The Netherlands’ Minimum Wage 1969–2022: Can We Learn from Decline?

4.1 Two Measures of Minimum-Wage Employment

Employment effects are the focus of most minimum-wage research. Due to its complex features, the S-MW implies such a varied wage concept that it cannot be used as a clear and uniquely defined driving factor in an economic analysis of employment effects, be it macroeconomic or microeconomic. The statistic of S-MW employment is basically a legal statistic, not an economic observation of low wages. This motivates the use here of taking together all wages up to the level of the A-MW as a single uniform concept of measuring what has happened to employment given the drastic changes in the minimum-wage level. While S-MW counts youth earnings at the specific Y-MWs only, A-MW includes all earnings above the Y-MWs up to the A-MW. This inclusion of all lower wages does away with measurement error below the A-MW and ignores enforcement problems.^[22]

The application of the A-MW still faces a complicated data situation of four separate distributions of wages that have been maintained by CBS during the life of the minimum wage since it was fully developed including the tail of youth minimum wages: 1974 to 1984, 1984 to 2005, 1991 to 2014 and 2006 to 2021 (unfortunately, missing 2022) (see Table A.1). The distributions differ by type of employees and concept of wage, and the first series also misses out on part of the economy, especially the public sector. The unit of observation makes a notable difference to the measurement due to the prevalence of part-time jobs: 5.1 % of jobs versus 4.1 % of hours worked for S-MW employment in 2021, and 11.9 % of jobs versus 5.9 % of hours, respectively, for A-MW. Jobs may shine a light on employers’ hiring efforts, but hours worked are arguably the better measure for an economic analysis of employment effects (Card and Krueger 1995, 185). Much of the minimum-wage literature observes persons but has difficulty answering the question what happens to their working hours. Hours also improve the consistency of the data over time for the purpose of this contribution and are insensitive to the significant shift towards part-time jobs especially for minimum-wage employment (Salverda 2010).

Nonetheless, A-MW employment shows a surprisingly consistent picture in spite of these data breaks (Figure 3). For measurement error above the A-MW, a sensitivity test counting people in up to 2.5 % above the A-MW lifts the employment share somewhat, on average by around 1 pp up to 1990 and 0.5 pp for 1991–2021, but leaves the long-run trend unchanged if not reinforcing it (not shown here). A-MW employment exceeds the S-MW shares by definition. It also smooths away the increase of the S-MW share in 1993 due to the inclusion of small jobs. High levels are reached before 1979, but their restriction to the private sector inflates the level compared to later years when the public sector became included. On balance, from 1979 to 2021, A-MW employment halves, from 12.9 % to 5.9 % (8.9 %–4.1 % for S-MW). The halving of the employment share is twice as steep a decline as that of the A-MW level itself.

Figure 3:

Employment shares (%) up to adult minimum wage (A-MW) versus statutory minimum wage (S-MW) and real adult minimum wage of 1979 (A-MW-79), 1974–2021. Source: author’s calculation using linear interpolation in classes of CBS wage distributions (see Appendix).

Notably, this sharp reduction differs from at least some rise that the simple competitive demand and supply model of the labour market (e.g. Neumark 2018, Figure 1) would in principle predict to result from a decrease of a minimum wage from a higher level down to, ultimately, the equilibrium wage.^[23] Often this is interpreted as an aggregate own-wage elasticity of demand (OWE) affecting aggregate employment. However, any lower wage, including the minimum wage, is potentially attractive for employers and there is no a priori that this would not affect minimum-wage employment itself. Lichter, Peichl, and Siegloch (2014, 3) extensively evaluate demand elasticities and find a higher elasticity at the lower end of the labour market, which makes the common measurement of minimum wage employment effects by the OWE less appropriate. Cengiz et al. (2019, 1,407) develop their approach of wage bins partly to show that increasing the minimum wage does not causally affect wages higher up the distribution; in addition to wage bins, Bezooijen, Berge, and Salomons (2022, 21) distinguish a specific minimum-wage elasticity MWE next to the general OWE (and actually find it to be positive in the case of increasing youth minimum wages that they studied). The elasticity works both ways, for declining wages as for increasing wages, and is symmetrical in case of a linear demand curve.^[24] For the minimum wage, this symmetry may not necessarily show up as increases are obligatory while decreases are not.^[25] That may be the case in the short run, but it seems largely irrelevant in the longer run as new supply may be hired at the lower minimum wage^[26] and long-run demand elasticities are thought to be higher compared to the short run.

