E/WP) and hours worked per person employed (H/E) by gender and, where possible, age group, as expressed by the following two additive functions:

$\frac{E}{WP} =\sum _{G\in \{<!--\; -->\left(15-24\right)men,\left(15-24\right),women, \left(25-54\right)men,\left(25-54\right)women,\left(55-64\right)men,\left(55-64\right)women\}}\frac{E_G}{WP}$ $\left(6\right)$

And

$\frac{H}{E}= \sum _{G\in \{<!--\; -->\left(15-24\right)men,\left(15-24\right),women, \left(25-54\right)men,\left(25-54\right)women,\left(55-64\right)men,\left(55-64\right)women\}}\frac{H_G}{E}$ $\left(7\right)$

Lastly, the model shown in equation 4 can be re-written in growth rates to give changes in economic output over time, as expressed by the following additive function:

$g_{\frac{Y}{P}}=g_{\left(\frac{Y}{E.H}\right)}+g_{\frac{WP}{P}}+g_{\frac{E}{WP}}+g_{\frac{H}{E}}$ $\left(8\right)$

where the g is the growth rate for the given factor, and with growth rates for the employment rate (g[E/WP]) and hours worked per person employed (g[H/E]) themselves calculated as

$g_{\frac{E}{WP}}=\sum _{G\in \{<!--\; -->\left(15-24\right)men,\left(15-24\right),women, \left(25-54\right)men,\left(25-54\right)women,\left(55-64\right)men,\left(55-64\right)women\}}{g}_{\frac{E_G}{WP}}$ $\left(9\right)$

and

$g_{\frac{H}{E}}=\sum _{G\in \{<!--\; -->\left(15-24\right)men,\left(15-24\right),women, \left(25-54\right)men,\left(25-54\right)women,\left(55-64\right)men,\left(55-64\right)women\}}{g}_{\frac{H_G}{E}}$ $\left(10\right)$

It is equations 8-10 that form the core of growth accounting exercise.

Data and estimation procedure

Estimation itself is conducted using the “Shapley decomposition” (Shorrocks, 2013[93]) – a procedure initially used in the inequalities literature to identify the contribution of different income sources to income inequality, but more recently also applied to other tasks such as, for instance, the decomposition of the drivers of poverty reduction (Azevedo et al., 2013[94]). Shapley decomposition runs through all possible sequences (effectively, combinations) of a given function in order to isolate the contribution of a given component – in this case, the contributions of growth in labour productivity (Y/(E.H)), growth in the working age share of the population (WP/P), growth in the employment rate (E/WP), and growth in average hours worked per person employed (H.E). It has the advantage of producing estimates that are additive (so that, in this case, growth in GDP per capita is estimated the sum of growth in the various components) and, at least when the number of components is fairly low, is also simple to calculate. A detailed overview of the Shapley decomposition method itself can be found in Shorrocks (2013[93]).

The estimates are produced in two stages. First, growth in GDP per capita is decomposed into each of its main components – as shown in equation 8 above, growth in labour productivity (Y/(E.H)), in the working age share of the population (WP/P), in the employment rate (E/WP), and in average hours worked per person employed (H.E). Second, the contributions of growth in both the employment-to-population rate (E/WP) and in average hours worked per person employed (H.E) are then themselves decomposed by gender and, where possible, by age group too, as shown in equations 9 and 10. It is these last two decompositions (of the employment rate and of average working hours) that are where the main interest lies – they provide information on the extent to which changes in men’s and especially women’s employment rates and working hours have contributed to economic growth. This procedure is run separately for each country.

The data used for the growth accounting exercise are based on a combination of official macroeconomic data from national accounts databases and employment and working hours estimates from labour force surveys. Data for the first stage are taken from the OECD National Accounts Database. The data series used include Gross Domestic Product (GDP), total population, total employment, and average working hours per person employed (see Table A B.1 for a summary). On occasion, important series are missing in the OECD database. Where this is the case, values are imputed or interpolated using information from alternative national accounts databases, such as the European Commission’s AMECO Database or those published by national statistical offices (see Table A B.2).

