Comparing Ethnicity Estimates

     A common saying in the genetic genealogy community is “friends don’t let friends test just for those percentages.”

     Ethnicity estimates can only tell a tester so much, and the real meat to chew on is found among your DNA relatives. However, ethnicity estimates are usually why people initially decide to test, 1. because let’s face it–they’re fun! and 2. because testers are often unaware of the bounty of information provided in their match list. Since ethnicity estimates are often where people begin (including me), I thought this would be a fitting topic for my first blog post.

First, there are three major DNA testing companies considered to be “legitimate”, known as The Big Three. They are:

  1. 23andMe
  2. AncestryDNA
  3. FTNDA

23andMe is widely considered to have the most accurate ethnicity estimates, with AncestryDNA a close second. FTDNA is a distant third in terms of intracontinental accuracy, but an update to their MyOrigins feature is expected later this year.

Below I explore the ethnicity estimates for my personal kit across a few platforms. I originally tested with AncestryDNA (23andMe won’t be covered in this blog), then transferred the data to FTDNA, DNA.Land, WeGenes, and Gedmatch. Let’s have a look at the discrepancies across these companies given the same raw data.

  First, let’s look at the AncestryDNA results.

AncestryDNA ethnicity estimate
These results pretty much match my paper trail, and I was impressed to see that the nailed down my Ashkenazi heritage to exactly 25%. I know this is accurate because I have one Ashkenazi grandparent.
The 2% Asia South was somewhat of a surprise, but considering our paternal lineage y-dna haplogroup is of a possible Persian Jewish origin, it didn’t strike me as inaccurate–or “noise”, as trace regions of small amount are often called.

Next, comparing the same kit to FTDNA’s ethnicity estimate, we see some slight differences.

FTDNA Ethnicity Estimate

Again, this is the same test based on my DNA sample at Ancestry. Notably, the Scandinavian component is significantly lower, and the Eastern European component higher. This is because 1. as mentioned, FTDNA is not considered to be quite as finely-tuned as AncestryDNA testing, and 2. different companies use different reference populations to calculate their ethnicity estimates. This can be frustrating for researchers, but also illustrates how ethnicity calculation is not an exact science–in fact, it is often referred to as “cocktail party conversation”, one of the reasons ethnicity reports are not widely emphasized by professional genetic genealogists.

Well-known genetic genealogist and web programmer Kitty Cooper demonstrates this in her blog post “One company’s British might be another company’s Scandinavian…these predictions reach further into the past than our family tree does and may reflect earlier migrations around Europe. Finally, predicting deeper ancestry is not yet firm science or these companies would be closer to each other in their results.”

Next up is DNA.Land.

While not a testing company per-se, this website is administered by researchers from Columbia University and the NY Genome Center. It is a non-profit, so participation is free and highly encouraged to contribute to scientific research.

dnaland-ethnicity-estimate is relatively new, and has a few bugs to be worked out in its ethnicity estimate. Most notably, it widely exaggerates the Ashkenazi component for testers. This is a known problem. As we saw in the AncestryDNA results, which were true to my paper trail, I’m 25% Ashkenazi–not 37%.

For some testers like Cooper, DNA.Land has actually underestimated Ashkenazi heritage–further confusing the issue. Like me, she has known Ashkenazi ancestry in the amount of 25%, but estimated her kit to be significantly less at 16%. 

Next we look to WeGenes

WeGenes is not among The Big 3 testing companies and considered practically useless for European testers. It is based in China, and is most reliable for Asian ethnicities. Because I have trace amounts of Asian heritage, I thought to include it here.

WeGenes estimate
Immediately obvious are the strange European results that do not match my paper trail, and can pretty much be ignored. I’m certainly not 25% Spanish. The Ashkenazi component was vastly underestimated, and French isn’t really an ethnicity to begin with (usually described as “Western European”, including other nationalities like German). These things are problematic.

The South Asian–Sindhi, Middle Eastern–Iranian, and Middle Eastern–Saudi, are more likely to be reflective of actual ancestry. I do not have a paper trail to India or Pakistan, where the Sindhi originate, however, once again our paternal haplogroup has origins in Ancient Persia. So the Iranian component could fit with what we already know. More on this below. is not a testing company per-se, but is considered the Holy Grail for DNA researchers. GEDmatch has a myriad of useful features, including a number of admixture (ethnicity) calculators, with region-specific reference populations to meet a wide range of research needs.

To begin with, let’s check the standard Jtest results.

