# Non-reproducible research

About 20 years ago, I realized that I had a fairly unique opportunity to test the hypothesis that 4-H was strongest where it was multigenerational – 4-H members grew up to be 4-H leaders, and the program was strongest where the multi-generational membership was the most common.

I was working with 22 counties, and 4 of them had Extension secretaries with 30 or more years of experience, and full records.  Complete records is more challenging than you might think – when I worked as a County Agent, the records were in the basement, and a cracked sewer line helped me make the decision that they couldn’t be recovered.  Obviously, if I had only 4 counties out of 22, reproducing the research would be difficult at best.  On the other hand, if it didn’t get done in the next year, retirements would make it impossible to do once.

In 1950, 18% of rural youth belonged to 4-H, with the membership plateau ending in 1976 (Putnam 2000, Bowling Alone), with a 26% decline in membership between 1950 and 1997.  And I was listening to folks who told me that the problem was a shortage of volunteer leaders.  It looked like I could find the numbers in those 4 counties with the oldest secretaries.

I was on a roll – the secretaries showed that 151 4-H families had at least one parent who had been a 4-H member as a child, 78 families where neither parent had been a member, and the parents of 6 families could not be determined.  We defined 4-H members who had belonged to a club four years or more as persistent, and contrasted their statistics with first-year members.  None of the six families whose 4-H history couldn’t be determined had any persistent members, so the sample, while not particularly large, was clean.

Well, the stats were simple – Chi square was calculated at 45.03, the probability of the distribution occurring by chance was less than 0.001.  The data supported the hypothesis that parental involvement in 4-H (as a club member) is the greatest single predictor of member persistence in 4-H.  Two thirds of the persistent members (4 years or more) had parents who had been 4-H members in their youth, while two thirds of the first-year members had parents who had not been 4-H members.   The kids most likely to drop 4-H were kids whose parents had not been in 4-H and were not 4-H club leaders.

The evidence was pretty solid that a multigenerational 4-H identity helped keep kids in 4-H – but it was equally solid that 4-H membership wasn’t random . . . it was hereditary, like the British nobility.    Still, making a conclusion about a national program from a sample of 334 people in 4 counties seems to be a stretch.  As I look at the Harry that was once an English prince, I wonder about researching the worldwide decline of royalty.

Non-reproducible research isn’t necessarily bad research, and it can provide some interesting conclusions – but it is better when you know it’s non-reproducible from the beginning.