Table A B.1. Summary of data series used in Section 3

Series	Details	Main source	Additional sources
GDP (expenditure approach)	Constant prices, constant PPPs, OECD base year (USD 2010)	OECD National Accounts Database	Statistics Norway
Total population	National concept, 1000s	OECD National Accounts Database	EC AMECO Database
Total employment	Domestic concept, 1000s	OECD National Accounts Database	EC AMECO Database; Statistics Norway (SSB); OECD Employment Database
Hours worked for total employment	Domestic concept, hours, millions	OECD National Accounts Database	Statistics Norway (SSB); OECD Employment Database
Population by age and gender	1000s	OECD Employment Database	Eurostat; Statistics Canada; Statistics Iceland
Employment by age and gender	1000s	OECD Employment Database	OECD Annual Labour Force Statistics Database; National Economic Institute of Iceland
Average usual weekly working hours by age and gender	Total employment, hours	OECD Employment Database	Eurostat; Statistics Canada; Statistics Finland; Statistics Iceland; Statistics Norway; Statistics Sweden

Table A B.2. Country-specific notes for the data series used in Section 3

Series	Country	Details
GDP (expenditure approach)	Norway	All data refer to mainland GDP only (i.e. excluding oil and gas). Data are taken from Statistics Norway.
Total population	Sweden	Data missing in the OECD National Accounts Database for years before 1970. Data for years before 1970 are estimated by trending backwards using alternative data on total population from the EC’s AMECO Database.
Total employment	Iceland	Data fully missing in the OECD National Accounts Database for all years. Replaced with alternative data on total employment (domestic concept) from the EC’s AMECO Database.
	Norway	All data refer to mainland employment only (i.e. excluding oil and gas). Data are taken from Statistics Norway.
	Sweden	Data missing in the OECD National Accounts Database for years before 1980. Data for years before 1980 are estimated by trending backwards using alternative data on total employment (domestic concept) from the EC’s AMECO Database.
	United Kingdom	Data missing in the OECD National Accounts Database for years before 1980. Data for years before 1980 are estimated by trending backwards using alternative data on total employment (domestic concept) from the EC’s AMECO Database
Hours worked for total employment	Canada	Data missing in the OECD National Accounts Database for years before 1997. Data for years before 1997 are estimated by trending backwards using data on annual hours worked per worker from the OECD Employment Database.
	Finland	Data missing in the OECD National Accounts Database for years before 1975. Data for years before 1975 are estimated by trending backwards using data on annual hours worked per worker from the OECD Employment Database.
	Iceland	Data fully missing in the OECD National Accounts Database for all years. Replaced with estimates based on data on annual hours worked per worker from the OECD Employment Database.
	Japan	No data on hours worked for total employment in the OECD National Accounts Database
	Norway	All data refer to mainland employment only (i.e. excluding oil and gas). Data are taken from Statistics Norway.
	United Kingdom	Data missing in the OECD National Accounts Database for years before 1994. Data for years before 1994 are estimated by trending backwards using data on annual hours worked per worker from the OECD Employment Database.
	United States	Data missing in the OECD National Accounts Database for years before 1998. Data for years before 1998 are estimated by trending backwards using data on annual hours worked per worker from the OECD Employment Database
Population by age and gender	Canada	Data missing in the OECD Employment Database for years before 1976. Data for years before 1976 are estimated by trending backwards using alternative data on population by gender from Statistics Canada. Estimates produced for all ages and 15-64 year-olds only.
	Denmark	Data missing in the OECD Employment Database for years before 1983. Data for years before 1983 are estimated by trending backwards using alternative data on population by gender from Eurostat. Estimates produced for all ages and 15-64 year-olds only.
	Iceland	Data missing in the OECD Employment Database for years before 1991. Data for years before 1991 are estimated by trending backwards using alternative data on population by gender from Iceland.
	United Kingdom	Data missing in the OECD Employment Database for years before 1984. Data for years before 1984 are estimated by trending backwards using alternative data on population by age and gender from Eurostat. Estimates produced for all ages and 15-64 year-olds only.
Employment by age and gender	Canada	Data missing in the OECD Employment Database for years before 1976. Data for years before 1976 are estimated by trending backwards using alternative data on employment by gender from the OECD Annual Labour Force Statistics Database. Estimates produced for all ages and 15-64 year-olds only.
	Denmark	Data missing in the OECD Employment Database for years before 1976. Data for years before 1983 are estimated by trending backwards using alternative data on employment by gender from the OECD Annual Labour Force Statistics Database. Estimates produced for all ages and 15-64 year-olds only.
	France	Data missing in the OECD Employment Database for years before 1968. Data for years before 1968 are estimated by trending backwards using alternative data on employment by gender from the OECD Annual Labour Force Statistics Database. Estimates produced for all ages and 15-64 year-olds only.
	Iceland	Data missing in the OECD Employment Database for years before 1991. Data for years before 1991 are estimated by trending backwards using alternative data on labour force participation and unemployment from the Icelandic Census and national estimates of employment derived from accident insurance records and provided to the OECD by the National Economic Institute of Iceland.
	United Kingdom	Data missing in the OECD Employment Database for years before 1984. Data for years before 1984 are estimated by trending backwards using alternative data on employment by gender from the OECD Annual Labour Force Statistics Database. Estimates produced for all ages and 15-64 year-olds only.
Average usual weekly working hours by age and gender	Canada	Data fully missing in the OECD Employment Database. Replaced with alternative data on average usual weekly working hours by age and gender from Statistics Canada.
	Finland	Data missing in the OECD Employment Database for years before 2001. Data for years before 2001 are estimated by trending backwards using alternative data on working hours (average annual hours actually worked) provided by Statistics Finland. Estimates produced for all ages only.
	Iceland	Data missing in the OECD Employment Database for years before 2000. Data for years before 2000 are estimated by trending backwards using alternative data on working hours (average actual working hours) provided by Statistics Iceland. Estimates produced for all ages only.
	Italy	Data missing in the OECD Employment Database for years before 1995. Data for years before 1995 are estimated by trending backwards using alternative data on average usual weekly working hours from Eurostat. Estimates produced for all ages only.
	Norway	Data missing in the OECD Employment Database for years before 1996. Data for years before 1996 are estimated by trending backwards using alternative data on working hours (average actual weekly working hours) from Statistics Norway. Estimates produced for all ages only
	Sweden	Data missing in the OECD Employment Database for years before 2002. Data for years before 2002 are estimated by trending backwards using alternative data on average usual weekly working hours provided by Statistics Sweden. Estimates produced for all ages only
	United States	Data refer to average hours for dependent employees only.

