“Howard Schultz famously state that unions ‘have a role to play’ but are ‘not the answer.’ But, are they not? Thankfully, there is an empirical answer to this question.”
Unions have occupied a prominent place in the public discourse over the last two years or so. In the United States, left-wing politicians like Bernie Sanders and Alexandria Ocasio-Cortez have revived interest in organized labor and made the health of unions a central talking point of their campaigns. The Democratic primaries have also helped in this respect, as union endorsements can provide an important boost for candidates. Yet, as it has just been reported, once again union membership hit a record low in 2019. According to figures published by the Bureau of Labor Statistics this year, the percentage declined by 0.2 percentage points to 10.3 percent of workers. The rate has been steadily declining for more than two decades. In 1983, the first year for which the BLS collected comparable data, the rate was 20.1 percent.
However, unions continue to play an important role in politics in the United States. Last year, the flight attendants’ union essentially ended the last government shutdown. With flight attendants on strike, the threat of a halt of air transportation across the country pushed the federal government to strike a deal with its employees. Elsewhere in the world, as social unrest continues to grow in Europe, Hong Kong, and other areas, unions have also been put in the spotlight. In France, unions have continued their lengthy battle against pension reform. In Hong Kong, people—as of late—have begun forming and joining unions as part of their protests against the Chinese-backed government. However, criticism of unions has been prevalent and have come from many sides of the political spectrum. Right-to-work laws, which have played a large role in the decline of membership in unions, have become prevalent in traditionally Republican states. Coffee magnate, could-have-been presidential candidate, and long-time Democrat Howard Schultz famously state that unions “have a role to play” but are “not the answer.” But, are they not? Thankfully, there is an empirical answer to this question.
First, I’ll provide a little more context. Schultz’s position, of course, is simply one of many along the long line of criticisms against unions. A common argument is that all unions do is stand in the way of productivity—and that they simply add transaction costs to the economy. They stifle creativity and workers’ individual initiative, instead promoting a stagnant climate. Another critique is that unions’ efforts in raising wages are misguided and, ultimately, fruitless because they only raise money wages, as opposed to real wages. Real wages are those that are measured taking into account the effect of inflation and only result from increases in productivity. Or put another way, they increase the cost of labor but not its value.
It is somewhat ironic because these libertarian factions tend to see themselves as heirs of the Enlightenment tradition and, as such, seek to protect individual freedoms. Among these freedoms is, by definition, the freedom of association, which must include a worker’s right to organize and collectively bargain for better wages and working conditions.
This line of criticism, of course, tends to come from the more libertarian strands of the Right, rather than the more religious, traditionalist, or even reactionary side. It is somewhat ironic because these libertarian factions tend to see themselves as heirs of the Enlightenment tradition and, as such, seek to protect individual freedoms. Among these freedoms is, by definition, the freedom of association, which must include a worker’s right to organize and collectively bargain for better wages and working conditions. Why this in itself is not considered initiative or even market forces at work is beyond me.
The second criticism is one that I find slightly more interesting. One can even accept the basic framing that productivity and real wages are important, as well as that rises in only nominal money wages are unsustainable. However, it is an observable fact that—at least in the United States—compensation has not increased in proportion to productivity since the early 1970’s. A now all-famous chart by the Economic Policy Institute (EPI) shows that average real compensation for workers is essentially flatline after 1973, while productivity continues to rise. This framing has been criticized because it does not take into account that wages are not the entirety of compensation—and that if one looks at full compensation, the gap disappears. The problem, however, is that average compensation can be a misleading indicator. If the compensation of top earners goes up (and that of low earners remains constant), average compensation will go up anyway. And, in fact, multiple sources confirm that this is what has been happening. A Pew Research study and a 2013 article in The Economist among others, note that most income growth in recent years has been captured by top earners. With this in mind, the EPI has updated the data to reflect the need to look at productivity vs. full compensation (as opposed to wages only) but also to control for the rise in top incomes. For this, the EPI use productivity and non-supervisory compensation instead of the average, which is much more indicative of what the typical worker earns. Once both factors are accounted for, the fact remains that compensation has barely grown with productivity.
This brings us back to unions. Analyzing the EPI data, economists Larry Summers and Anna Stansbury sought to find possible causes of this divergence, given that—prior to the 1970’s—production and compensation grew hand-in-hand. Their findings suggest that productivity might be having the same effect but that external structural factors are pushing wages down. They do not analyze these possible causes in detail, but one of the potential explanations they mention is the decline in unionization. So, the original critique that unions are ultimately useless because the only thing that matters is the rise in real wages through productivity starts to look a little shaky. Sure, productivity matters, but we, at least, have a suggestion that for the typical worker, unions might play a role in ensuring that compensation reflects productivity. Of course, it is important to be cautious to not go in the complete opposite direction and suggest that unions are the sole explanation. In social issues, there are always multiple competing variables. Changes in the supply and demand of labor, for example, are probably a factor. What we need is to show that unions are one of them.
Fortunately, there is also empirical data that allows us to shed some light on this hypothesis. The Organization for Economic Cooperation and Development (OECD) provides fairly detailed statistics concerning the labor force in the world, including unionization rates and various measures of worker compensation. This permits a look at the relationship between these variables. For this, I use data for 2013 or the closest available year. As for why 2013 is the year being used, the data that is reported by the OECD is provided by each country, and not every year is available for every country; as such, 2013 is the most recent year for which a large share of countries have available information. Regardless, across the whole period covered in the data, there is much more spatial than temporal variation. This means that the difference among countries on the same year is much greater than the difference from one year to the next within the same country. This is true for the whole period covered by the data from 2000 to 2017, but it is particularly true for years close to one another. This means that (in general) data for 2014, for example, is a good approximation of data for 2013.
