Labour market concentration since the British Industrial Revolution

By Kirill Kushnarev

The UK Historical Data repository has been developed jointly by the Bank of England, ESCoE and the Office for National Statistics. The site is a repository of historical UK data and relevant statistical publications, as well as a hub that links to other data websites and sources.

ESCoE recently launched a master’s dissertation programme open to all UK master’s students writing dissertations on economic measurement. Students were challenged to use sources from the historic repository and write a blog linked to their project. The winning blog on labour market concentration since the British Industrial Revolution was written by Kirill Kushnarev, a former Economic History student at LSE who is now doing a pre-doc at Trinity College Dublin. Here he explains his findings and the use of historic repository sources.

Does share of labour income decline over time? Recent observations stress the decline of labour income, referred to as the labour share.[1] From the 1970s, labour share steadily reduced in most advanced economies, including the United Kingdom.[2] Many economists view the recent global decline as a negative phenomenon, emphasising the rising market concentration as “bad concentration,” characterised by a decline of workers’ compensation per-unit of output.[3] However, from the historical perspective of the United Kingdom, labour share has generally increased since the British Industrial Revolution, starting from around 60% in the mid-18th century and heading toward 70% in the 2000s, as Figure 1 shows.

Two cautionary comments should be made regarding this trend. First, given the different approaches to measuring labour share, I rely on publicly available calculations from the ESCoE historical data repository that use weighted factor shares, excluding housing.[4] This approach imposes certain assumptions on how compensation per employee is weighted, particularly the treatment of the self employed, but remains the only long-run publicly available measure.[5] Second, the co-movement of real wages and output starting in the 1850s was already famously framed by Robert Allen in Engels’s pause puzzle.[6] However, more research is needed to explore how and when technological change during the British Industrial Revolution contributed to the rise in the labour share, as well as whether there were centuries of equal growth prior to the well-documented decline in the 1970s.

For a long time, the stability of labour share has been considered a key macroeconomic variable of interest, early noted in Kaldor’s 1961 paper and widely applied in neoclassical models.[7] On one hand, recent literature extensively challenges this idea, particularly regarding the recent decline of labour share. Therefore, it is worth investigating both the decline and rise of labour share. On the other hand, inequality and technological literature provide insights that technological change and growth may cause a rise in market shares of top firms, referred to as market concentration, and some recent empirical papers show that market concentration could rise along with the labour share.[8] Combining both fields of literature, I ask a research question of how market concentration is related to the labour share from a historical perspective.

Why is this important in practice? At the core, changes in labour share capture the pace of business dynamism – defined as the entries, exits, and other population variations, such as the age, of firms – considering that declining labour share led to the rise of market concentration and a corresponding decline in business dynamism. There are two key examples of why high market concentration could be harmful. Employees may not receive the relative payoff in comparison with capital holders, leading to inequality between 1) workers, as those employed by more productive firms earn higher wages, and 2) capital holders, as firms could exercise market power in output markets by charging prices above marginal costs (markups). However, the British Industrial Revolution provided unique settings for investigating whether the market concentration may rise alongside an increasing labour share.

I start my analysis from the mid-18^th century for labour and capital shares, and from the mid-19^th century for market concentration.

Data

My dissertation uses ESCoE historical repository data to construct labour and capital costs.[9] I use labour share as ONS’s definition of the total compensation divided by Nominal GDP at factor cost, excluding rents on dwellings. In addition, I use the Feinstein tables provided by ESCoE[10] to construct capital cost measures, which are calculated as the product of the rate of return and the nominal value of the capital stock. For the rate of return, I use the yield on perpetual annuities/consols. And for the nominal value of the capital stock, I use the values of non-dwellings stocks.

Furthermore, for the market concentration analysis, I explore the Integrated Census Microdata (I-CeM) and the British Business Census of Entrepreneurs (BBCE) data between 1851 and 1911, provided by The Cambridge Group for the History of Population and Social Structure.[11] The individual-level source comes in two editions: 1) I-CeM, which covers a population of workers and entrepreneurs, and 2) BBCE, which focuses only on entrepreneurs. The datasets cover the universe of workers and entrepreneurs collected by census enumerators. My research uses data from entrepreneurs’ statistics, focusing on those with more than one employee on their accounts, excluding the self-employed. This is because, in practice, calculating concentration requires more than one value for the number of workers.

Summary

To emphasise my research question in economic history, I investigate the distribution of growth during the second phase of the British Industrial Revolution (1851-1991), which literature treats as relatively equal between workers and capital holders, known in the context of Engels’s pause.[12] I use the concept of market concentration as a measure of market power, specifically local labour market concentration. I use the Herfindahl-Hirschman index (HHI) restricted by occupations within each labour market to formalise the measure of concentration. For local concentration, I calculate the HHI for each occupation and then average these values within each local labour market. In all other cases, I aggregate HHI values at the occupation level. These concepts formalise the role of large businesses in determining business dynamism, as concentrated local labour markets tend to have lower wages and higher entry barriers. In my dissertation, I present three main results:

1. Market concentration doubled between 1851 and 1881. The rise in concentration coexists with a growing inequality between the top and bottom of business distribution

Measured as the Herfindahl-Hirschman index for 51 sectors and more than 200 labour markets, I have found that concentration grew from 35 to 75 points for thirty years, lower than the 31-36 points concentration for the US between 1992-2017.[13] I report these results in Figure 2, showing a relatively steady rise in market concentration. As this is a shocking result from a theoretical perspective, I use another specification of market concentration, the Concentration Ratio (CR) of the top 1% of the largest businesses. CR ratio shows the same trajectory, although accounting for a minor surge in concentration, from 52 to 53.75% of the share of the top 1% of businesses in the top 10%. Also, I justify an increasing inequality using a 90/10 ratio of employment growth, showing that inequality between the top and bottom 10% of firms increased from 12 to 18 times.