The second stage of the growth accounting exercise requires data on population, employment, and working hours that are disaggregated by age and gender. Because national accounts databases do not typically disaggregate information by age or gender, these data are estimated using information from labour force surveys – in short, the overall national accounts series on population, employment and hours are split and ‘allocated’ across the various gender- and age groups according to the distribution of the given series provided by labour force survey data. As our interest is largely on the working-age population only – and also because national accounts series are not always precisely consistent with estimates from labour force surveys, for various coverage, definitional and methodological reasons – the labour force survey estimates for employment and working hours are re-scaled prior to this ‘allocation’, so that the aggregated estimates for the working-age population match exactly those from the national accounts series.

The labour force survey data used for this second stage are taken primarily from the OECD Employment Database or from data series supplied by the Nordic national statistical offices in response to an OECD questionnaire (see Table A B.1). Where data are missing, values are imputed using information from alternative sources such as the Eurostat Database or alternative information from national statistical offices, or are estimated by trending the nearest observation back or forward using alternative but similar series (see Table A B.2). Even so, the length of available series differs across countries (see Box 3.2).

Development accounting estimates in Section 3.3

The analysis presented in Section 3.3 is based on development accounting – a procedure similar to the growth accounting technique discussed above, but used to identify and decompose the sources of differences in economic output across economies, rather than growth within a single economy over time. The principle in this case is to use macroeconomic data to split differences in output into parts accounted for by differences in labour, in capital, and in total factor productivity. As with growth accounting, data permitting, the contributions of each of these main components can then be decomposed further into their own sub-components.

