When the Numbers Tell Two Stories

When the Numbers Tell Two Stories

G. Monray on Spain’s minimum wage, unemployment, regional differences and the statistical paradox that can change how we interpret labour markets

Few economic questions generate as much debate as the relationship between the minimum wage and employment. Conventional economic models suggest that if a legally imposed wage rises above the market-clearing level, employers may demand less labour. Yet empirical research has repeatedly produced more nuanced results, with effects varying according to labour-market institutions, sectors, worker characteristics and the size of the wage increase.

In their study “Minimum Wage and Effects on Unemployment: The Case of Spain and Its Implications on Simpson’s Paradox and Geographical Mobility,” Jorge Monray and Juan Morillo examined the Spanish experience using longitudinal data covering 2010–2023. The research divided the evidence by gender, age group and Spain’s Autonomous Communities, using a dataset of 3,094 observations and applying correlation analysis and one-way ANOVA.

One of the study's most intriguing findings was that the relationship between minimum-wage variables and unemployment was generally negative rather than positive in the groups examined, with Pearson correlations between approximately -0.4 and -0.6 in most categories. The authors also found a notable reduction in unemployment among young males despite repeated increases in Spain's minimum wage.

But the most important methodological question went beyond the sign of the correlation. When the researchers compared aggregated data with disaggregated data, some relationships moved in different directions. This raised the possibility that Simpson’s paradox could affect how aggregate labour-market statistics are interpreted.

An interview with G. Monray

Q: What motivated you and Juan Morillo to examine the minimum wage in Spain?

G. Monray: The starting point was an apparent contradiction. The traditional competitive labour-market model suggests that, under certain conditions, increasing the minimum wage can reduce the demand for labour and therefore increase unemployment. Yet when we looked at Spain over an extended period, the empirical picture was not that straightforward.

Spain offered an interesting laboratory because the minimum wage had been increased repeatedly while unemployment was simultaneously influenced by enormous changes in the Spanish economy. We therefore wanted to examine the relationship longitudinally rather than looking at a single year or a single policy change.

Q: What makes the Spanish case particularly interesting?

G. Monray. Spain is not a homogeneous labour market. There are substantial differences between Autonomous Communities in economic structure, productivity, industrial composition, unemployment, wages, demographics and geographical mobility. There are also important differences between groups of workers. A young male worker in one region may face a completely different labour-market environment from an older female worker in another region.

That led us to a fundamental methodological question: what happens when we move from the national aggregate to the individual categories underneath it?

Q: How did you structure the research?

G. MonrayWe constructed a longitudinal analysis covering 2010 to 2023, dividing the information by gender, age and Autonomous Community. The dataset contained 3,094 observations across the 17 Autonomous Communities and the different demographic categories. We also adjusted the minimum-wage data using mean and mode calculations.

We then used Pearson correlation analysis to examine the relationship between minimum-wage variables and unemployment, followed by one-way ANOVA to examine differences across groups. The purpose was not simply to produce one national coefficient. It was to see whether the relationship changed when we looked inside the aggregate.

Q: And what did the correlation analysis show?

G. Monray.The results were striking because the correlations were generally inverse. In most of the groups examined, Pearson coefficients fell between approximately -0.4 and -0.6. In other words, within the observed data, increases in the minimum wage were associated with reductions rather than increases in unemployment. That is obviously interesting when compared with the simplest version of the traditional theoretical expectation. But I would emphasize the word associated. A correlation does not establish that increasing the minimum wage caused unemployment to fall.

Q: Why is that distinction particularly important in this research?

G. Monray. Because unemployment is determined by an enormous number of variables. Economic growth, inflation, interest rates, investment, demographic changes, migration, sectoral composition, productivity, labour-force participation and institutional factors can all influence employment. If the minimum wage rises during a period in which the economy is also expanding strongly, we cannot simply attribute the subsequent fall in unemployment to the minimum wage.

The research was therefore about identifying patterns in the data, not claiming a simple causal mechanism.

Q: One of your findings concerned young men. What did you observe?

G. Monray. The study found evidence of unemployment reduction particularly among young males, despite the repeated increases in Spain's minimum wage during the period examined. This is interesting because young workers are often considered one of the groups potentially most exposed to minimum-wage effects.

One possible explanation discussed in the paper concerns labour mobility and changes in the opportunity cost of employment. But these are mechanisms to be considered, not causal conclusions demonstrated by the correlation analysis.

Q: This brings us to geographical mobility. Why did you include it?

G. Monray. Because labour markets are geographical. Suppose wages increase in one location but employment opportunities increase somewhere else. Workers who are sufficiently mobile can respond by moving. That means the observed unemployment rate in a particular region may reflect not only what employers and employees are doing locally, but also who enters or leaves that regional labour market. Geographical mobility can therefore act as an adjustment mechanism.

Q: Did the research find evidence that geography matters?

G. Monray. Yes. The ANOVA analysis indicated a relevant estimated effect size (eta) when comparing Autonomous Communities and their influence on unemployment among people aged 55 and over. That was important because it demonstrated that regional differences should not simply be treated as statistical noise. The Spanish labour market contains substantial geographical heterogeneity, and the effect of economic variables can differ considerably depending on where people live and work.

