Week 3 - Harasztosi and Lindner (2019) Who Pays for the Minimum Wage?

What is a summary of this paper?

• Aim: study the causal impact of increases in the minimum wage on most affected firms, using a natural experiment in Hungary which saw a large, unexpected, and persistent increase in the value of the minimum wage.

• Empirical strategy: The authors adopt a difference-in-differences methodology comparing firms more affected by the reform (more of their workers were paid below the new minimum wage before the reform) vs. firms less affected by the reform (less of their workers were paid below the new minimum wage before the reform) after vs. before the reform.

• Findings:

  • The policy decreased employment (in part by substituting labour with capital), but still led to large increases in workers’ income

  • A large share of the economic incidence of the policy was shifted from firms to their consumers

  • Disemployment effects were larger in industries were shifting the cost of the policy to consumers was more difficult.

B. What are the key assumptions under which the estimates in the paper capture the causal effect of the minimum wage, and what evidence do the authors provide to indicate that these assumptions are plausible in their setting?

• Because DiD, the first key assumption that their analysis relies on is that of a “parallel trend” between “treated” and “control” firms. Note that, in this paper, the “treatment” variable is not a indicator variable taking the value of 0 or 1, i.e., in which some firms are clear treatment firms and others are clear control firms. Instead, they have a “continuous” treatment variable: they look at how outcomes compares after vs. before the policy for firms with different values of F A, which is “the fraction of workers for whom the 2002 minimum wage binds” (who were paid below the 2002 minimum wage prior to the policy).

  • To support the plausibility of this assumption, the authors systematically show that trends for their key outcome variables were parallel for firms with different values of F A in the years prior to the reform. They show that graphically (i.e., that the estimated βt were all around 0 prior to the reform) in Figures 2, 3, and 4. They also show that in the regression analysis (i.e., the changes between 2000 and 1998 were similar for firms with different values of F A) in Table 2 (col 5 and 6), Table 3 (col 3), and Table 4 (Panel C).

• Another assumption that the authors discuss in the paper is what is called the “SUTVA” assumption in econometrics: it must be that firms with much lower values of F A are indeed much less affected by the policy than firms with much higher values of F A. So it cannot be that the bigger increase in labour costs for firms with higher values of F A somewhat spills over to firms with lower values of F A through interactions between firms in the overall Hungarian economy (e.g., through the labor market, the output market, or the capital market).

At the top of Section III, the authors write: “The previous section shows that the minimum wage increase had a large positive effect on real wages and a small negative effect on employment. The simple consequence of this finding is that the income of low wage workers increased in response to the minimum wage.” Why can they infer the statement in the second sentence from the statement in the first sentence?

The income of low wage workers is the product of employment times average wage.

If employment decreases by about 10% and average wage increases by about 50%, then income increases by about .9 × 1.5 = 1.35 or 35%.

What is the purpose of Table 5 and what is the main takeaway from Table 5?

The purpose of Table 5 is to calculate the incidence of the policy on firm owners vs. consumers, i.e., how much of the change in total labor costs falls on consumers vs. firm owners.

The incidence on consumers is calculated as the change in the amount that firms receive from consumers in exchange of the value that they created (revenue net of material and other expenses). The incidence on firm owners is calculated as the change in firm profits plus depreciation expenses.

The bottom line is that 75% of the incidence falls on consumers vs. 25% on firm owners.

What do the authors show in Figure 5 and how does it help the authors conclude that “our evidence underscores that the minimum wage is an effective redistributive policy”?

The authors have shown that the policy increases the income of lower-income workers, who are most affected by the minimum wage. In that sense, the policy is redistributive.

The point of Figure 5 is to study whether lower-income consumers are more likely to buy goods from firms most affected by the policy compared to higher-income workers.

If that was the case, then the incidence of the policy would fall more heavily on lower-income consumers (we know already that it falls mostly on consumers in general), which would undermine how redistributive the policy is: lower-income workers would have higher income but also face higher prices!

Figure 5 shows that this is not the case: lower-income consumers spends a similar share of their consumption as higher-income consumers on goods from firms most affected by the policy. Therefore the policy is actually redistributive.

What explanation do the authors propose for the finding that the disemployment effects are much larger in the tradable and exporting sectors than in the non-tradable and service sectors?

The argument is about the elasticity of demand faced by firms in their output market. Firms in the tradable and export sectors are more likely to compete with foreign firms. As a result, buyers from these firms could easily switch to goods produced by non-Hungarian firms if the price charged by Hungarian firms were to increase following the increase in the minimum wage. In other words, Hungarian firms in the tradable and export sectors face a more elastic demand for their products. This is why they are not able to shift as much of the increase in labor costs to their consumers and they thus have more incentives to reduce their labour costs, by using less labour, in order to maximize their profits, leading to a larger disemployment effect. By comparison, firms in the non-tradable and service sectors are more likely to compete with other Hungarian firms that face a similar increase in labor costs. As a result, buyers from these firms could not as easily switch to goods produced by firms unaffected by the policy if the price charged by Hungarian firms were to increase following the increase in the minimum wage. In other words, Hungarian firms in the non-tradable and service sectors face a less elastic demand for their products. As a result, they are able to shift a larger share of the increase in labor costs to their consumers and they thus have less incentives to reduce their labour costs, by using less labour, in order to maximize their profits, leading to a smaller disemployment effect.

What is the main takeaway from the paper?

The main takeaway from this paper is that an increase in the minimum wage does not necessarily lead to large disemployment effects – a serious concern in policy circles – when firms are able to shift the increase in labor cost onto their consumers, which is more likely to happen when all firms in the same output market are subject to the same minimum wage.

What is the main limitation of this paper?

A limitation with this paper is that even firms who have a lower value of F A might be somewhat affected by the policy, for instance, through the output market: if their competitors with a higher F A raise their prices because of their higher increase in labor costs, firms with lower value of F A may be able to increase their profits either by increasing their market share (if they don’t increase their price to the same extent) or by increasing their price.

So the estimates from this natural experiment may not capture exactly the answer to the question of interest: what would be the impact of increasing the minimum wage on the most affected firms. To answer that question, you would need a control group of firms not affected by the policy at all, which is difficult to find in practice (Section V shows the results of empirical exercises doing a bit more on this if you are interested).

However, this limitation is very unlikely to invalidate the main takeaway from this paper (the estimates might just be slightly different from the estimates that we would get if firms with lower value of F A were not affected by the policy at all).