Theoretical model

The theoretical framework used for the development accounting exercise is analogous to that used for the growth accounting exercise above. Using the standard Cobb-Douglas production function outlined above and based on the re-arranged model shown in equation 4, relative differences in GDP per capita between two economies (A and B) at a given point in time can be expressed using the following multiplicative function:

$\frac{\frac{Y_A}{P_A}}{\frac{Y_B}{P_B}}=\frac{\left(\frac{Y_A}{E_A.H_A}\right)}{\left(\frac{Y_B}{E_B.H_B}\right)}.\frac{\left(\frac{{WP}_A}{P_A}\right)}{\left(\frac{{WP}_B}{P_B}\right)}.\frac{\left(\frac{E_A}{{WP}_A}\right)}{\left(\frac{E_B}{{WP}_B}\right)}.\frac{\left(\frac{H_A}{E_A}\right)}{\left(\frac{H_B}{E_B}\right)}$ $\left(11\right)$

where A is the economy of interest, and B is a benchmark economy chosen to provide a point of comparison. The choice of benchmark economy is largely arbitrary, though different choices can alter the ease of interpretation. In the results shown in Section 3.3, the OECD total is chosen as the benchmark ‘economy’. Results therefore show the relative gap in GDP per capita between a given economy and the OECD as a whole and the factors that contribute to the relative gap in GDP per capita between a given economy and the OECD as a whole.

Relative differences in the employment rate (E/WP) and hours worked per person employed (H/E) can then be further decomposed using the following two additive functions based on equations 5 and 6:

$\left(12\right)$

$\frac{\left(\frac{E_A}{{WP}_A}\right)}{\left(\frac{E_B}{{WP}_B}\right)}=\frac{\left(\frac{E_{{\left(15-24\right)men}_A}}{{WP}_A}\right)}{\left(\frac{E_{{\left(15-24\right)men}_B}}{{WP}_B}\right)}+\frac{\left(\frac{E_{{\left(15-24\right)women}_A}}{{WP}_A}\right)}{\left(\frac{E_{{\left(15-24\right)women}_B}}{{WP}_B}\right)}+\frac{\left(\frac{E_{{(25-54)men}_A}}{{WP}_A}\right)}{\left(\frac{E_{{(25-54)men}_B}}{{WP}_B}\right)}+\frac{\left(\frac{E_{{(25-54)women}_A}}{{WP}_A}\right)}{\left(\frac{E_{{(25-54)women}_B}}{{WP}_B}\right)}+\frac{\left(\frac{E_{{(55-64)men}_A}}{{WP}_A}\right)}{\left(\frac{E_{{(55-64)men}_B}}{{WP}_B}\right)}+\frac{\left(\frac{E_{{(55-64)women}_A}}{{WP}_A}\right)}{\left(\frac{E_{{(55-64)women}_B}}{{WP}_B}\right)}$

And

$\left(13\right)$

$\frac{\left(\frac{H}{{WP}_A}\right)}{\left(\frac{H_B}{E_B}\right)}=\frac{\left(\frac{H_{{(15-24)men}_A}}{E_A}\right)}{\left(\frac{H_{{(15-24)men}_B}}{E_B}\right)}+\frac{\left(\frac{H_{{(15-24)women}_A}}{E_A}\right)}{\left(\frac{H_{{(15-24)women}_B}}{E_B}\right)}+\frac{\left(\frac{H_{{(25-54)men}_A}}{E_A}\right)}{\left(\frac{H_{{(25-54)men}_B}}{E_B}\right)}+\frac{\left(\frac{H_{{(25-54)women}_A}}{E_A}\right)}{\left(\frac{H_{{(25-54)women}_B}}{E_B}\right)}+\frac{\left(\frac{H_{{(55-64)men}_A}}{E_A}\right)}{\left(\frac{H_{{(55-64)men}_B}}{E_B}\right)}+\frac{\left(\frac{H_{{(55-64)women}_A}}{E_A}\right)}{\left(\frac{H_{{(55-64)women}_B}}{E_B}\right)}$

These three equations (11-13) form the core of the development accounting exercise.