Q: How does Simpson’s paradox enter the story?

G. Monray. This is probably the most intellectually interesting part of the research.Simpson’s paradox occurs when a relationship observed in several separate groups changes or even reverses when those groups are aggregated.

Imagine, for example, that the relationship between two variables is negative in several individual regions, but when all the regions are combined, the aggregate relationship becomes positive. The aggregate number can therefore tell a different story from the underlying groups. Our results showed differences between aggregated and disaggregated data that were consistent with the possibility of Simpson’s paradox.

Q: Why should economists and policymakers care about that statistical issue?

G. Monray. Because policy is frequently discussed using national averages. A minister, journalist or economist might say: unemployment increased by X percent while the minimum wage increased by Y percent. But that national average may conceal completely different experiences among young and older workers, men and women, or workers in different regions. The danger is that we may attribute an observed aggregate relationship to the minimum wage when the underlying composition of the labour market is actually driving part of the result.

The broader lesson is simple: before interpreting an economic relationship, look underneath the average.

Q: Does your research challenge the traditional economic theory of minimum wages?

G. Monray. I would not describe it that way. The study does not demonstrate that traditional labour economics is wrong. Instead, it shows that the relationship between minimum wages and unemployment in Spain is more complicated than a simple one-variable explanation. Economic theory provides mechanisms that may operate under particular assumptions. Empirical labour markets contain many additional mechanisms.

For example, employers may respond through prices, productivity, hours, margins, technology, hiring standards or changes in the composition of their workforce. Employees may respond through participation, mobility or changes in job-search behaviour.

The actual outcome can therefore be considerably more complex than a textbook supply-and-demand diagram suggests.

Q: Did the paper find that increasing the minimum wage reduces unemployment?

G. Monray. That would be too strong a statement. What the study found was an inverse statistical relationship between minimum-wage variables and unemployment in most of the groups analyzed. It also found unemployment reductions in particular categories, notably young males. But the research was observational and longitudinal rather than a randomized experiment or a quasi-experimental identification strategy capable of establishing a clean causal effect.

Therefore, the appropriate conclusion is that the Spanish data do not support a simple assumption that minimum-wage increases necessarily produce higher unemployment across all groups.

Q: What does this tell us about averages in economics more generally?

G. Monray. It tells us to be careful with aggregation. Averages are extremely useful. They allow us to summarize enormous quantities of information. But they can also conceal heterogeneity. In economics, the average worker does not necessarily exist. The average region does not necessarily exist. And the average response to a policy may conceal several very different responses underneath it.

This is why segmentation by age, gender, geography and economic activity can materially change the interpretation.

Q: What was the broader contribution you wanted the study to make?

G. Monray. The contribution was methodological as much as empirical. We wanted to demonstrate that the Spanish minimum-wage question should not be reduced to a simple statement such as “higher minimum wage equals higher unemployment” or the opposite statement “higher minimum wage equals lower unemployment.” Both are potentially too simplistic.

The evidence suggests a labour market in which different groups and regions can behave differently, and where aggregation can obscure those differences.

For me, the most interesting lesson is that the way we organize economic data can influence the economic story we think the data are telling us.

Q: So what should a reader ultimately take away from the research?

G. Monray. The first lesson is that minimum-wage policy should be evaluated with disaggregated evidence. The second is that correlation should not automatically be interpreted as causation. And the third is that geographical and demographic composition matters. The Spanish labour market is not one market in the analytical sense. It is a collection of regional and demographic labour markets interacting with each other. Once you recognize that, apparently contradictory results become much easier to investigate.

The significance of the Monray–Morillo research extends beyond the Spanish minimum-wage debate.

Its deeper message is methodological: economic aggregates can conceal economically important differences between groups.

The study found inverse correlations between minimum-wage variables and unemployment in many of the categories examined, including Pearson coefficients around -0.4 to -0.6 in most groups. It also identified a significant regional effect in the unemployment of people aged 55 and over.

But these findings should not be interpreted as proof that minimum-wage increases cause unemployment to fall. The research identifies associations in observational longitudinal data, while unemployment is influenced by many other factors.

The reference to Simpson’s paradox is therefore particularly valuable. It encourages economists, policymakers and business researchers to ask a deceptively simple question before drawing conclusions from a national statistic:

What happens when we break the average apart?

In an increasingly regionalized and heterogeneous labour market, that question can be as important as the headline number itself.

Publication

Monray, J., & Morillo, J. (2024/2025 publication issue). *Minimum Wage and Effects on Unemployment: The Case of Spain and Its Implications on Simpson’s Paradox and Geographical Mobility.* International Journal of Economics and Finance, 17(2), 26–44. DOI: 10.5539/ijef.v17n2p26. The article was received October 27, 2024, accepted December 18, 2024, and published online December 25, 2024; the journal volume is 17(2), 2025.

Historical and methodological note: The analysis covers Spanish data from 2010–2023 and therefore should not be presented as a measurement of the Spanish labour market in 2026. Its statistical associations should likewise not be interpreted as causal estimates of the effect of minimum-wage increases.