Productivity VI: Detailed Labour Productivity Statistics.

The Netherlands are well prepared for future challenges, ranking high on competitiveness, social progress, and human development. The Dutch population is highly educated, has high living standards, and values a good work-life balance, which puts further pressure on working hours. To pursue the countries high ambitions on living standards and work-life balance on the one hand, and housing, healthcare, education and the environment on the other hand, political and business leaders need to urgently define productivity improving policies to mitigate the impact of labour scarcity on the economy. In this article, strategies are discussed to identify candidates to enter in productivity improving programs.

1. Introduction

In previous articles, several labour productivity indicators were analysed by at the lowest level economic activity. This exercise concerned thousands of enterprises in the Netherlands, and every enterprise had, of course, its own contribution to value added, employed persons, full-time equivalents or hours worked. Consequently, every enterprise had its own labour productivity.

In the literature, there are numerous publications on the relationship between productivity indicators and company size, like e.g., Productivity in SMEs and large firms, in OECD Compendium of Productivity Indicators 2024, OECD Publishing, Paris. The OECD article states:

“…Productivity tends to increase with firm size, as large firms can benefit from increasing returns to scale. Firm-level productivity also depends on the industry enterprises are operating in. In addition, large firms tend to adopt new technologies more than small firms, unless the latter are new or younger companies…”

It was, therefore, a big surprise that we found in a previous contribution that (Belgian and Dutch) medium sized enterprises seemed more productive than large enterprises in the years 2021 and 2022. However, the latest data show that this unusual effect vanished in 2023.

Therefore, we take a deep dive in this article into the latest available detailed statistic data on labour productivity.

2. Labour Productivity by Company Size

In the article Productivity IV: Labour Productivity by Sector we demonstrated that (Belgian and Dutch) medium sized enterprises (from 50 to 249 employed persons) appeared to be more productive than large enterprises (250 employed persons or more) in the years 2021 and 2022. Yet, in the most recent data for 2023, this unusual phenomenon of “diseconomies of scale” vanished.

Figure VI.1 shows the result of analysing Statistical Netherlands data on small and medium-sized enterprises (SMEs) as part of the program Staat van het MKB (in Dutch only), which includes information on labour productivity by company size for 2023.

Figure VI.1 demonstrates that in 2023 the apparent labour productivity per full-time equivalent (ALP_fte) increases with company size. For companies with 500 employed persons or more there is still a small decrease in the apparent labour productivity per employed person (ALP_ep), as compared to one size class smaller (due to more people working part-time in those enterprises).

Therefore, the disturbing 2021 and 2022 results concerning “diseconomies of scale” were probably caused by the aftermath of the Covid-19 crisis. Possibly, larger enterprises needed more time to adapt to the new situation than smaller ones. The result also serves as a warning to use 2021 and 2022 statistical data with caution.

Figure VI.2 shows the contribution to gross value added (GVA) and full-time equivalents in the business economy (aggregate [B-N, excl. K, incl. 95]) by company size in 2023. The GVA of the business economy was 573,800 million euros, the number of employed persons was 7,380 thousand, which represented 5,778 thousand fte in 2023.

The solo self-employed persons (SSEs) delivered 11.5 percent of GVA and 21.3 percent of fte in the business economy. Medium sized companies (from 50 to 249 employed persons) delivered 23 percent of GVA and 18.3 percent of fte in the business economy. Large companies (250+ employed persons) delivered 38.7 percent of GVA and 27.8 percent of fte in the business economy.

Figure VI.1 confirms the low (average) relative apparent labour productivity of SSEs. At first sight, this seems to validate the suggestion from the article Productivity IV: Labour Productivity by Sector that the economy (i.e., gross value added) would fare well if part of SSEs (e.g. SSE – labour), (re)incorporated themselves in larger companies, as these are (on average) more productive.

However, there is an enormous spread in labour productivities among SSEs (and economic activities), just like there is a vast scatter of wages in and ?labour productivities among enterprises. Of course there are numerous SSEs with a higher labour productivity than underperforming large companies. Indeed, in paragraph IV.4.B.iv, we saw that there is quite some dispersion in hourly rates of SSEs.

The latter is not just true for (solo) self-employed persons, but for every company size class in any economic activity, as we shall see at continuation.

3. Company Size and Productivity Percentiles

The program Staat van het MKB (in Dutch only) supported by Statistical Netherlands on small and medium sized enterprises (SMEs) also includes information on labour productivity by company sizes and productivity percentiles for the years, 2010, 2020, and 2021.