Naturally, this is all in a ceteris paribus world and actual employment effects may depend on a host of other control factors commonly found in the literature (e.g. Manning 2021b). For example, such a lowering may discourage the supply of labour following its own-wage supply elasticity.^[27] Thus, in practice, employers might not always find employees willing to be hired at the lower MW. This, however, seems highly unlikely in the present case in view of the ample supply of labour that became available at the same time (and the high personnel turnover in footnote 25). The steep decline of the MW went together with an unprecedented escalation of the rate of unemployment from 3.9 % in 1979 to 11.7 % in 1983, with a strong overrepresentation of youth and less educated persons (Salverda 1992, 1998) who are often candidates for minimum-wage jobs. This was followed by a gradual decline to an unemployment rate of 7.5 % in 1994, when the long period of minimum-wage employment decline ended – still with an ample supply at twice the level of 1979 and without any visible effect of rising demand. While from 1981 to 1983 employment declined by 5 %, it grew significantly thereafter (see Figure A.1) but minimum-wage numbers continued to fall with the declining share.

4.2 Employment Just Above the Minimum Wage

For an evaluation of the expectation that a lower minimum wage would not specifically benefit minimum-wage employment but employment growth more broadly, I consider the evolution of employment above the lowered minimum wage, where new wage options were opened for employers. This takes inspiration from the state-of-the-art studies of minimum-wage increase that aim to disaggregate employment effects by ‘wage bins’ below and above the minimum wage (Cengiz et al. 2019 – up to $ 4 below and above; Bezooijen, Berg, and Salomons 2021 – up to € 2.50).

I take a look, first, at the original, real level of the A-MW in 1979 (further: A-MW-79, Figure 3 dashed line) during the rest of the period, i.e. all wages below € 5.15 in 1979 inflated to € 13.26 in 2021, opening up a gap of € 2.42 to the A-MW (€ 10.84). Importantly, the employment share of A-MW-79 has not grown but has remained basically unchanged instead, hovering around 15 % of total hours worked albeit with some clear ups and downs. As such, this employment has evolved in line with total employment (Figure A.1) while the frozen level of real pay implies a decreasing share in total earnings, given the rise in the mean hourly wage (Figure 2; footnote 2). Apparently, the employment coverage of the A-MW-79 has hardly altered and its bite relative to the mean has fallen (not shown). The two results of A-MW and A-MW-79 should be seen as stylised facts of decline and stability, respectively. Equally important, though the share of the current A-MW fell from 100 % of A-MW-79 employment to around 40 %, this has hardly affected the wage distribution within the A-MW-79 segment. The upper half of total A-MW-79 employment remains largely unchanged at roughly between 83 % (€ 11.49) and 100 % of the wage level of the A-MW-79 – representing an unchanged wage bin with a width of € 2.35. The employment share up to half the level of the A-MW-79 mostly drifts around 2 %–3 % of total employment, which is close to half of A-MW employment since about 1990.^[28]

In addition to this picture of ‘within’, little or no change occurs ‘above’, i.e. in the wage bands up to 10 % or 20 % above the threshold of A-MW-79 – two further wage bins of € 1.33 each (Figure 4). In total, this implies an unchanged density for the bottom quarter of the wage distribution. Evidently, it means a general long-run standstill of real wages for a large chunk of employees at the lower end of the wage distribution with a virtually unchanged density distribution within that chunk. The average increase in wages concentrates in the upper three quarters of the wage distribution as a whole. The very long duration implies a stable situation, unaffected by firms’ new hirings at the lower wages that were in principle enabled by the falling minimum wage. This represents a third stylised fact.

Figure 4:

Employment shares (%), up to adult minimum wage of 1979 (A-MW-79) and bands above, 1979–2021. Source: author’s calculation using linear interpolation in classes of CBS wage distributions (see Appendix).

