This is the second part of the diary posted last weekend, Economics in the Real World: Macro versus Micro. This case study is the reason why I chose the Minimum Wage as the focus for looking at the difference between the Macroeconomic Perspective and the Microeconomic Perspective. It is an elaboration of an earlier diary, but this time I go into the nitty gritty details.
Traditional marginalist economists wear blinders that restrict them to an incomplete view of the world. What are we able to see when we take those blinders off? That is the fundamental issue of this series. In today's case, we will see some marginalist economists who are willing to put the assumptions of the standard model to the test, and discover that the assumptions do not hold up. The story begins beyond the fold ...
Chapter 1: Economics in the Real World: Macro versus Micro. Part 2: Case Study.Note: If you are joining from Part 1 last week, you can skip the first section if you wish. If you are not interested in the empirical analysis of the raw data and want to get into the debate in the economic literature, skip the second section.
The Model for Public Discourse: True By Definition
The official looking report on why a Minimum Wage increase in Ohio will supposedly result in job loss can be found at the "E"PI propaganda outlet (See the References
Turning to the report, its "economic model" is extremely simplistic:
The proportional impact is the Percentage Wage Increase times "Labor Elasticity". The impact on employment is the number affected by the wage increase times the proportion impact.
And what is "Labour Elasticity"? That is the "Percentage Change in Employment divided by Percentatge Wage Increase".
Quite clearly, if you have the right number for "labor elasticity" the rest is true by definition. So the whole prediction of job loss is based on a negative value for "Labour Elasticity". If this value is negative, then wage increases drive employment down. If it is positive, then wage increases drive employment up.
And if it is zero, or more precisely "not significantly different from zero", then there is no positive or negative impact that can be detected.
Stepping into the looking glass ...
If this was a Microeconomic issue, and only a Microeconomic issue, then we would know something about this "elasticity". In a single, well-organized market, when you increase the price of a service for sale, and there is no other change in market conditions, then "no reaction" leaves the quantity sold unchanged. The increase in price will make the service less attractive to buyers, so if there is a reaction, the reaction will be fewer sales of the service.
In brief, at the Micro level, inside a single market, and with stable market conditions, the observed "Price Elasticity" will have 0 as an extreme value, and head off into the negative range. In fact, this observation is so fundamental and so entrenched that when reporting price elasticities, economists often omit the sign, since it is so widely understood to be negative.
... and backing out again ...
Of course, a wage is not only the price of labor services, it is also the hourly rate of income earned. And a change in income is in market conditions. If we were only talking about a wage increase at one fast food establishment, it might be safe to ignore the impact ... but when we are talking about increasing the wages of everyone on the minimum wage and most people near the minimum wage, it can't be assumed. So the overall response will be driven by the "price elasticity" and the "income elasticity". We can't just assume that the overall impact is negative.
At the same time, we know that the present level of activity in the economy is below the full employment level. This means that across most of the "markets" with people selling labor services, there is a quota. Businesses will not hire more people once they have reached the limit of what can be sold.
In a market operating under a binding quota, there is no single equilibrium price. Instead, there is a trading range. Any price in the range between the lowest amount the supplier (worker) will take and the highest amount that the demander (employer) will accept results in the same quantity sold.
Since the prevailing wage in a number of labor "markets" all across the country is above the minimum wage, it is clear that much of the wage increase involved with increasing the minimum wage will involve sliding people up through this trading range. And the "price elasticity" inside the trading range is 0, zero, nada, zilch.
I have no reason to believe that the entire "price effect" across all industries will be zero ... but with zero price effects for much of the impact across many low-wage employers, a very small "price elasticity" is certainly possible. And in that case, we can see a zero overall impact from only a small "income elasticity" effect.
This is why we cannot simply assume that there will be job losses, and ask how many. We have to ask whether there will be job losses, in the situation that we are looking at. We need to get some idea of the sign of the overall price elastivity, rather than starting from an assumption that it is negative.
Indeed, the publication by the "E"PI predicting a job loss is not really an analysis at all. It simple quotes a wage elasticity from somewhere else, and then uses that as a magic number.
How Plausible is the EPI's Magic Number?
Before going into dueling studies, consider the basic challenge facing us when we try to establish the impact of raising the minimum wage.
- The minimum wage increases
- There are changes in export demand
- There are changes in government spending
- There are changes in investment spending
- There are changes in the housing market
- There are changes in consumer tastes
Still, having a handle on these overall patterns is still useful, since it gives us an idea of the plausible magnitude of impact. So first, lets consider the simple correlation.
There are a number of ways to test for the direction of this relationship. Let's pick one that does not rely on strong assumptions about the nature of the relationship. Suppose that employers react to changes in the contract wage rate. Then the years with minimum wage increases should be different from the years with no increases. Therefore, I take year to year change in June employment for workers 16 and over, and for the subgroups 25 and over, 20-24, and 16-19 (these are standard Bureau of Labor Statistics breakdowns, from Table A-6 Historical Data). To pick out "higher than normal" and "lower than normal" employment growth, I assign an index of of 1 to the top quarter, an index of -1 to the bottom quarter, and an index of 0 to the two quarters in the middle.