The statistics on the business economy in this section is different from section 2, as neither the economic activity [L] Renting, buying, selling real estate are included, nor self-employed persons (and partnerships) in enterprise sizes.

This results in 203,265 enterprises being included in the study that contributed €359,790 million to gross value added (GVA) and employed 4,072 thousand full-time equivalents. Therefore, the average apparent labour productivity (ALP_fte) of the business economy was 88.4 k€ in 2021.

Figure VI.3 shows the percentiles of the apparent labour productivity (ALP_fte) by company size for the business economy, which is aggregate [B-N, excl. K+L, incl. 95], in 2021. The company size classes (ep=employed persons, number of enterprises in brackets) are defined as micro (1-9 ep, 132 265), small (10-49 ep, 37 120), medium (50-249 ep, 7 975) and large (250+ ep, 1 555) enterprises.

The percentile p10 indicates that 10 percent of enterprises in the population had a lower productivity than p10 (and 90 percent higher). Figure VI.3 demonstrates that there can be huge differences between p10 and p90. In the case of micro enterprises (1-9 ep) it is even more than a factor 10. Of course, this was to be expected as the enterprise population of the business economy includes many economic activities.

The average apparent labour productivity of the business economy (aggregate [B-N+S95_X_K+L]) was 88.4 k€ per full-time equivalent in 2021. This is lower than the average apparent labour productivity of entire Dutch economy ([A-U] All economic activities), which was 100.5 k€ per full-time equivalent in 2021, as the two very labour productive economic activities [K} and [L] were left out.

Yet, for simplicity, we will calculate relative apparent labour productivities per fte in this section against the average of the business economy and not against the total economy. Using this definition, figure VI.3 shows the levels RALPfte = 100%, RALPfte = 75% and RALPfte = 50%, which we will use to classify the performance of enterprises using the naming convention given in table VI.1.

Figure VI.3 demonstrates that about 70 percent of micro, 65 percent of small, 60 percent of medium, and 55 percent of large enterprises in the business economy underperformed, i.e., RALP_fte < 100%, in 2021. Similarly, it shows that about 55 percent of micro, 45 percent of small, and 35 percent of both medium and large enterprises in the business economy could be considered very labour intensive (i.e., RALP_fte < 75%) in 2021.

Finally, about 35 percent of micro, 20 percent of small, 15 percent of medium, and 20 percent of large enterprises in the business economy could be considered extremely labour intensive, i.e., RALP_fte < 50%, in 2021. In appendix VI.6.B, we discuss an estimated (cumulative) productivity distribution derived from figure VI.3 that helps us to better understand the economic dynamics behind the data.

Figure VI.4 shows the relative apparent labour productivity per fte by company size and productivity percentile classes p0-p10, p10-p25, p25-p50, p50-p75, p75-p90 and p90-p99. The size class “Total” refers to all company sizes, while we will use p0-p99 (which includes 99 percent of enterprises) as the total of all productivity classes.

The RALP_fte by company size of p0-p99 in figure VI.4 (black bars) shows the earlier mentioned maximum at medium sized enterprises. This (temporary) phenomenon was discussed in the previous section. The scatter of the RALP_fte within the distinct company size classes are huge.

Figure VI.5 shows the contribution of companies with different sizes and productivity percentiles to gross value added (GVA) and full-time equivalents (fte) in the business economy in the year 2021.

Figure VI.5 demonstrates that large companies (250+ ep) contributed to 42.9 percent of GVA in 2021, with 38.9 percent of fte. In contrast, the contribution of productivity percentile p90-p99 (all sizes) to GVA was 22.3 percent, with just 7.9 percent of fte. This small share (9 percent) of enterprises (all sizes) outperformed the rest with an average labour productivity per fte (RALP_fte) of no less than 284.2% (see figure VI.4).

What stands out from figure VI.5 is that the productivity percentile classes p50-p75, p75-p90, and p90-p99 have similar contributions to gross value added (GVA) for all enterprise sizes. However, the largest contribution comes from p50-p75 enterprises, and, in particular, from large enterprises.

4. Potential Targets for Productivity Improving Policies

The information from figures VI.3, VI.4 and VI.5 could help us to determine the impact of productivity improving policies on the economy (GVA) and the labour force (fte).

Figure VI.6 shows the share of gross value added and full-time equivalents by productivity percentile class in the Business Economy (aggregate [B-N, excl. K+L, incl. 95]) in 2021. Figure VI.6 ?confirms that the productivity percentile classes p50-p75, p75-p90, and p90-p99 of all enterprises together have similar contributions to gross value added (GVA). Indeed, the largest contribution comes from p50-p75 enterprises.