Only the third band to A-MW-79 plus 30 % shows more shrinkage (−9 pp), implying that part of it has shifted to higher real hourly wages. By contrast, low-wage employment, which is commonly defined as wages below or at 66 % of the median wage (€ 14.36 in 2021), does grow – from 11.5 % in 1979 to 18.6 % of all hours worked in 2021 (not shown). It reflects the 8 % rise in real value of the median wage while A-MW-79 remains unchanged: the growth of low-wage employment rests on the persistent stagnation of pay. A-MW employment happened to be almost identical to low-wage employment in the year 1979 but amounts to less than one-third of it in 2021. By then, A-MW-79 employment falls short of low-wage employment by one quarter (4.5 pp), while the A-WM-79 plus 10 % band virtually matches it in 2021. Note that the rise of the low-wage employment share implies an equally large shrinking of the band between 66 % and 100 % of the median wage, reducing its size from 39.5 % to 31.6 % of all hours. While the 10 % and 20 % bands above the minimum wage remain virtually unchanged, the thinning of the 30 % band may have other reasons than the evolution of the minimum wage, such as the famous job polarisation, in other words a contraction of middle-level occupations due to technological change (e.g. CPB 2023).

Potentially, the presence of CLAs could offer an explanation for the simultaneous fall of A-MW employment and A-MW level as they could prohibit employers to actually pay the lower S-MW where this falls short of the lowest CLA wage scales. However, witness Figure 2, those scales have been lowered down to the level of the very S-MW and became available for hiring. Consequently, the most of the long-run decline in minimum-wage employment may be interpreted as that employers have massively refrained for reasons to be clarified from following the decline of the A-MW and profit from the opportunity of paying lower wages than before.

4.3 Composition

The demographics^[29] of general employment change significantly over the period: female employment expanded very significantly and youth employment shifted drastically from full-time to part-time in an overlap with prolonged participation in education. Surprisingly, these changes are hardly reflected in minimum-wage employment (S-MW): the share of females (all ages) actually decreases from 60 % to 50 % and that of youth remains virtually the same, on balance, at around 50 %. For A-MW-79 in 2021 (taken as S-MW plus 20 %), the female share declines somewhat less to 55 % and the youth share does show some decline to 45 %. These modest demographic changes go together with substantial changes in supply behaviour (in terms of jobs).^[30] The share of part-time employment among S-MW employment doubles, from 31 % to 60 % in 2021.^[31] Secondary earners in higher-income households have come to play a dominant role in low-wage employment and may entertain different preferences regarding, e.g. pay and occupational level versus working hours and commuting time (Salverda and Rook 2023; Salverda 2018b; Salverda and Thewissen 2018).^[32] Demographic stability and simultaneous behavioural change define a fourth stylised fact.

Limited information is available on the characteristics of employers employing minimum-wage workers, but, fortunately, we can fathom the changes in firm size. Between 1979 and 2021, the share of small enterprise (< 10 employees) in S-MW employment halves from 39 % to 19 % while the role of larger enterprise (10+ employees) rises from 61 % to 81 %.^[33] Behind this is not a shrinking share of small firms in total employment but a spectacular fall in the incidence of S-MW among them from 22 % to 6 %, while that of larger enterprise hardly declines (6 %–5 %). It leaves small firms only slightly overrepresented nowadays. For A-MW-79 employment in 2021 (again S-MW plus 20 %), the incidence among small firms (20 %) lies below that of 1979 while among larger enterprise it expands significantly from 6 % to 13 %.^[34] About half of these changes concentrate in the period between 1979 and 1989 when the declines are steepest. Over those same years while S-MW employment more than halves the distribution of S-MW employment over branches of industry^[35] does not really change. The combined share of the main contributors: agriculture, trade, and hotels, restaurants and catering (horeca), remains unchanged at 62 % while manufacturing and construction (−12 pp) trade places – plausibly quite literally – with growing temporary agency work (+11 %) as part of business services. After 1989, the incidence among small firms continues to decline gradually while that among larger enterprise grows. Compared to 1979, A-MW-79 employment in 2021 reduces trade employment (−24 pp) to the advantage of horeca (+6 pp) and, again, business services (+19 pp). Because small firms make up the vast mass of all firms (> 80 %), the halving of the minimum-wage incidence among them suggests a great reduction in the number of firms that are actually involved with minimum-wage employment. The increased role of larger enterprise points to a meaningful change in the nature of firms involved in minimum-wage employment, and largely within the same industries – a fifth stylised fact.

