national membership count issues

[Joe Dehn]

As you all probably know, I have been monitoring membership statistics and posting graphs of them for many years.  Specifically, I regularly post a graph showing numbers relating to both state and national membership in California over the past several years.  Normally I try to do this once per month, because the numbers come from reports that are published monthly by both the national and state party organizations.  But because one of these reports is often delayed, sometimes I post a graph updated with just some numbers, and then another when both reports have come out.

The most recent one of these that I posted, both here and on Facebook, this past Tuesday, looked like this:

The lower two lines show two categories of state membership, just "central committee members", people who are qualified to vote on state/county LP business, and the larger "total" membership which includes "associate members", people who have contributed financially but don't qualify for central committee member status for one reason or another, e.g., they haven't signed the certification regarding initiation of force or they are not registered Libertarian. The last numbers for those lines in this graph came from the state membership report for the end of October.

But the discussion which follows relates to the other two lines.  The dark purple line represents the number of "sustaining" national members with addresses in California. In this context, "sustaining" means they contributed at least $25 to the national LP in the past year, or had at some time in the past paid an amount that would qualify them as a "life member". To be included in this count they also should have signed the certification. There have always been some people who paid "dues" or made other financial contributions to the national party but who had not signed the certification, and those people have always been included in the national database, and for decades a count of those people was included in the monthly membership reports, but those reports  were changed several years ago to not report that number (and some other numbers as well, such as the total number of people who had signed the certification whether or not they had paid any money recently). That's why there are two lines in my graph showing counts for the state level but only one for national -- because the national report now only gives us one number, "sustaining members", people who both signed and gave money.

The fourth line on the graph does relate to national membership, but it's not a count.  It show the fraction of all national "sustaining members" who have addresses in California.  This fraction (or actually a very slightly different number) is used in calculating things like how many delegates our state gets to send to national conventions and whether a group of states (or in our case, one state) is big enough to have a representative on the LNC.  As shown by the graph (looking at the scale on the right), this fraction has been roughly around 10% for the entire period shown. Also significant -- it is the value of this fraction at a very specific time, relative to each national convention, that is used in these calculations. And so lots of people pay special attention to the membership statistics at those times, and sometimes make special efforts to boost their state's membership count leading up to it..

Meanwhile, some time after our current state EC took office, our state chair started receiving from the national party files labeled as containing lists of "active" national members in California, and then he started sharing those files with our state Membership Committee (of which I am currently a member).  Such information could be useful to us in many ways, but right here there is one in particular that is relevant. I don't know exactly when they started sending us this kind of membership information in this particular format -- possibly they have been doing so for years and I just didn't know about it because I wasn't on the Membership Committee, or perhaps the previous state chair wasn't sharing them with anybody. But once Loren started forwarding them to us, and I noticed that they seemed to be produced on a weekly basis, I got curious about what exactly "active" meant in this context. Was this actually a list of the very same people who are counted as "sustaining members" in the monthly reports?  So I started comparing the two series of numbers and pretty quickly became convinced that they were in fact just two labels for the same set of people. Which meant that the count of the records in each of these files could be used just as well to see how we were doing -- and would give a count for the end of any given month, or close to it, significantly before we received the formal monthly report (which sometimes has been close to the end of the following month).

And so that's what you are actually seeing on the graph above as the right-most value of the purple line.  Not the number from any formal "report",. but the number of records that I found in the most recent "active" file that they sent, which just happened to be on 31 October. (And that's also why the light purple line, the one representing our fraction, ends one month earlier -- I have no corresponding nationwide count from which to calculate the fraction.)

To illustrate why it would (normally) make sense to use this number, the count from a data file, in this way, as a substitute for a number taken from a "membership report", here is a graph of numbers for various dates over the past couple of months, two of which came from the monthly "reports" and the rest from the "files". Without checking the X-axis to see which are exactly at the end of a month, you would not have any way of telling which are which.  They look like they are a single series -- slowly but steadily declining with a slight uptick at the end -- because I believe they are in fact a single series..