Data and estimation procedure

Estimation itself is again conducted using the Shapley decomposition (Shorrocks, 2013[93]). It is again run separately for each country, and once more takes place in two stages. First, the relative difference in GDP per capita between the given country and the OECD total is decomposed into each of its main components – as shown in equation 11 above, into the contributions of differences in labour productivity (Y/(E.H)), in the working age share of the population (WP/P), in the employment rate (E/WP), and in average hours worked per person employed (H.E). Second, the contributions of relative differences in both the employment-to-population rate (E/WP) and in average hours worked per person employed (H.E) are then themselves decomposed by gender and age-group, as shown in equations 12 and 13. Again, it is these last two decompositions that are of primary interest – they estimate the contribution of the relative level of men’s and women’s employment and working hours to any gap in GDP per capita between the given country and the OECD as a whole.

As with the growth accounting exercise, the data used are based on combination of official macro-economic data from national accounts databases and employment and working hours estimates from labour force surveys. More specifically, data for the first stage are taken from the OECD National Accounts Database (Table A B.1), while data for the second stage are estimated using information from labour force surveys, in the same way as above. The labour force survey data themselves are taken primarily from the OECD Employment Database (Table A B.1). Because the development accounting exercise is run just for the latest year available (2015), there are far fewer problems here with missing data. On the few occasions that data are missing, data points are filled using the same steps as outlined above (see Table A B.2).

It is worth noting here that complete national accounts data is currently available for only 30 OECD member countries. The exceptions are Chile, Iceland, Japan, Mexico and Turkey, all of whom are missing information on one or more key measures in the OECD National Accounts Database. As a result, the OECD ‘total’ used as the reference economy for the development accounting exercise refers to OECD-30 total (that is, the weighted total across the remaining 30 OECD member countries), only.

Forward-looking labour force and GDP per capita projections in Section 4

The forward looking projections presented in Section 4 are produced based on a combination of estimates from the OECD’s in-house labour force projection model – a dynamic age-cohort model that estimates labour participation up to the year 2060 – and a modified version of the OECD’s long-term growth models (as presented in OECD (2014[92]), OECD Economic Outlook No. 95). The projections look to model the impact of a range of different ‘gender gap’ scenarios, each of which assumes a different trajectory for gender differences in labour force participation and working hours:

• The baseline scenario, where labour force participation rates of men and women (15-74) are estimated using the OECD’s standard dynamic age-cohort model, which projects participation rates (by gender and five-year age-groups) based on current (2007-16) rates of labour market entry and exit, and average usual weekly working hours for each gender and five year age group are held constant at their 2016 values. This scenario services as our reference or baseline scenario.

• Scenario A: gender participation gaps reduced by 25% by 2025 and by 50% in 2040. In this scenario, male participation rates are held at the baseline and female participation rates are projected so that the gender participation gap observed in 2012 within each five-year age-group falls by 25% by 2025, and 50% by 2040. This scenario incorporates the G20 “25% by 2025” target. Average usual weekly working hours are held at the baseline.

• Scenario B: gender participation gaps reduced by 50% by 2025 and 100% by 2040. In this scenario, male participation rates are held at the baseline and female rates are projected so that the gender participation gap observed in 2012 within each five-year age group falls by 50% by 2025, and 100% (i.e. is fully closed) by 2040. Average usual weekly working hours are held at the baseline.

• Scenario C: gender participation gaps and gender working hours gaps reduced by 25% by 2025 and by 50% by 2040, with women increasing working hours. This scenario assumes that gender gaps in both labour force participation rates and usual weekly working hours decline for each five-year age-group by 25% by 2025 and 50% by 2040, with the gender working hours gap closed entirely through increases in female working hours (male hours follow the baseline).

• Scenario D: gender participation gaps and gender working hours gaps reduced by 25% by 2025 and by 50% by 2040, with men decreasing working hours. This scenario is the same as scenario C, but gender working hours gaps closed entirely through decreases in male hours, rather than increases in female hours. Female hours follow the baseline.