Together with the steep decline of the real and relative minimum wage, the five stylised facts raise important questions about the nature and significance of the bite of the minimum wage and its relationship with employment – Neumark’s issue. The diverging coverage of the current A-MW and the A-MW-79 and the shift towards larger firms cast doubt on traditional assumptions of firm behaviour regarding the minimum wage. Changing preferences of labour supply regarding the nature of jobs may have contributed to the standstill of real wages for the bottom quarter of wages.

5.1 Dutch and International Literature

Also in the Netherlands, minimum-wage employment effects are a bone of contention in the political debate, perhaps more strongly even than in other countries because of the coupling with benefits and public finance. However, no Dutch study is available of the striking experience of the much-lowered minimum wage, let alone that minimum-wage employment more than withered with it. There is no lack of attention paid to minimum-wage effects in the country but independent studies are very few, also on minimum-wage increases. The CPB, the government’s main economic advisory body,^[36] habitually advises on policy proposals concerning the S-MW. However, this mostly regards the effects on public finance following from the idiosyncratic coupling. Understandably, this is the government’s most pressing concern, and it may also be easier to examine than employment effects. Recent reports of CPB that deal with those effects notably derive the elasticities from the international literature (literally taking the median of six American studies, one German and one Hungarian), which are then applied to Dutch wage and employment data (CPB 2020a,b). Thus, the reports tend to move with the trend in the minimum-wage literature towards lower or zero elasticities and nowadays they proclaim much smaller effects than before (e.g. CPB 2016). Without independent checks of national elasticities, this work is better considered as simulations of foreign situations on Dutch data without regard for the specificities of the S-MW including the long tail of Y-MWs.

Three independent contributions do derive their results on the basis of Dutch microdata. All are partial and focus on the very recent event of changes made to Y-MWs in 2017–19, which have lifted the 22 and 21 year olds from their Y-MWs to the A-MW, and they have nothing to say about the effect of the S-MW overall nor on the lowering. Bezooijen, Berge, and Salomons (2021) adopt a difference-in-difference design with the wage-bins approach of Cengiz et al. (2019) and find a ‘robust absence of negative effects’ on employment and a substantial increase in income from earnings, which is largely due to spill-over effects. Two other studies performed at the request of the Ministry of Social Affairs and Employment (Werff, Zwetsloot, and Weel 2018; Zwetsloot et al. 2020) use also a difference-in-difference method and find no (short-run) effects on labour force participation and educational participation together with an increase in earnings. An interesting fourth study (Kabátek 2021) evaluates the effects that Y-MWs as a system may have on the employment chances of young individuals when they age and qualify for the next-higher Y-MW, with pretty steep steps (16 % on average).^[37] It finds a clear effect that youth tend to be fired from their jobs when they approach their next birthday with its correspondingly higher minimum wage and are then replaced by a younger and cheaper person.

Therefore, I turn to the international literature where ‘the employment effect of the minimum wage is one of the most studied topics in all of economics’ (Schmitt 2015). Wolfson and Belman (2019) mention 800 English-language studies since 1991; plausibly 200 more may be added in other languages. In a long, especially American tradition, there are so many studies that comparative overviews and critical metastudies^[38] are also multiplying (e.g. Belman and Wolfson 2014; Doucouliagos and Stanley 2009; Neumark and Wascher 2006), which, in turn, are scrutinised by further overviews and metastudies (Dube 2019; Neumark 2019; Neumark and Shirley 2022; Schmitt 2015; Wolfson and Belman 2019). Labour econometricians seem to keep each other occupied in a play of leapfrog – if not ‘trench warfare’ (Manning 2021b, 9). Along the way, the dominant view slowly but increasingly seems to become that employment effects are barely visible. It is as if workers in the ‘minimum wage’ zone (and perhaps a spill-over zone above) can be given a higher wage at will. An earlier qualification of the debate as generating ‘a lot of heat but not much light’ (Dolton, Bondibene, and Wadsworth 2012) may now be less to the point.

