Here’s the data I’ve gathered and that I use to make most of the maps and charts here. You can find the actual testosterone levels from each study, the age of the people involved, the location, and a link to the study. When possible I used the mean testosterone, but I used the median if that was all that was available. In some cases I had to measure or guess the mean from a chart, when the actual mean was not written in the study.
Spreadsheet of data from locations all around the world
In the following national averages spreadsheet I either used data that was collected from people all over the nation, or I used data from locations that seemed to be nationally representative based on population density, or I averaged results from several locations around the country, trying to take into account population density, when this was possible:
Spreadsheet I used to come up with national averages
I hope all of this is useful. I would love it if professional scientists can use this data with more accurate methods than mine, and to get more insights than I’m able to do.
Characteristics of the Populations
Much of the data came from the control groups in studies of other things. As much as possible, I tried to use only studies where the men reflected the average population’s characteristics. I generally didn’t use testosterone data from groups that were selected to be either more or less healthy or obese than the rest of the population, since testosterone varies greatly with sickness and obesity.
In a few cases where I couldn’t find any better data, I “reverse-engineered” a group that reflected the national average. For example, if a study compared the obese with the lean in a population, I could find the proportion of obese and lean people in that population, and then calculate a weighted average from the two groups in the study that would reflect the whole population. Generally, I’ve noted what I did in the spreadsheet.
My Age-Adjustment Method
Testosterone rises from puberty until around the early 20’s, and then slowly declines with age. For example, a 20 year old with 700 ng/dl testosterone might decline to 565 ng/dl at age 50 and 425 ng/dl at age 80.
The studies that I get my data from are rarely studies of the entire population. They usually are studies of one small age group. Sometimes they are young and sometimes they are older. Therefore the testosterone levels in each study are not directly comparable at all. But if you assume that testosterone falls at about the same rate in all the populations, and you know that rate, then you can guess what testosterone at age 50 would be from the testosterone levels of 30 year olds, or from the testosterone levels of 60 year olds.
A good statistician can hopefully do a better job of this than I did, but my best attempt at it was the following:
I found the linear rate of testosterone decline across a population study that included all ages from 20 to 80. This was the NHANES study of the US population from 1999-2004, which is meant to be nationally representative. Link to study: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6735742/figure/fig1/
The rate of decline in this population is -0.20681.
The average testosterone of the whole population, which was reported by the study, happened to be the same as the testosterone at age 50 in this population.
So now I could determine how much higher testosterone at age 20 or 24 or 36 was than the average of the whole population, or how much lower testosterone at age 57 or 65 or 72 was than the average of the whole population.
And I could take the testosterone level at age 20 or at age 70 and calculate what testosterone level was predicted for the whole population. This “age adjustment” is what I did to all the testosterone levels that I found in all the studies, and these are the numbers that you see on my charts.
I used the mean age of a group to do my calculations when this data was available, but when it wasn’t, I used the median of the age range.
Sometimes a study had multiple groups of different ages. In this case I predicted the total population’s testosterone separately for each age group, and then averaged the results of each prediction.
Accuracy of the Age-Adjustment Method
This is not an incredibly accurate method, but I hope that it does reflect some real information about the average testosterone of a population, and when it has been applied to all the studies, I think we can at least use it to compare the populations.
I did try to keep in mind some of the inaccuracies of this method when I did my calculations, for example I very rarely used studies of the very old to predict whole population’s testosterone levels. I tried not to use studies where I knew my results would be pretty tenuous.
For example, there was one Swiss study from the 70’s (link) that included 10 people. Six of them were under 18, and I almost never used under-18 data, because I never determined exactly what their test levels would predict. (Teens have lower testosterone than adults, but by how much? And it could vary a lot because people of the same age could be at different stages of puberty) So there were only 4 people left, and their ages were 19, 20, 21, and 54. The mean age of these 4 men would be 28.5, but this age is accurate for none of them. So I didn’t use that study.
Studies with Not Enough Data
When I had an alternative, I did not use studies that didn’t tell me the age or the characteristics of the men involved.
In some cases though, I’ve used studies where I don’t really know – 70’s USSR, 60’s and 70’s everywhere.
Time of Day of the Study
Testosterone rises through the night and falls through the day. For this reason, testosterone is usually measured around 8AM, when testosterone is at its highest. Most of the studies used morning measurements. When possible, I excluded studies that measured testosterone at times other than the morning. In a few cases, I really wanted to include a study which unfortunately had measured at some other time. So I “reverse-engineered” what the morning testosterone level would be, assuming the men had the standard circadian rhythm of testosterone through the day, for their age group (I used the data from this study (link) to calculate this) Again, where I did this I generally noted what I did on the spreadsheet.
Some older studies will tend to undercount testosterone because these tests were not as sensitive as the tests that were done more recently, and also because they may have collected blood later in the day than the more or less standard 8AM collection that is used now, when testosterone is highest. The variation in testosterone throughout the day does not seem to have been widely known until the 70’s. I did my best to take these old studies into account. When the time of day (or age of participants) was unknown, I generally didn’t use the study, unless there were other reasons to use it.
Saliva, Plasma, and Serum Testosterone
I don’t think I used any saliva testosterone measurements in the charts. It’s not clear to me how saliva testosterone compares with blood testosterone.
Testosterone in the blood can either be measured in plasma or in serum, and to the best of my knowledge these measurements are basically the same.
Measurements in serum and plasma are usually the same, except for a few things which aren’t hormones, according to this: https://www.vet.cornell.edu/animal-health-diagnostic-center/laboratories/clinical-pathology/samples-and-submissions/chemistry
Here’s a study showing that estradiol measurements are basically the same in plasma and serum: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6447065/
So testosterone measured in plasma should be comparable to measurements in serum.
Unconventional Testosterone Data
I have included my attempts to guess testosterone levels long before testosterone could be measured in the blood. There is one study of testosterone in hair. I tried to calculate what blood level that would predict.
I also found info on testicular size from a long time ago. Then I found what testosterone levels were associated with different sizes today.From that info, I could calculate what testosterone levels might have been a long time ago.
To be honest, this is not very solid research! But maybe it suggests something real. It’s fun to try to guess.
Approximately between the 1940’s and the 1960’s, hormones were measured in excreted urine. As of writing this, I haven’t used these studies. But in the future I may try to calculate what blood levels these studies would predict.