We can start, then, by looking at a straightforward question: What is the relationship between average real wages (also controlling for Purchasing Power Parity) and unionization?
SOURCE: My own elaboration based on OECD Labor Statistics
The data indicates a positive relationship; countries where unionization rates are higher also have higher real wages. However, it is not an overly strong relationship and, as already mentioned, average wages are not necessarily the best indicator of what a typical worker makes—therefore, not necessarily the areas where unions would have the highest impact.
But there are other indicators of compensation that can provide more insight into this relationship. One example is the correlation between the real yearly minimum wage and unionization rate. This does away with any effects that top incomes might have on the average compensation, and, furthermore, it addresses the most vulnerable workers in the labor force, namely those that earn the minimum wage.
SOURCE: My own elaboration based on OECD Labor Statistics
Again, there is a positive relationship: the higher the rate of unionization, the higher the annual real minimum wage (again, also controlling for Purchasing Power Parity). Now, even if the analysis uses the real minimum wage, it is true that it is unrealistic to expect Mexico to have the same real minimum wage as Belgium, simply because Belgium is a much richer country per capita. There is still some value to this however because even among richer nations, it appears that higher unionization translates to a higher real minimum wage. Compare, for example, Belgium and Luxembourg to Korea and the United States. And finally, if the countries with higher unionization rates are among the richest, it casts some doubt on the position that unions have a negative impact on economic activity.
But there is a further relationship to be explored, one that does away with the effect that each country’s per capita income has on the analysis. The incidence of low pay provided in OECD statistics is defined as, “the share of full-time workers earning less than two-thirds of gross median earnings of all full-time workers.” In other words, low pay is relative to each country’s labor market, rather than based on an absolute benchmark. This allows for comparisons between countries that have widely different levels of per capita income because it simply indicates what percentage of workers are in a bad economic situation. Incidentally, this is also a measure of economic inequality: when workers’ earnings are more similar to each other, the percentage who earn a low fraction of the median becomes lower. Higher income equality, thus, translates to lower incidence of low pay and vice versa.
SOURCE: My own elaboration based on OECD Labor Statistics
Once again, higher unionization rates have a positive impact on workers. This time the statistical relation is negative—that is, the higher the rate of unionization, the smaller the incidence of low pay, which is essentially the fundamental goal of unions.
It is true that these are all mere correlations and, as the Internet is fond of reminding us, correlation does not imply causation. But this criticism is unwarranted. For one, while it does not imply it, it at least hints at the possibility. The inverse relationship still holds: causation should imply correlation. Put differently: if unions really are the reason for these positive effects, the data would look the way it does now. In any case, in statistical terms, there is little more one could ask of the data. The correlation coefficients of the last two datasets are around 0.5. In layman’s terms, this is a number that tells us how close the relationship is to being perfect, with 0 being no relation, and 1 and -1 being a perfect relation (positive or negative depending on whether the relationship is direct or inverse). However, we can go beyond a simple correlation and instead run a regression that will tell us the precise magnitude of the relationship—and how likely is that relation to be real or a statistical fluke. The first value is given by the regression coefficient. In the case of the incidence of low pay, for example, the coefficient for unionization as the explanatory variable is -0.1466. This means that for every percentage point increase in unionization, the quantity of workers with low pay is reduced by 0.1466 percentage points. This may not seem like much, but we have to keep in mind that in social phenomena there are always countless factors at play. This is where the other number, the p-value comes in, which in this case is 0.006. A not entirely accurate but close enough explanation of this number is that when it is below 0.01, the probability that the coefficient is not a statistical fluke is 99%, which is the case here. This is about as good as it gets with social phenomena.
Going back to the more general criticism, causation is notably hard to establish philosophically speaking, as David Hume argued. One could even say that almost nothing ever establishes causation, making this criticism a tautology. Thus, this kind of almost metaphysical speculation is unwarranted and mostly useless. If we stick to the evidence on the other hand, we have some clear findings: in the places where unions are stronger, the objectives that they set to achieve (and which should be celebrated by anyone regardless of political leanings) are being fulfilled. Workers are earning better wages, and income equality is higher where unions are more firmly established.
The argument here is neither that unions are a silver bullet, nor that any criticism of them is unwarranted. Cases of corruption among union leadership should obviously be criticized. This should be even more true from the Left, if this starts to hinder unions’ ability to effectively represent workers’ interests. But two central issues remain true. The first is that unions, in principle, are nothing more than the expression of the workers’ individual rights to associate and to use their collective power to improve their bargaining ability. This is something that even right-leaning libertarians will sometimes acknowledge. The second is that the general objectives of unions—namely, fair compensation for a worker’s labor and better working conditions—should also be objectives of any decent society. It happens, however, that empirical evidence suggests that unionization is actually an effective vehicle in achieving these goals. Unionization as a general aim, therefore, should be desirable for any society, and valid criticisms of the union movement should not go as far as to call for their abolition.
Néstor de Buen holds an M.A. in social sciences from The University of Chicago. He has previously written at Quillette. He can be reached at email@example.com or on Twitter @nestor_d