Gedmatch Jtest pie chart  
Jtest is known for underestimating Ashkenazi, as weird as that is. I’ve read that 100% Ashkenazi testers routinely recieve around 30% AJ on the Jtest, so with my result here shown as 8%, that seems to mesh pretty well with AncestryDNA’s real result of 25%. Pie charts are pretty to look at, but what’s really interesting is something called the “oracle” (sounds interesting already, right?)

Gedmatch Jtest oracle
Here is the oracle for my Jtest results. The secondary population is representative of deeper ancestry, and here we see not only Ashkenazi (AJ), but also Druze-modern inhabitants of the Levant-, South Italian, and Samaritan. Jtest is a starting point for Ashkenazi testers on Gedmatch, but to dig into these components further, we need to check out other calculators.

 Gedmatch Gedrosia Near East Neolithic k13

 Gedrosia Near East Neolithic k13 pie chart 
Neolithic is a term meaning “New Stone Age”, or the time period concurrent with the advent of farming. These samples that comprise the reference panel for this calculator were taken from ancient human remains. Perhaps what’s notable here is the 9.54% Iran_Neolithic, which could be tied into our assigned paternal haplogroup. More on this in the oracle below.


What’s interesting about these oracles is that they very nicely display that I have one Jewish grandparent. Moreover, it gives us a look further back in time before Ashkenazi Jews existed as a culture to get a feel for where my grandfather’s family originated from. The recurring Jew_Iranian and Jew_Iraqi would be in line with what we know about the origins of the haplogroup R2a/M124 arising in Babylon and Persia, during the period of the Jewish Captivity in Babylon. More on this in an upcoming blog post once FTNDA finishes processing our y-37 upgrade. Those subjects are beyond the scope of this introductory post, and will have their own dedicated post soon.

The Eurogenes EUtest V2 k15 admixture calculator

The Eurogenes EUtest V2 k15 admixture calculator doesn’t tell us much by the pie chart alone, but once we click on the oracle we get some really interesting information.




What’s interesting here is my kit is suddenly clocking in as SEPHARDIC! I reached out to an expert on Sephardic and Mizrahi DNA, and she confirmed that this calculator has a reliable Sephardic reference population.  Unlike with the paternal haplogroup suggesting a Mizrahi origin for my grandfather’s line, I had no idea that there was any Sephardic DNA in our family at all. The Eurogenes reference panel is not from ancient remains like the Gedrosia Near East Neolithic, so we’re looking at a much more recent event here. Historically it is well-known that Mizrahi Jews moved into the area of the West Mediterranean and into the Iberian Peninsula, probably through Rome. So although it was a personal surprise to find Sephardic heritage, it certainly meshes well with what we know about the origins of European Jews.

This will be covered more extensively in an upcoming blog post, as mentioned. I’m looking forward to sharing the research I’ve uncovered, along with some terrific reference material and expert opinions in the next post.


Cooper, K. (2016, April 16). Ancestry Composition Comparisons: a Case Study. Retrieved March 06, 2017, from

Further Reading

Estes, R. (2017, January 15). Concepts – Calculating Ethnicity Percentages. Retrieved March 08, 2017, from

For an exhaustive overview of the difference ethnicity analysis tools available on the web that aren’t covered here, ISOGG-International Society of Genetic Genealogists- is the go-to Wiki.



8 thoughts on “Comparing Ethnicity Estimates

  1. Great post—very interesting to see the comparisons. I tested on 23andme, my brother and mother at FTDNA, and all results are on GEDMatch, but I’ve not focused on the ethnicity breakdown since I know that I am 95+ % Ashkenazi, so didn’t really care. I tested more for the chance of finding relatives. But now I might go back to see how the tests came out differently. Welcome to genealogy blogging!

    Liked by 1 person

  2. Great post! I could never make sense of the different options for ethnicity on Gedmatch – maybe you could do a future post explaining the differences?

    Paper trail I’m 100% Bulgarian as far back as I’ve gotten. Out of the big 3 only 23andme had me as 70% Balkan, the other two, ancestry and FT, had me as combination of Eastern and Western Europe, and other bits thrown in. In other words – no practical use 😉

    Liked by 1 person

  3. Hello,

    I just want to leave a quick comment to say thank you for this very helpful post! I recently tested with Ancestry and was surprised to be given an estimate of 14% European Jewish. That line (3/4 of my grandparents are Irish, so this is comparatively easy to isolate) goes back in the Church records for a couple centuries, but the family genealogist suggested this might be from my GG-grandfather, as my great-grandfather is listed as illegitimate, but in his small hometown, 20% of the population was Jewish at the time. Your blog post gave me an idea of which calculators to look at on GEDMatch to see how they compared to Ancestry’s estimates. The 14% seemed a bit high for a GG-grandfather, but using the Gedrosia Oracle-4 above as you suggest, it came out 4/80 — pretty close to the expected 6.25%!

