Wednesday, April 17, 2013

LDS church growth

Back in 2000, my seminary teacher quoted to my class the Rodney Stark projections for church growth, estimating that there would be 64 to 267 million members by 2080. Other researchers have predicted much lower numbers, around only 30 million.

Another sky-high projection was made by a Mormon recently, projecting that there could be 2.6 billion members by 2120. I'd like to consider whether those projections are realistic, but first let's distinguish between growth of the reported membership, and growth of either self-identified or active members.

The official reported church membership is much higher than the number of self-identified Mormons. I’ll just give a few examples:

Members claimed by church in Mexico, 1999 846,931
Self-identified members from census in Mexico, 2000 205,299
Members claimed by church in Chile, 2001 520,202
Self-identified members from census in Chile, 2002 103,735
Members claimed by church in US, 1990 4,175,000
Members projected by ARIS survey, 1990 2,487,000
Members claimed by church in US, 2008 5,974,041
Members projected by ARIS survey, 2008 3,158,000

Sources: here and here. Based on worldwide data, this researcher concludes "approximately 40% of individuals claimed as members by the LDS Church worldwide identify the Church as their faith of preference."

The comparison between the ARIS survey projections and the official membership is interesting because the growth rate implied by the official membership numbers from 1990-2008 is an impressive 30% (~1.47%/year, or 15.7%/decade). The ARIS projections only indicate growth of 16% (~0.83%/year, or 8.6%/decade), which is about the same as US population growth.

So if the church is overstating both total numbers and its growth rate, how can the true worldwide growth rate be estimated? Many observers think that the active members per congregation (meaning wards and branches) has been fairly consistent, so the true growth rate might be close to the growth rate of congregations.

This chart doesn't include the most recent data points, and unfortunately it's not my chart so I can't update it, so I'll list the most recent data for increase in congregations in this table:

Year:Total congregationsIncreaseIncrease as %

As a comparison, the US population growth rate is ~0.9%/year, and world population growth rate is ~1.15%/year. I do expect to see a temporary increase from the change in missionary age over the next 2-3 years, and possibly a slight long-term increase as well. Based on this data, I'll make a few conclusions:
  • The church's growth rate is similar to background population growth.
  • The conclusions drawn by David Stewart, based on a lot of data, are bit out of date but are probably still accurate.
  • Because the church has a higher fertility rate than average, achieving only population growth means they are actually losing members on the conversion side, in spite of an aggressive proselytizing program.
  • The church cannot sustain long-term exponential growth.
  • Therefore, I think the long-term Loomis and Anderson projections are much more believable than the sky-high Stark or Koltko-Rivera projections.


Sarah Kate said...

two and a half BILLION!? oh my. that guy's dreamin. that's interesting though.

ZDM said...

Always the first and most important thing to identify is the question of interest.
The main take-away from your post seems to be that it is quite possible that Stark is measuring something different than Anderson and Loomis are measuring.
Particularly if Stark is talking about members of the LDS church (people who have been baptized and have not had their names taken of the records of the church) whereas Anderson and Loomis are talking about people who identify as 'Mormon'.
These are different measures and should not be considered to be the same. It is very possible that both the LDS church records and the Mexico census records are accurate in the measure that they are purporting to measure.

ZDM said...

'taken off' not 'taken of', I apologize.

ZDM said...

One more comment. . .
While the census records and the LDS church are obviously measuring different things, the researchers may be measuring the same thing but using different models.
When one is measuring growth over a 'long' time frame (bacterial growth in a test tube, population growth of a community, nation, or world) models that incorporate a leveling off (S-curve models) are typically more appropriate than those that remain in exponential growth indefinitely. An exponential growth curve does indeed include an period of exponential growth, but that does not continue indefinitely.

"All models are wrong, but some are useful."
- George Box

Bennion said...

Thanks Sarah!

ZDM, I think the situation is the second one you identified: they are meauring the same thing (headline official number of members) but are using different models. Stark and Koltko-Rivera are projecting long-term exponential, while Loomis (and I think Anderson) are projecting a decaying exponential, like what you're referring to.

I think, because the headline numbers are different than the active or self-identifying members, that the Loomis or Anderson models are more likely to be correct.

ZDM said...

Yeah, I figured that was the case. I let myself get thrown off for half a second by fact that you were discussing the first situation elsewhere in the post. (With the first point it becomes obvious that your use of the word 'overstating' is patently unfair and inaccurate given the truth is a difference in metric.)

The assumption of a long-term exponential model is ridiculous when extrapolating that far, while the exponential association model is clearly more appropriate. (But you should remember to be consistent on the wrt growth models in general.)
On the other hand, I have found exponentiation association models like that Loomis used to be rather annoying to fit when I was running them for collaborators. (My data set was smaller than theirs, however, so they probably didn't have this issue.) I also have yet to find a paper in a statistics journals studying the validity of such a model. They may exist and just do not come up under the name that I know them by, 'exponential association models'. So, I don't know the statistical properties of this model. Finally, I suspect both models in that they extrapolate further than I think is reasonable.

ZDM said...

One thing that Stark got right was reporting a confidence interval (assuming that this is why the 'low' and 'high' estimates) the width of which appropriately got larger as the time increased.
From what we see here, Loomis includes no such interval and is only reporting a mean estimate. I have little doubt that his intervals should be large as well.

Bennion said...

Sorry if "overstating" was misleading, I just meant that they're counting more people than would count themselves. I guess I consider their reported membership numbers misleading to people who don't understand how they count. I should have stated that more clearly. A confidence interval would have been convenient for the other estimates.

Post a Comment