Then I add up the indices in the years that had an increase in the minimim wage rate. The results were:
- -1 for all workers, 16 and over
- -1 for workers 16-19
- +3 for workers 20-24
- 0 for workers 25 and over
Now, two of these results have the sign that a traditional marginalist economist expects, but the -1 out of a maximum of -14 is too close to 0 for comfort, and two are not what a traditional marginalist economist would expect. (Note that I could test the statistical significance of each result, if you are really interested.)
However, at this point, a traditional marginalist economist will object that what is important is not the contract wage rate, but the wage rate corrected for purchasing power. And when we use some form of purchasing power index, we leave the situation of 23 years with minimum wage increases and 34 years with steady minimum wage rates. Instead we have a much richer set of data, with the minimum wage rate sliding down at different rates for 34 years and rising at different rates for 23 years.
So, lets makes use of that extra richness. Now I'm looking at change in the minimum wage after its been adjusted for consumer price inflation (using the CPI), and assigning the same type of index -- +1 for minimum wage increase in the top quarter of the change, -1 for minimum wage decreases in the bottom quarter of the change, and 0 for the "normal" changes between the extremes.
Then I multiply the two indices, and sum the result. If things go as the traditional economist expects, opposites will attract, so the product will be -1, while if they go against expectation, indices of a feather will flock togther, so the product will be 1. And the results are:
- +3 for all workers, 16 and over
- +2 for workers 16-19
- +2 for workers 20-24
- -2 for workers 25 and over
Now, that is a pattern that can't be a significant negative effect, because for overall employment and for youth employment, the sign is wrong. What is the objection now?
Oh, this is the post hoc ergo propter hoc fallacy ... that is, "it follows so its caused by" (oddly enough, traditional marginalist economists remember this fallacy far more often when things go against their expectations) ... it could easily be the case that the increase in employment happened in those years in spite of the increase in the minimum wage ... and that without the change in the minimum wage, there "would have been" greater growth in employment.
So, for example, we cannot combine the fact that the Bush "recovery" has the weakest job growth of any recovery since World War II, and the fact that inflation has been reducing the purchasing power of the minimum wage every single year of the Bush administration ... to prove that a slipping minimum wage (in terms of purchasing power) is bad for employment.
This form of evidence is suggestive, and for some people it will be persuasive, but it will never be conclusive. It is simply too easy to construct an alternate story.
For example, minimum wage increases are more likely to be supported by Democrats, and it is well known that the economy tends to do better under Democrats ... so, the argument would go, employment grows during minimum wage increases because if the employment benefits of other policies motivated by concern for working Americans, which turns out to be stronger, under the argument, than the negative impact of minimum wage increases.
Has anyone tried to correct for these problems?
And this takes us to the article by Card and Krueger (1994) (full citations in References), published in the leading journal of traditional marginalist economics in the US, the American Economic Review. This is the article that eventually raised such a stir that one of our discussants elected to debunk it in advance, in the discussion during Chapter 1, Part 1. Card and Krueger (1994) perform an analysis of changes in employment in the fast food industry in counties on both sides of the Pennsylvania / New Jersey border, in the period when New Jersey raised their state minimum wage above the nationwide minimum.
Why study fast food? If we think about it, the fast food industry is one that can fine tune its employment. It can accept slightly longer lines on average - with a slight competitive disadvantage - by squeezing employment down slightly, or pursue slightly shorter lines by boosting employment slightly. If there is no clear negative impact on employment even in the fast food industry, then there is no reason to predict a negative impact across the board.
Why study two states side by side? If we study two states, before and after one of them changes the minimum wage, we are going to be "holding constant" a large number of economy-wide variables. The really big problem with the analysis above is it looked at different rates of employment in different years, with different levels of exports, government spending, investment, consumer confidence, construction activity, and everything else that can influence employment. Studying two states that have different minimum wage rates at the same point in time eliminates that variability.
Why study adjoining counties across a state border? If we control for variability due to different points in time, there is still variability in different parts of the country as to how they react to economic changes. Looking at a cluster of neighbouring counties minimizes this variability as much as it is practical to do.
What Card and Kruger did was to conduct telephone surveys with establishments in three fast food chains - Burger King, Roy Roger's and Wendy's - both before and after New Jersey raised its minimum wage from the national minimum wage of $4.25 to a state minimum wage of $5.05. In effect, New Jersey followed on the two national minimum wage rises in 1990 and 1991 with a third in 1992 that applied to New Jersey alone. Since some locations in New Jersey were paying more than the previous minimum wage when the $5.05 rate went into effect, Card and Krueger were also able to compare the impacts when stores had to increase wages by the full $0.80 and when they were faced with a smaller increase.
Card and Krueger did not find a drop in employment as a result of an increase in the minimum wage. Instead, as discussed in Neumark and Wascheran (2000), they found an increase ... elastities ranging from +0.29 to +0.51 (where "+1" would be a 1% increase in employment from a 1% increase in wages).