However, 31 October was not a "normal" date. It was one of those magical "end of the 7th month before the national convention" dates that are of special concern to so many people. And I know from past experience that people tend to do things a bit differently around such dates. State activists are pushing people to join or renew, a greater than usual number of payments are coming in at the last minute, leading perhaps to some transactions not being entered into the database until later but still treated as having met the deadline, and state activists and national staff are making extra efforts to correct errors that might affect some state's count. So I figured that the "official" count for 31 October, the one that would actually end up being used for delegate calculations, would very likely differ from what I was seeing in that 31 October file.

And so, when I originally posted the first graph above, I specifically noted that possibility, that the "official" number might be different.  I didn't have any way of knowing how different, but I was pretty sure it would be higher. And so, while I and other activists all across the country were waiting to see that official report for the end of October (which still has not yet been published), I was also eagerly awaiting the next weekly data file, which I expected would have at least a slightly larger number of records than the previous one.

So I was pretty shocked when I got a look at that file, dated 7 November, and discovered that it had about 100 fewer records than the one dated 31 October.  I was expecting the count to go up a least a bit, due to a last minute "push" for people to pay dues by the deadline, and perhaps some delay in processing last minute transactions (or simply people paying online just before midnight, after that file had been generated, during the day). But instead, it appeared that we had actually lost about 10% of our members, right at the time of the critical deadline!.

At first I believed that the most likely explanation was that there were so many people who had joined just before the deadline for the previous convention, for which the deadline had also been 31 October, that a drop now would be a natural consequence. Some of the people who joined then would not have renewed a year later, but of those who did some of them would have not renewed now -- and if there had been enough of them originally the number failing to renew now could easily exceed the number of new people being added now, which is, unfortunately, not many.

But as I looked more carefully at these files in an attempt to confirm this theory, what I found made no sense. Yes, there was a bump around the end of October in people due to lapse, but only a small one, nowhere near enough to account for the 10% drop in the count. Somehow a much larger number of people had disappeared from our set of "active" national members even though their memberships were nowhere near expiring.  So I decided to do a full comparison of the content of the two files,the one from 31 October and the one from 7 November.

I found that there were 8 people in the new file who were not in the old file. That seemed reasonable -- they were new members, or people whose memberships had lapsed a bit earlier and then decided to renew just before the deadline, or perhaps they had just changed their address from some other state and so were now appearing in our file rather than some other state's.

But what I found in the other direction made no sense at all. There were 107 people in the older file who were not in the newer file. Most of these people had, in the old file, expiration dates significantly in the future. Many of them had no "expiration date" at all because they were life members.  And one of those life members who had disappeared from the file was me! 

Looking over that set of 107 records I could not discern any pattern. They were not at the beginning or the end of the file, they had start dates ranging all the way from the 1970s to this year, they had addresses all over the state, they didn't have any particular flags in common, their names were not from one end of the alphabet.  Just a lot of people missing for no obvious reason, with nothing in common that I could see -- based on the information in the file.  

But something was clearly very wrong, either with the national database itself, or how they had generated this file compared with the previous one. So we reported the problem to the national staff.

It didn't take them long to respond. First they said that they could account for a difference of 34, because they had made a change in how they counted people who had paid dues but for whom they had no record of a "pledge" (by which they meant the membership certification).  They kindly even provided a list of those 34 people. Huh?  That's a pretty basic element of the counting -- why did it change, now of all times?  And they couldn't explain the rest, suggesting that it was just a lot of people who had memberships expiring at that time, which I already knew was not the case.

So I went through the two lists, the list of 107 that I had found missing and the 34 they said they could explain, discovered that most of the rest were life members, and reported that back.  But in the meantime apparently they had figured that out themselves,and their new answer was that the the same change in accounting for the "pledge" had affected the counting of life members.

