STUMP » Articles » Public Employee Unions and Pensions: Are There Connections? » 1 March 2018, 06:12

Here is the Geek-Out version of yesterday’s post.

via GIPHY

I questioned the connection between public pension benefits and unionization of public employees.

UNSURPRISINGLY: RESEARCH EXISTS

Before I did my skin-deep data slice, I thought to check out if the Center for Retirement Research had looked into this.

Of course, they did.

From July 2011 – Unions and Public Pension Benefits

The brief’s key findings are:

Recent debates suggest that union power has led to higher public pension benefits.

Interestingly, union strength appears to have no impact on the level or growth of benefits.

Because pensions are legislated, not bargained, lobbying expertise may be more important than union size.

In contrast, union strength does seem to raise employees’ wages.

Because wages are determined through collective bargaining, union numbers appear to matter.

Finally, union strength appears to reduce the relative size of the public workforce.

Given the effect on wages, reductions in employment would not be unexpected.

These results should be viewed as only a first step in understanding the influence of public unions on employee compensation.

The bolded line is what I said yesterday. That said, there is an interesting mix of findings there.

Let’s grab some graphs. And let me give you a nutshell of some of the stats.

Let’s start with figure 4, which gives unionization rates.

The private sector unionization peaked at about 35% in 1955…. and kind of just collapsed like a flan in a cupboard.

See?! I’m not just making this stuff up!

The 1962 jump is this: Executive Order 10988 by JFK, which allowed federal employees to collectively bargain.

And then public employee unionizations seems to have topped out around 40%, but just under. Mark that 40% in your mind, because we’ll be coming back to that.

The next figure I want to show you is Figure 5:

So, this is inflation-adjusted, but … well, anyway, I will commend them for starting their y-axis at 0. That’s appropriate here.

So finally, let’s look at those regression variables.

Slice 1:

I have lots of issues with this specific visualization. And those choices of… “variables”. Buy me a drink, and I’ll explain. This is my blog, so enjoy this otter:

via GIPHY

But I will give you a short version from the appendix:

I want to direct you to the R-squared statistic. The 0.11 result is really low — essentially, the “variable” choices barely explains the result.

So I’m just going to quote Dierdre McClosky at ths point:

Statistical significance is not the same thing as scientific importance or economic sense.

So that regression with a 0.11 r-squared may come back from analysis as “statistically significant”, but it doesn’t mean it’s meaningful.

This is a really important paper, by the way, if you think to use statistical analysis for making policy prescriptions.

Anyway, basically, union membership has little to do with pension benefit increases.

OTHER RESULTS

That’s not true of everything they tried.

So unionization boosted wages… and the fit?

The r-squared of 0.32 is actually pretty good for anything involving humans. So unionization does seem to lead to higher wages for public sector workers… at least when you measure it against the private sector.

This one is weird:

I’m not going to bother with the regression results here — the r-squared stat again is low (0.18) – but the message is that higher public unionization leads to lower levels of public employees compared to private employees. So public employee unions don’t necessarily lead to more public employees.

To be sure, this covers only a very limited number of years of data. But I thought it interesting.

PRETTY PICTURES OF STATE UNIONIZATION

In yesterday’s post, I linked to this piece which had unionization stats by state.

I’m linking this into my favorite new visualization: gird maps.

So here’s the 2014 public employee unionization rates.

Here’s 2016 Employer Normal Cost by % of payroll:

Hmmm. Okay.

Let me calculate a correlation coefficient for these series.

Normal cost & unionization rate? Correlation of 4%. That is the same as 0, in practical terms.

So, I’m not seeing much proof that unionization rate has much to do with normal cost.

Here’s a scatterplot:

So here’s the main thing: no, public employee unionization is not the driver behind pension benefits.

WHAT ABOUT AS PERCENT OF VOTING POPULATION?

Now one thing I didn’t measure here was the % of public sector employees v. voting population.

I found this spreadsheet.

So, for convenience, let’s pretend all public union members vote. (Yes, I’m using 2014 union #s against 2016 voting numbers… just flow with it.)

So here’s a map:

Correlation between union membership as % of vote & normal cost is also 4% – not too surprising, as this is not going to differ much from unionization rate for a state.

And I also looked at determining the number of all the public employees (unionized or not), and again, assuming 100% of them vote:

Hmmm, interesting — and the correlation is even weaker here. 1%. Again, essentially zero.

Now, I don’t have data for benefit increases to check these against – I’m just looking at normal cost. Of course, normal cost is affected by choice of method to calculate, but also the valuation assumptions that go into that.

The big problem in doing any of this analysis is this: the percentages are in a constrained range, for both normal cost and unionization rates. And there are only 50 states (plus 1 D.C.) – that’s not really a lot of data points for me to work with.

My only point is that I don’t assume that public employee power comes from the unions. It comes from there being public employees.