In summary, about half (49 percent) of the enterprises (all sizes, p50-p99) is responsible for 70.4 percent of GVA with about half (49.7 percent) of fte. This implies that the average labour productivity per fte (RALP_fte) of p50-p99 was 141.5 percent. This is far more than the average labour productivity per fte of large companies (p0-p99), being 110.1 percent, as figure VI.4 demonstrates.

In Productivity V: the Economic Value of Labour Scarcity we saw that the economic value of labour scarcity since 2021, presently estimated in about 400 thousand unfilled vacancies, was about 3 to 5 percent of the economy. Figure VI.6 shows that the productivity percentile class p0-p10 contributes about 3.6 percent to GVA with 10.5 percent of fte.

This implies that the productivity percentile class p0-p10 employs about 400 thousand fte and has little impact on the economy (GVA). This makes it a good candidate to further scrutinize as a candidate to target for productivity improving policies. These policies should help those enterprises to implement productivity improving measures, and educate and release resources that can be deployed in more productive economic activities that suffer from labour scarcity.

Figure VI.4 demonstrates that the average relative apparent labour productivity of p0-p10 was 34.5 percent, and is, therefore, considered extremely labour intensive (RALP_fte < 50%). This means that we could alternatively scrutinize for extremely labour intensive enterprises as potential targets for productivity improving policies, which comprised, according to figure VI.3, about 35 percent of micro, 20 percent of small, 15 percent of medium, and 20 percent of large enterprises in the business economy in 2021.

In appendix A (table VI.2) we give the detailed results for the business economy showing that the extremely labour intense enterprises (red marked) represent 42,925 enterprises (21.1 percent), a gross value added of €8,272 million (2.3 percent of GVA), and 388 thousand full-time equivalents (9.5 percent of fte) of the business economy.

This shows that the condition RALP_fte < 50% would also yield about the right amount of resources to reallocate, while the risk to the economy would be limited. However, it would target 21.1 percent, versus 10.1 percent in p0-p10 (all sizes), of enterprises in the business economy, being 25.1 percent of micro, 10.0 percent of small, 10.1 percent of medium, and no large enterprises.

Many, but not all, of those small and medium sized enterprises (SMEs) will operate in very labour intensive economic activities in the business economy. The very labour intensive economic activities identified in section IV.2.C were: [F] Construction, [N] Renting and other business support, [I] Accommodation and food serving, and [S95] Repair of computers and personal and household goods.

As these economic activities already have a low relative labour productivity, there will be more enterprises operating in them that have very low performance.

Figure VI.7 shows the labour productivity of the productivity percentiles p25 (lower quartile), p50 (median), and p75 (upper quartile) of small and medium sized enterprises in the business economy by economic activity in 2021, as well as the number of enterprises (right axis, logarithmic scale). The straight line is RALP_fte = 50%. The average labour productivity of an economic activity usually lies (due to top performers) somewhere between p50 and p75.

Figure VI.7 demonstrates that the lower quartile (p25) labour productivity values dance around the extremely labour intensive limit RALPfte = 50% line for most economic activities. There are a few economic activities of which the median (p50) values are below the line, being [I] Accommodation and food serving, ?[I55] Accommodation, [I56] Food and beverage service activities, and [N79] Travel agency, tour operator and other reservation service and related activities.

An important labour intensive economic activity, according to section IV.2.C, that is missing from the list is [F] Construction, which is due to the fact that solo self-employed persons (SSEs) are not included in this analysis. In tables IV.5 and IV.8 it was already shown that when only employees are considered, the sector [F] Construction is actually performing above average.

Of course, targeting underperforming enterprises with programs to help them become more productive and release resources necessary in economic activities with higher productivity would not be enough, as we saw in the article Productivity I: Business Clusters and Scarcity. It is also necessary to define policies for highly productive large firms and help them unblock their full potential (like, for instance, the lack of space and energy for the company ASML, and address the housing challenge for its employees).

Yet, we will use conditions like RALP_fte < 50% in next studies to define productivity improving policies for enterprises and include the (very or extremely) labour intensive economic activities not included in the business economy, like: [Q] Health and social work activities, [R] Culture, sports and recreation, [S] Other service activities (except [S95]), and [T] Activities of households.