    It was great being able to see how you started down the path, and be able to replicate some of the same methods on my own. Thanks again:)

    Liked by 1 person

  4. I really enjoyed your post: it’s approachable, covers a lot of ground, and is exceptionally well-written. However, I think there’s a misunderstanding about how DNA is inherited that runs through the essay that might deserve a little attention.

    When talking about the genetics you inherited from your Ashkenazi grandparent, you said:

    “As we saw in the AncestryDNA results, which were true to my paper trail, I’m 25% Ashkenazi–not 37%.”

    The Estes article you cited does a pretty good job of explaining why you might get a result from your testing that’s not reflective of the statistics implied by your “paper trail.” Let’s revisit it.

    Estes wrote, “Ancestral DNA isn’t divided exactly in half, by the “one for you and one for me” methodology. In fact, DNA is inherited in chunks, and often you receive all of a chunk of DNA from that parent, or none of it. Seldom do you receive exactly half of a chunk, or ancestral segment – but half is the AVERAGE.”

    Think about genetics like a card game. The hands we’re dealt, no matter how well the cards are shuffled, aren’t always perfectly even distributions of the cards. Even though the whole deck is evenly distributed among the suits, some hands lean more towards some suits than others.

    To carry the analogy further, let’s play a game.

    Pretend the four equal suits of a standard card deck represent the genes of my four grandparents, as spread across both of my parents. For the sake of argument, let’s say that each of my grandparents has only one ethnicity and each of the four grandparents is different — each is one easily distinguished single suit of cards. (That’s a highly unlikely situation, but it’s only an example, so we can start with whatever we want.) Dad’s dad is Hearts, Dad’s mom is Diamonds, Mom’s dad is Clubs, and Mom’s mom is Spades.

    Let’s also say that, like in genetics, each of my parents got half of each of their parents hands, nicely divided into two equally represented suits. So in this metaphor, each player gets a hand of 26 cards. Dad had his 26 red cards that he got from his parents and Mom had her 26 black cards that she got from hers.

    Now it’s time for Mom and Dad to make a new hand and give me some cards. But I only get 26 cards total, just like my parents got — 13 cards from Dad and 13 from Mom. Dad’s going to shuffle his hand and deal down 13 red cards. Mom’s going to do the same with her hand and deal down 13 black cards.

    In the end, the 26 cards I get will be equal parts Mom and Dad, but they won’t be equal parts Diamonds, Hearts, Clubs, and Spades. I could even get a hand with nothing from one or two suits. It’s unlikely, but it’s possible.

    And then this same thing keeps happening, from hand to hand, from generation to generation. Some ancestral contributions are stubbornly over-represented, others fade away into obscurity. That’s what may have happened with your Ashkenazi roots. It’s also possible that a little Ashkenazi on the other side of your family is stubbornly hanging on to tilt the balance beyond what you’d expect from the one grandparent’s contribution.

    It’s also true that some of the tests are better at identifying the contributions of certain regional populations than other tests are, because they’ve chosen to identify particular markers with particular groups. The Western European, Iberian, and Scandinavian examples you mention in your essay are good examples of that. Those populations are handled very differently by the different testing companies — witness the hugely different results the same person gets for those regions from different companies’ tests. That can also be tilting your Ashkenazi result this way on one test and that way on another.

    So no matter what the paper trail is, no matter who the ancestors truly were, you might still really not have exactly 25% Ashkenazi markers in your genetics. And that’s perfectly consistent with your ancestry.


  5. @Elizabeth today, I’ve been searching about JTest 🙂 Funny… I discovered it only now, in 2018. Anyway, I googled a bit more, spotted on ur blog, remember that I’ve read ur post about Sephardic vs. Aszkenazi. And I updated my article today (, mentioned ur article. Thanks. This topic is not very critical (significant) for me, due to very small level of AJ DNA value, but still in general it’s interesting – different companies, different values, different algorithms.
    And btw, I didn’t know also about WeGens, I see it accept only 23andme or Ancestry format, and I have FTDNA format. I maybe, will search converting tool like this ( but in opposite direction 🙂


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s