The growing econometrical sophistication along the way (Allegretto et al. 2017) is one explanation of the new insights but the prime reason is that the analysis was liberated by Card and Kruger (1994) from the a priori that employment could only decrease in response to an increase and that the question was by how much? Their ‘new economics of the minimum wage’ has fostered attention on the employer side and its heterogeneity. Traditionally a black box, light is now increasingly cast into this darkness with the help initially of researchers’ own establishment surveys (e.g. Card and Krueger 1994; Machin, Manning, and Rahman 2003) and progressively of large matched employer-employee datasets. This approach revives older arguments (Card et al. 2018; Lester 1946) arguing that when the price of labour mounts employers may have other ways at their disposal to respond than simply cutting jobs while they may vary also in their use of those other ways.

5.2 The International Significance of Minimum-Wage Declines

At the same time, virtually all studies and metastudies examine effects of raising the minimum wage, saying nothing about lowering it. It would be a mistake to consider the Dutch lowering an exception, a peculiar outcome needing no further attention. Scrutiny of the OECD’s minimum-wage database points to a much broader, international significance of minimum-wage reductions (Table 1). The database contains about 1,000 annual changes in 25 countries’ minimum-wage levels, nominal and relative, over the period 1961–2022. Admittedly, nominal lowerings basically concern only the Netherlands (1984) and, much later, Greece (2012, 2013).^[39] However, the initial Dutch decrease is followed by successive nominal freezes generating a long-run decline in the real and relative value of the minimum wage, which are as material to firms’ employment decisions as formal increases of the minimum wage.

The OECD data indicate that real declines (deflated by CPI’s) concern 35 % of all annual observations, with an average effect of −2.4 %, sufficiently large to allow and deserve further study.^[40] Relative declines to the median or the mean Kaitz indices occur in no less than half of all cases, significantly more frequently even than positive changes and implying also that when the real minimum wage increases the median and average wage often rise more. These high frequencies imply serial declines that endure beyond a single annual change, witness the downward shift by 4.4 pp on average between 2000 and 2022 of the median-wage Kaitz in seven countries.^[41] Thus, minimum-wage declines constitute a full-grown elephant in a room unnoticed by analysts who are busy scrutinising the little hairs at the end of the animal’s tail.

Unfortunately, the OECD offers no data on minimum-wage employment shares to match with these results, precluding adequate empirical research. But, tellingly, the long decline, both real and relative, of the Dutch minimum-wage level and employment since 1979 seems to be entirely shared by the American federal minimum wage (Figure 5), which is the principal subject of minimum-wage research. Secular trends of the level are largely the same in the two countries (panels A and B).^[42] Ultimately, the greater volatility of the saw-tooth pattern caused by the notoriously difficult American political decision making on any rise and the more gradual development due to the Dutch regular uprating both produce almost the same result. The American employment share (panel C) concerns workers paid by the hour (55–60 % of all workers, next to salary workers).^[43] In addition, this decline (known from 2002 to 2022 only) is largely shared between the US states (31) with own state minimum wages, which commonly exceed the federal one, and those (20) where only the federal minimum wage applies.^[44] It leaves a surprisingly small difference in federal minimum-wage incidence in 2022: 1.1 % for the former 31 states versus 1.7 % for the other 20 states. Clearly, if the Dutch minimum wage would be a black swan of (joint) decline, it is in good company of another much sturdier one.

Figure 5:

Minimum-wage levels (1979 = 100) and employment shares (%), Netherlands and United States, A and B: 1969–2022, C: 1979–2021. A: Real MW, 1979 = 100^a. B: Relative MW, 1979 = 100^b. C: MW employment share, %^c. Notes: Dutch A-MW and US federal MW. ^aReal MW: deflated by CPI, ^brelative MW: Kaitz median-based index, ^cemployment share US among hourly paid workers and Netherlands as in Figure 4. Sources: see Appendix, OECD databases and BLS, characteristics of minimum-wage workers.

