Union Membership and Wages in the US over Time


Gold Koala
Quinn Bello, Diana Flores Valdivia, Lily Pan, Henning Schade


The data comes from the Union Membership, Coverage, and Earnings from the CPS by Barry Hirsch (Georgia State University), David Macpherson (Trinity University), and William Even (Miami University). It contains three separate data tables: demographics, which has information about how many workers are in unions vs not in unions in the U.S., wages, which tells us about the pay of union and non-union workers, and states, which conveys union data by state. All of the variables are calculated annually, using the data collected by the monthly Current Population Survey. To understand trends in union membership and earnings across the U.S., we focus on two of these tables for our analysis: ‘states’ and ‘wages’.

We utilized 5 variables from the ‘states’ dataframe: ‘year’, ‘state_abbreviation’, ‘members’, ‘employment’, and ‘sector’. The first two encode the year the survey data was collected and the postal abbreviation for the state the data pertains to. Each economic sector of the state is encoded as ‘sector’: public, private, and even more specific industries within the public and private sectors. The ‘members’ variable is a count of union members within that sector in each state, while the ‘employment’ variable is a count of all workers within that same category, union and non-union. The ‘wages’ dataframe also includes ‘year’ and ‘sector’ variables, as well as ‘union_wage’ and ‘nonunion_wage’. Each of the latter contains the average wage made by a union worker and a non-union worker, respectively, for each year and sector. As a whole, these variables will allow us to explore geographic and industrial trends in union membership, as well as the trends in one of the issues unions on which unions tend to campaign: wages.

Question 1: How has union membership changed over time in the United States?


The evolution of union membership shows the dynamics of labor and society in the United States. In asking this question, we want to explore patterns of union participation over the years, nationally and within each state. In order to do this, our analysis will draw from the states.csv dataset and employ features such as ‘state’, ‘members’, ‘year’, and ‘sector’. These features will allow us to capture specific details on membership variation based on state and sector over the years.

Through this analysis, we hope to get a sense of any changes in union membership over the last five decades. Hopefully, this will grant us some understanding of the strength of labor organization in the US and any trends that might have contributed to the recent rise in support for unions and labor action over the last couple of years. In addition, we hope to be able to see any regional differences: are unions growing in one region and shrinking in others, or are all states seeing similar trends? This information could be helpful to any workers hoping to join or organize a labor union in their industry, by giving them a sense of how strong organization might be in their home state.


To address this question we construct two visualizations, each exploring a different aspect of the change in union membership in industries over time. First, we will build a time series displaying the trend in union membership as a percentage of overall employment in each of the 50 states and the District of Columbia, utilizing the geofacet package to display 51 geographically-arranged line graphs. This plot allows us to clearly visualize the trends in each individual state using the intuitive form of a line. Displaying these plots in a roughly geographical arrangement using geofacet and color-coding the overall change in membership can additionally give us an idea of regional trends.

For the second plot, we will compare the trends in union membership across different U.S. economic sectors using a faceted bar graph. Using wide bins of 10 years will allow us to visualize trends in union membership nationwide while reducing the visual noise of graphing 50 individual years. Bar graphs utilizing the same scales are also easy to compare across multiple facets, allowing us to see the differences between several sectors without placing them all on the same plot.



The first plot, displaying trends in union membership in all 50 states (plus the District of Columbia), reveals that membership has been decreasing since 1983. Every single region analyzed experienced a decline in the percentage of workers belonging to a union. Reductions in union membership also seem to be concentrated in the Midwest, with Indiana, Wisconsin, and Michigan indicating a drop of nearly 20%. As all three of these states are generally considered manufacturing-heavy, this may indicate that union membership in the manufacturing sector is experiencing a downturn.

The second plot seems to uphold this theory: private manufacturing displays a significant decrease in union membership over the past 50 years. There is a general decrease displayed in the private sector as a whole, as well. Meanwhile, public-sector employees actually seem to be joining unions in increasing numbers. It is possible that unionization is easier, or unions are stronger in general, in public industries; this would explain the increase in public-sector union membership and the decrease in private-sector union membership. This same fact, combined with the overall decline of manufacturing in the United States, could also be responsible for the rapid decline in union membership for manufacturing workers.

Question 2: How have union and non-union wages changed over time?


This question aims to encourage discussion around how the wages of union and non-union employees differ and how they have changed over time. Additionally, how have these differences evolved across industries? To answer this question, we need the ‘year’, ‘non_union_wage’, ‘union_wage’ and ‘facet’ variables from the wages dataset. One potential limitation of our data is an expansion in the definition of ‘union’ in 1977. This may cause a visible shift in the trend chronologically early in the data, which might be interesting to explore.

We are interested in addressing how the potential effect of unions has endured over the years - it is general knowledge that unions used to be stronger in past decades, but are they still as strong today? Might any change have affected the benefits workers see? Also, are unions stronger across certain industries, and are they even detrimental in some areas? As support for unions increases, we want to investigate how unions might claim to impact workers’ compensation.


We are going to create two visualizations: an area chart that displays union vs non-union wages over time for all wage and salary workers, and a faceted line chart that shows union vs non-union wages across different sectors. The area chart provides a clear comparison between union and non-union wages, highlighting the trends and changes over time. The use of an area chart allows us to visualize the magnitude of wages and the difference between union and non-union wages in a way that is easy to interpret.

The faceted line chart, on the other hand, offers a detailed view of how union and non-union wages vary across different sectors and industries. By faceting by sector, we can see the trends of the gap between union and non-union compensation in different work environments. The line chart is a straightforward choice for this visualization, as it effectively communicates changes in wages over time without overwhelming the viewer, even when displaying multiple facets.



The overlaid area chart shows that wages overall have risen over time (although this does not account for inflation) and union wages have generally been higher than non-union wages. However, in the last few years, non-union wages have been rising more quickly than union wages to the point that the two were almost equal in 2022. This could be due to the recent “labor shortage” as workers have more bargaining power and the free market can increase worker wages without the help of a union, and unions are focused on gaining more incremental increases in wages.

The faceted line chart is more interesting. We can see that the difference between union and non-union wages does not stay constant across industries. For example, in construction, union wages have been on average much higher than non-union wages, and in some industries such as the public federal sector and manufacturing, wages have historically been equal but recently non-union wages have been higher. In the rest of the industries, union wages are slightly higher than non-union wages, closer to the average. These patterns could be due to workers in public federal sector/manufacturing having more bargaining power, or that unions are stronger in construction and other industries.


Our presentation can be found here.


Harmon, J. (2023). Union Membership in the United States [Data set]. TidyTuesday. https://github.com/rfordatascience/tidytuesday/tree/master/data/2023/2023-09-05. Accessed 2/9/2024.


Tidy Tuesday datasets