Are these findings consistent with the labour share dynamic? As I mentioned in the motivation, current literature suggests that the rise of market concentration coincides wiht the decline of labour share. My results show that the conclusion is invalid, as market concentration can rise when there is either a declining or rising labour share. Moreover, a decline or a rise in labour share tends to increase concentration via different channels. The booms in labour share trigger uneven growth, rooted in the uprising of market entrants and their less capital-intensive enterprises compared to incumbents. The new entrants spur incumbents’ growth. The busts in labour share are caused by redistribution via exits, accounting for different ages of new market entrants and incumbents. The new entrants more frequently leave, and incumbents grab their labour while raising fixed costs.

2. Despite the nationwide market concentration, local labour market concentration declines when labour share rises

The local labour market concentration also rose between 1851 and 1871 but fell in 1871 as labour share grew, as outlined in Figure 3. This dynamic suggests that different measures, such as nationwide and local labour concentration, have different statistical properties and do not capture labour share similarly. Figure 3: Local Labour Market Concentration, 1851-1881. 

However, should we reject or confirm the relationships between market concentration in labour share based on these calculations? I disentangle this question into two perspectives. From a practical perspective, they do not contradict each other, as local labour market concentration could give more precise estimations for the distribution of labour than nationwide concentration or top 1% firm ratios. Theoretically, using only certain specifications of market concentration while excluding others is challenging, as data is not always precise and unified. Because of that, I highlight business dynamism as a critical driver of changes in the labour share.

3. The business dynamism amplifies declining and rising labour share while discussing whether it could solve the abovementioned disparities between practical calculations and theoretical implications. I specifically explore three questions:

3.1. How often did businesses enter and exit the market? Business dynamism, measured by entry rate, fluctuated between 5% in 1851 and 15% in 1911. These calculations in Figure 4 are comparable with modern entry rates, reported as between 15% and 4%. However, the data shows an excessive spike in entries in 1881, heading toward 40%. This spike must be analysed cautiously due to differences in data construction. While other censuses were derived from the I-CeM dataset, the 1871 census was independently reconstructed from enumerators' books by the BBCE project. This could potentially lead to spurious results in identifying new businesses through linkages.[14] However, this inconsistency does not affect the firms’ characteristics and probably only generates excessive entrants’ results for the 1871 census. Exit dynamics also confirm such surprising numbers, indicating that the number of businesses that exited markets increased over time. Where does business dynamism come from? I argue that creative destruction causes bigger business dynamism, as creative destruction keeps both reallocation and entry rates high. To provide an argument for that, I use counterfactual analysis, hypothesising what happens if I increase creative distraction of incumbents, the largest firms in the local labour markets, to 1 standard deviation. On average, it gives a 50% increase in new entrants between 1851 and 1881. Business dynamism was significantly amplified when labour share began to rise, as shown by continuing excessive entries and the shifting of the distribution of companies toward larger businesses. Hence, the declining labour share coexists with lower business dynamism, as the opposite rise of labour share is supported by higher business dynamism. However, market concentration and inequality at the occupation level continue to rise even though the labour share has grown.

  3.2. Did the businesses become more stable over time? Businesses usually exit markets in their first ten years, and just a minority of companies stay in the same market for longer than that, as shown in Figure 5. However, there was a significant improvement in the maturity of businesses after 1871, as companies' average age grew from less than one year to more than five years in 1911.

3.3. Lastly, from which part of the distribution does employment growth originate? While it is a conventional statement from the literature that employment growth comes from small businesses, the net employment growth shifted from small to medium companies between 1851 and 1881. The growth shifted toward 20-29 workers businesses and significantly moved to 100-1000 companies, as seen in Figure 6. These results are consistent with the modern macroeconomic literature, which also mentions that most firms exit the market in their first years.[15]

 

Conclusion

To summarise my three core findings, I have found that the second phase of the British Industrial Revolution experienced a rise in market concentration nationwide and locally. Among the three measures of market concentration, only local market concentration confirms that a decline in labour share is associated with rising market concentration. In contrast, other market concentration specifications rise despite growing labour share. It makes sense in the theoretical framework when both growth and decline of labour share have an uneven distribution. Moreover, the rise of labour share started at a time of enormous creative distraction, and the growing labour share after 1871 coexisted with significant improvements in business dynamism, such as age, maturity, and number of new entrants.

Why do these findings matter?

The three implications of the core results reported in this blog are:

ESCoE blogs are published to further debate.  Any views expressed are solely those of the author(s) and so cannot be taken to represent those of the ESCoE, its partner institutions or the Office for National Statistics.