5.3 Researching Minimum-Wage Declines

These findings make it hard to understand that employment effects of minimum-wage declines have never been the explicit subject of economic analysis (except once for Greece – see below). Note that the high frequency of declines implies that they will be present in many time-series studies, but unfortunately stay under the radar which remains fixated on the effects of increases. It seems a mistake to focus minimum-wage analysis exclusively on increases and ignore the many decreases. These deserve examination in their own right and their analysis may also benefit the analysis of raises. Technically, such analysis may be more demanding, as declines usually do not figure as clear events or breaks and may be harder to disentangle from other, confounding trends in the economy. However, taking a longer-term perspective and using firm-level data can help as declines cumulate over time and influence employer expectations.

Methodologically, in time-series analyses, decreases are accounted for on the same basis as increases, generating a single measure of employment elasticity be it upward or downward. In other words, ‘the implicit assumption is that, controlling for the other explanatory variables, employment would be the same over time if the minimum wage were constant’ (Card and Krueger 1995, 183). It implies that any prima-facie difference between upward and downward effects needs to be attributed to and entirely accounted for by control variables.^[45] If not, the elasticity for both will be misestimated. In addition, the ceteris paribus condition for estimating long-run elasticities of minimum-wage hikes as distinct from the short-run ones is not satisfied in the case of declines after the hike. ‘If minimum wage increases are sufficiently temporary … there would be little difference in the observed short- and long-run employment responses’ (Sorkin 2015, 317). In Sorkin’s view, firms’ expectations regarding the ‘temporariness’ of the hike matter for their long-run labour demand, depending on the dissimilar nature of their production.^[46] This temporariness affects both time-series, cross section and panel studies. Neumark (2019, 317) deems this a ‘different perspective’ on the bite of the minimum wage, but it may be more fundamental than that given the high frequency of declines.

Theoretically, employment effects of a lowering may differ as much as those of an increase between the two main strands of minimum-wage research: the mainstream neoclassical one and the ‘new-economics’ monopsonistic one. In their textbook versions, the former is fully competitive with price-takers on both sides: firms and employees, while in the monopsonistic approach employees can be price-takers while firms are not if they enjoy some degree of labour-market power, or employees and firms can be mutually non-competitive to the extent that firms’ power depends also on workers’ heterogeneous non-wage job preferences regarding firms (Card 2022; Card et al. 2018; Manning 2021a). As the degree of power may vary across firms and puts their individual behaviour on the research agenda, heterogeneity is an essential corollary of monopsony. Following the neoclassical strand, minimum-wage employment will increase if the minimum wage is lowered^[47] to a level closer to or at the equilibrium wage, while following the monopsonistic strand, minimum-wage employment and/or the wages of the employees will decrease depending on the level at which the minimum wage is situated before and after the lowering relative to monopsonists’ profit-maximising wages (cf. Elliott 1991, Figures 10.3 and 10.5).

Differently from these textbook versions, a firm will be legally obliged to pay an increased minimum wage but not a nominally decreased one, let alone a nominally unchanged one facing real or relative decline. Commonly, employers can hire at will but not necessarily fire because of institutional regulations. Upwards, the minimum wage is an obligation for the firm and provides an attraction for new and existing labour supply. Downwards, by contrast, employees at the initial minimum wage may retain their pay turning the minimum wage into a label attached to a shrinking part of a firm’s labour force and a disincentive to new labour supply. In the longer run, job mobility will substitute new hires paid the lower wage. This legal difference lends a degree of freedom to an employer who in his right mind would trade off maintaining the existing situation – freezing the previous minimum wage – at no more costs than before, against the uncertain net benefit of lower labour costs minus the deduction of the costs that would be incurred by the firing and hiring actions needed to bring the new situation about. Such a freeze can offer an attractive lowering of labour costs in a longer-run perspective of increasing productivity or passing through to prices (Aaronson 2001) without the need to adopt the lower minimum wage.