Their "fix" for this was to assume that all life members had signed the "pledge", and they sent us back a new file with all those life members added back in. But not the non-life members. For those people they made the opposite assumption -- that they had not signed the pledge!  But they said they were willing to update that information if we could tell them that a particular person had signed.

So that's where things stand right now. Our "active" / "sustaining" national member count is, as of today (we got another file today), about 4% lower than it was two weeks ago -- not because we did anything, not because any members in California did anything, not because any members in California failed to renew -- but because the national LP changed the way their system counts people for whom they don't have a record of the "pledge", and then further (after we pointed out the result) in effect reversed that change, but only for life members.

Note that none of this is likely to threaten our 10% status, because presumably the same factors had similar effects on the counts for other states.  And meanwhile Mimi has been working on getting the records for non-life members in California corrected, in cases where we have our own information about them having signed the certification. So I expect that by the time they actually get to doing the delegate allocation calculation, our count will be back up close to what it would have been if none of this had happened.

But this has some pretty disturbing implications otherwise. This means that all of the numbers they were reporting previously, for some number of years, were wrong  because they were doing the calculation wrong. Further, it appears that the reason they were doing it wrong was that they couldn't do it correctly because the database did not contain the correct information. And they knew about this, but didn't address the issue adequately until now. Maybe they wanted to and just didn't have the resources. Presumably they are only scrambling to fix it now because it's time for that delegate allocation process. (And what did they do, or try to do, about this two years ago?)  But they can't actually fix it properly because they lost track of the necessary data!

By "they" I don't mean the people currently working on the database. The current staff apparently have known about this for some time but they may have had nothing to do with the data being lost. They seem to believe it happened during the switch to "CiviCRM", and I don't remember who was in charge of what at that point.

This database has existed, in one form or another, for almost 50 years. These same two pieces of information -- who has "signed" and who has paid money -- have been relevant to our definition of membership for basically the whole history of the party.  Generations of activists kept track of this information over the decades as it moved from index cards to minicomputers to personal computers to "the cloud". Until somebody, either through incompetence or neglect, threw it away, when they moved everything to "Civi". We aren't using that system anymore, but that doesn't help --the information was already lost, so they don't have it available now, and that's why they have been forced into fudging things.

What does this mean for the arguably much more important question of how we are doing with recruiting and retaining members?  Well, it certainly makes it harder to see how we are doing, if the real meaning of the reported numbers is in flux! But making some reasonable guesses as to what the numbers "would have been", I think it's obvious that we have not been doing well. There is nothing in these numbers, even accounting for this fudging, to suggest that our membership is or has been growing. Considering the past several years, it's still pretty clear that we have been losing more members than we have been adding. Maybe we have now hit bottom? And hopefully with this correction -- or rather these corrections -- the reported numbers going forward will start being comparable to each other again.

--

Joe Dehn

Northern Area Coordinator

Libertarian Party of California

n...@ca.lp.org

[Joe Dehn]

On 2025-11-15 02:11, Joe Dehn wrote:

As you all probably know, I have been monitoring membership statistics and posting graphs of them for many years.  Specifically, I regularly post a graph showing numbers relating to both state and national membership in California over the past several years.  Normally I try to do this once per month, because the numbers come from reports that are published monthly by both the national and state party organizations.  But because one of these reports is often delayed, sometimes I post a graph updated with just some numbers, and then another when both reports have come out.

 

The most recent one of these that I posted, both here and on Facebook, this past Tuesday, looked like this:

 

A question about this graph came up during the meeting today.  Apparently there was some confusion about the meaning of the labels along the x-axis.  Those are not labels of every date for which there is a y-value -- they are just yearly labels spread across the entire time span.  The last values for the three lines that represent counts are for the end of October, not June. (The last value for the light purple line is for the end of September, because no nationwide count is available for the end of October from which to compute such a percentage.)