So where do "E"PI get their magic number?
And this takes us back to the publication that the "E"PI used to obtain their magic number for Labor Elasticity. The source for the "E"PI magic number is a "comment" on Card and Kruger (1994). In the reply, Neumark and Wascheran (2000) look at the same question, in much the same way. However, as they argue, they use payroll data instead of telephone survey data, and they arrive at a substantially different conclusion. They conclude that the elasticity is -0.21, or a drop of employment by 0.21% for every 1% rise in the minimum wage.
Of course, you may wonder how Neumark and Wascheran (2000) got their hands on payroll data. And that is covered in the author details and shout out footnote on the first page of the reply:
... We are grateful to Carlos Bonilla of the Employment Policies Institute (EPI) and to participating franchise owners and corporations, for providing us with the payroll data. The EPI is funded by business contributions and generally opposes minimum wage increases. However, the analysis of the data described in this paper was conducted independently of the EPI and neither author received any renumeration for conducting this research. ... (2000: 1362)
Cut, Print, That's a Wrap
Of course, if you followed the preview of the Neumark and Wascheran (2000) comment on the original refereed (1994) article, you also know that there is a set of talking points to "debunk" Card and Krueger (1994), based on the way they went about doing their survey and the way they analyzed their data.
What you would have missed is the fact that Neumark and Wascheran (2000) is not a refereed article, but merely a comment. And when a journal editor elects to publish a comment on a refereed article, the original authors have the right of reply - normally in the same issue of the journal, sometimes in the following issue.
And, by jove, don't you know, but there is it, Card and Krueger (2000). And this "reply" showed that the negative elasticity is based on flawed data. First the "reply" repeats the original analysis with better data, from the Bureau of Labor Statistics, and the results of the analysis are mostly positive values, but not high enough to be statistically significant. Then the "reply" looked at why the results from payroll data vary from results based on data reported by firms to the Bureau of Labor Statistics.
And it turns out that the subset of fast food restaurants in Pennsylvania used in the "comment" article is significantly different from all fast food restaurants in that part of Pennsylvania. In other words, the cross border comparison relies on a false picture of what is happening in the lower wage state, and that is the source of the negative elasticity.
In particular, it is data from a single Burger King chain in Pennsylvania that provides the statistical significance for the study that gives the negative labor elasticity. It is therefore likely that the employment growth that Neumark and Wascher (2000) attribute to a lower wage is in fact due to the fact that their Pennsylvania data was biased by data from an operation that was growing its share of the fast food market at the time.
What does this all mean?
Card and Krueger were so excited by their results, which confirmed the results of a number of similar cross sectional studies, that they wrote a book. You can see them getting up to speed in the Card and Krueger (1994) when they present their theories as to how such a contrary result is possible.
On the other hand, in an economy where demand-constraints can exist on the amount of employees you can hire, there is no need to be surprised. In that situation, employers will generally be able to push wages toward the bottom of the trading range, due to slack "markets" for unskilled labor. Therefore, much of the wage increase will be pushing back up through the trading range. With that reduction in the "price effect" from a wage increase, there does not need to be much of an "income effect" to push the overall "labor elasticity" into the positive range.
I must stress that Card and Krueger would definitely not agree with that view. I would not be surprised if they labelled that view as being old-fashioned, and long since debunked. There are, after all, advantages in treating a theory as old-fashioned and long since debunked. That means that you never have to actually go through the hard work of actually constructing an argument against them. And it works equally well even if their debunking was as shaky as the Neumark and Wascher "comment" on Card and Krueger's conclusions.
Next Chapter ...
I can't promise to put up a part of a Chapter each week, but I'm going to try to do that for Chapter 2 at least. Chapter 2 is Money ... where it comes from and where it goes.
Card, David and Alan Krueger. 1994. "Minimum wages and employment: A case study of the fast-food industry in New Jersey and Pennsylvania." The American Economic Review. Nashville: Sep 1994. Vol.84, Iss. 4; pg. 772-93.
Card, David and Alan Krueger. 2000. "Minimum wages and employment: A case study of the fast-food industry in New Jersey and Pennsylvania: Reply." The American Economic Review. Nashville: Dec 2000. Vol.90, Iss. 5; pg.1397-420.
Neumark, David and William Wascher. 2000. "Minimum wages and employment: A case study of the fast-food industry in New Jersey and Pennsylvania: Comment." The American Economic Review. Nashville: Dec 2000. Vol.90, Iss. 5; pg. 1362-1396.
Official Data: Bureau of Labor Statistics Historical Time Series.
- Economic Policies Institute (the "real" EPI).
- Center for Full Employment and Price Stability (CFEPS), Kansas City, Missouri.
- Center of Full Employment and Equity (CoFFEE), Newcastle, New South Wales.
Industry propaganda mill to help the fight to retain windfall gains: Employment Policy Institute, denoted in diary as "E"PI..