So far, we studied labour productivity by economic activities at the lowest level ([A] .. [U]), and we considered enterprises at the highest level by company size and productivity percentile classes, as well as economic activities at level 2. In chapter VII, we will also screen economic activities at higher levels (level 2, 3 or 4) to identify the most likely targets for productivity improving policies.

5. Conclusion

In this article we found that the scatter of labour productivity data for enterprises within a certain economic activity are considerable. The ratio between the upper and lower labour productivity quartiles (P75/p25) is on average 2.6 and ranges from 1.6 ([M75] Veterinary activities) to 9.4 ([M72] Scientific research and development) in the encompassing economic activities.

This demonstrates that the variation within a certain economic activity is almost as big as the one between distinct economic activities. In table IV.15, we found that the ratio between the highest and lowest relative apparent labour activity of all economic activities (level 1) was 29.6 in 2023, and ranged from 41 percent ([T] Activities of households) to 1199 percent ([B] Mining and quarrying).

Also company size matters. Larger companies tend to have a higher labour productivity than smaller companies. In figure VI.1, we found that in aggregate [B-N, excl. K, incl. 95], very large enterprises (500+ employed persons) had an apparent labour productivity per full-time equivalent of 2.6 times the one of self-employed persons in 2023.

Altogether, this means that when targeting enterprises for productivity improving policies we will have to prioritize searching small and medium sized enterprises (SMEs) within very or extremely labour intensive economic activities. This strategy would have the lowest impact on the economy (less than 5 percent), while having the highest impact on the possibility to reallocate resources in more productive economic activities (at least the number of unfilled vacancies, i.e., presently about 400 thousand).

The main objective of productivity improving policies should be to help labour intensive enterprises to become more productive. This would enable them to release part of their personnel who can be educated to be employed in more productive economic activities.

It was found that the easiest way to target enterprises for entering productivity improving programs is to search within extremely labour intensive economic activities (RALP_fte < 50%) at the highest level possible (i.e., level 2, 3 or 4).

In the next article, we will do this by scrutinizing the Eurostat business statistics database for candidates.

6. Appendices

A. Detailed Productivity Percentile Results for the Business Economy in 2021.

Table VI.2 contains the content of the information for the Business Economy in 2021 offered by Statistical Netherlands for the program Staat van het MKB (in Dutch only).

Table VI.2: Detailed results of (relative) apparent labour productivity by company size and productivity percentile classes; very labour intensive classes (RALP_fte < 75%) are marked gold, extremely labour intensive classes (RALP_fte < 50%) are marked red; (Source: Statistics Netherlands – “Staat van het MKB”).

Very labour intensive (RALP_fte < 75%) classes are marked gold and extremely intensive classes (RALP_fte < 50%) are marked red in table VI.2. The sum of all extremely labour intensive company size and productivity percentile classes is given in de last row (see section VI.4 for details).

The condition RALP_fte < 50% comprises a similar amount of full-time equivalents (about 400 thousand) and share of the economy (GVA) as the productivity percentile class p0-p10 (all sizes), but contains more enterprises (20 percent versus 10 percent of all enterprises) in the business economy, and, in particular, more micro enterprises and SMEs. In fact, it would not include large enterprises.

B. Estimated Labour Productivity Distributions.

Figure VI.8 shows the result of fitting the data of figure VI.3 for the cumulative distribution of the labour productivity (x) with a two parameter model using the incomplete Gamma function:

The labour productivity distribution is defined by parameters α?and β?is equal to:

Here Γ(α)?is the Gamma function.

The average labour productivity <x>?of the distribution P(x; α, β)?is equal to: <x> = αβ.

The parameters α?and β?are found by a non-linear optimization scheme on the least squares of the difference in measured and estimated probabilities, and given in table VI.1. The fits to the cumulative distributions are reasonable, except for the largest labour productivities.

In fact, the tails should have been even more pronounced to fit the higher productivities better. However, introducing more parameters with just ten data points yields unrealistic fits and causes high predicted values of the average labour productivities. Unfortunately, more detailed data are not available to the general public at Statistics Netherlands due to privacy regulations.

We will not further use these estimations for any purposes, but, the estimated distributions shown in figure VI.6 are very illustrative. All distributions ?possess a long tail, which is due to a few very productive enterprises, that exist in any company size class, although more pronounced in large companies.

Table VI.1 demonstrates that, in most cases, the calculated average labour productivity is underestimated as compared to the ALPfte determined from the value added and full-time equivalents per company size class in the whole population of p0-p99. This means that the high productivity tail is in fact bigger than estimated in figure VI.8. This high productivity tail exists for any company size.