The least one can say is that a lowering potentially incites a more heterogeneous response than an increase. Georgiadis, Kaplanis, and Monastiriotis (2020) provide the one and only study of the employment effects of a lowering that I have been able to discover. With matched employer-employee data, they examine the effects of the extreme decrease of the Greek minimum wage in 2012, wrought by the Troïka during the Eurocrisis. The authors stress in their results the heterogeneity of firm responses ‘that are not entirely consistent with the competitive model of the labour market’ (p. 3). Naturally, we cannot infer more heterogeneity from this in lowering than in increasing the minimum wage, but the finding underlines the importance of paying attention to the firm side of the equation. Importantly, a significant decline in minimum-wage employment may go together with a shrinking and changing population of firms that actually pay the minimum wage. Remember the drastic shifts towards larger firms and part-time work between 1979 and 2021 found above for the Netherlands. Similarly, the relevant labour supply may get more restricted. The combination of both sides is important if the monopsony power of firms depends also on heterogenous preferences that people may entertain for non-wage employer characteristics.

All of this draws attention to the economics of the ‘bite’ of the minimum wage. Studies that find little or no employment effect seem to attribute this ex post to a relatively low level of the minimum wage, i.e. a weak ‘bite’, but without a firm criterion what that might be. The concept of a bite purely defined as ‘how high the minimum wage is relative to the wage distribution’ (Neumark 2017, 11) risks being tautological to the share of minimum-wage employment.^[48] Instead, the concept should be given actual economic content, independent from the level of minimum-wage employment itself. That seems to be the point of Neumark when he further advocates a ‘more serious consideration of the economic factors to help explain long-standing differences in minimum-wage employment effects across studies’ (p. 300). Behind the bite lingers the question what part of the economy does the minimum wage affect and for what reasons? This applies as much to changing minimum wages as to unchanged ones. Different bites, but also similar bites, may have different effects because of differences in those reasons. For example, assuming a difference in and heterogeneity of behaviour of monopsonistic firms compared to competitive ones, either the contributions of those two categories of firms to the relevant segment of the labour market may change and give rise to a different bite or, alternatively, the monopsonistic behaviours and/or their heterogeneity may change, perhaps even at the same time, and affect the bite in a different way.

It suggests to systematically extend the question ‘Who is affected by the minimum wage?’ (Belman, Wolfson, and Nawakitphaitoon 2015), pertaining to workers receiving the minimum wage, also to the firms paying it. These authors extract the answers for a range of demographic characteristics and a few specific industries from the existing literature. Unfortunately, they pay no attention to changes over time. On the firm side, Cengiz et al. (2019) discuss the bite by industry and London Economics (2016) does by firm size, but both do not match with employee data while Wursten and Reich (2023) do by firm size. Matched employer-employee data will be essential (Manning 2021a, 16), but very little is available over a longer time span and their use is only just beginning. What has become available pays no attention to changes over time. Bachmann and Frings (2017), who study monopsony in great detail, contend that their conclusions based on the years 2000–2002 still hold in 2015, the time the German minimum wage was introduced. It is unfortunate that matched data do not stretch back to the 1980s for a study of the times that American and Dutch declines were steepest, to probe whether the growing role of larger enterprise implies an intensification of monopsony. Shorter stretches of matched data could enable studies of more recent years but variation may be insufficient.^[49] However, pure individual matched data will need to be supplemented by the firms’ functioning in the relevant low-wage labour market, e.g. their contribution to labour market concentration (Azar et al. 2021). Also, a concentration of monopsonistic firms may actually affect the behaviour of competitive firms and blur the analytical distinction between the two.^[50]

Obviously, such a study is beyond this contribution to undertake. All in all, minimum-wage analysis must not ignore such perfectly reasonable economic facts of declining minimum-wage costs, often gradual but also accumulating over time, and try to find a way to deal with declining minimum-wage levels and shares and scrutinise the implications of significantly differing levels of minimum-wage employment. Hopefully, the growing use of matched employer-employee data together with a focus on the combined changes over time in their populations and behaviours can give birth to minimum-wage analysis 2.0. That would suit David Card’s (2022, 1,086) plea to drop the axiom that ‘markets set wages’ and instead consider firms’ wage setting an integral part of labour economics.