• Scenario E: gender participation gaps and gender working hours gaps reduced by 50% by 2025 and by 100% by 2040, with women increasing working hours. This is a “full convergence” scenario. It assumes that gender gaps in both labour force participation rates and usual weekly working hours decline for each five- year age-group by 50% by 2025 and 100% (i.e. are fully closed) by 2040, with the gender working hours gap closed entirely through increases in female working hours (male hours follow the baseline trend).

Production of the estimates themselves takes place in two stages. First, estimates of the size of the labour force and overall average working hours under each scenario are produced by applying the assumed labour participation rates and working hours to the OECD’s in-house labour force projection model. The model is a dynamic age-cohort model that, under baseline conditions, projects future labour participation by gender and five-year age-group using current rates of labour market entry and exit. For the various hypothetical ‘gender gap’ scenarios, participation rates and, where needed, working hours for the relevant gender and age-groups are forced so that they meet the given assumed gender gap targets by the given target year. Adjustment is assumed to occur linearly between the projection start year (2017) and the target year. The resulting estimates are then summed across both genders and all five-year age groups to produce estimates of the size of the overall labour force (15-74 year-olds – see below) and overall average working hours.

Second, estimates of GDP per capita and GDP per capita growth under each scenario are produced by combining the labour force and working hours estimates with a modified version of the long-term growth models presented by the OECD in OECD Economic Outlook No. 95 (see Johansson et al. (2013[89]) for technical details). The theoretical foundation for the long-term growth models is similar to that outlined for the growth accounting exercise above. The models estimate GDP based on a standard Cobb-Douglas production function, with the usual long-term growth determinants (i.e. labour, physical capital, human capital and total factor productivity). Potential GDP across the projection period (here, 2012 to 2040) is estimated by projecting trends and changes in the various input components, with projections of the components themselves based on both long-term dynamics within the given country and on convergence patterns between countries (see OECD (2014[92]) and Johansson et al. (2013[89]) for details on the measures, data and assumptions used to project the individual components).

Potential GDP per capita and GDP per capita growth under each scenario is estimated by adjusting projections from these long-term growth models according to the assumed change (relative to the baseline) in the overall labour force participation rate and the assumed change (relative to the baseline) in overall average usual weekly working hours. No change is assumed in the baseline scenario, so the estimates of GDP per capita in this scenario are identical to those in the OECD Economic Outlook No. 95. In each case changes and developments in all other production factors – such as physical capital and human capital and the remaining sub components of potential employment and labour efficiency – are held steady at the baseline.

It should be pointed out that the projections used in these scenarios are simply mechanical. In other words, they assume that any changes in labour force participation rates or weekly working hours do not interact with, or have any indirect effects on, other labour inputs or any other production factors, including physical or human capital. It is possible, for example, that changes in labour force participation rates and weekly working hours among, say, parenting-age women (25-54 year-olds) could lead to changes in participation and/or hours among older workers if, for instance, grandparents or older friends and relatives are used as substitute carers for children. If any such indirect effects occur, the impact of changes in patterns of paid work on the overall labour supply may differ from those estimated here. It should also be noted that, just as with the growth accounting estimates presented in Section 3, the projections do not factor in any possible effects of changes in patterns of paid work on household production. Again, to the extent that changes in male or female labour supply lead to changes in household production or to shifts between measured and unmeasured economic activity, the estimates shown here may not fully capture the effects of a change in patterns of paid work on economic output.

Lastly, as touched on at the start of Section 4, it is worth pointing out that the measures and units used for these forward-looking projections differ slightly from those used in Section 3. Specifically, while Section 3 concentrates on employment rates and a 15-64 year-old age group, here estimates are based on labour force participation rates (i.e. the employed plus unemployed population) and a 15-74 year-old age group. This is to help ensure compatibility with the inputs used for the OECD’s standard long-term growth models – which use labour participation as a core input, and the age-group 15-74 (OECD, 2014[92]) – and because the participation of over-65 workers is likely to increase in importance in future decades.