Question Regarding Accuracy of AncestryDNA Ethnicity Estimate

Question

Hi, All.  I'm new here, and have two related questions, one general and one specific, about the accuracy of AncestryDNA's ethnicity estimate.  I apologize if either has already been asked and answered here.

My general question is, how accurate do you find AncestryDNA's ethnicity estimate to be?  Does it generally correspond closely to what you see in your family tree?  I assume that for each ethnicity population, they are relying on SNPs that are largely specific and exclusive to that population, but have any of you gotten high percentages for populations that are similar/related to your family tree, but that your tree doesn't actually include?  My specific question, below, will illustrate what I mean.

My AncestryDNA ethnicity estimate projects my highest ethnicity percentage as Scandinavian.  I've determined my family tree with a pretty high degree of confidence, to at least 5 generations on all line, probably about 8 generations on average.  It includes a lot of "German" (i.e. swiss german anabaptist) lineages, but little to no Scandinavian.  I think the potentially most likely explanations for this apparent discrepancy are: (i) Scandinavian and central-european-german are similar enough that AncestryDNA essentially mistakes one for the other, or (ii) my tree will show a substantial Scandinavian component, if I just trace it back a few generations further.  Any opinions on what is the more likely explanation?

Answer 1

I have the same high % of Scandinavian ethnicity despite no known ancestors living there.  I think the best explanation is history.   I.e. there were wars and movement of people.  Of course, I'd have to look into how Ancestry, com determines what was original and what was due to movement.  So say, if segment A is found largely in England today, is it because the original settlers of England had segment A or because of vikings conquering England and interbreeding or for that matter because of merchants from England peddling their goods to the Swedes and gaining sexual favors in return.  Can these possibilities be decided just from the current gene patterns or do certain assumptions have to be made which could be incorrect?

Comments

Thanks, Dave.  I think your "history" explanation is substantially the same as my "trace it back further" explanation.  Meaning that, while you or I may have some allegedly English or German ancestors, if we go back a few generations further, we will see that those ancestors are actually the descendants of Danish invaders or something similar.

---

You need to know how they analyze English.  If they don't show it as English but analyze it into European types, Americans with some English ancestry are going to get very misleading results.  Where the English originally came from is neither here nor there and you can't get back that far genealogically.

---

Thanks, RJ.  I think you've largely hit upon my question.  What time frame do the SNPs that Ancestry relies on tie to?  Maybe they don't even know.  If you have an ancestral line that lived in East Anglia 500 years ago, and lived in Denmark 1000 years ago, will that show up as English or Scandinavian?

Answer 2

You are right to be skeptical about the ethnicity percentage estimates from Ancestry DNA.  Ethnicity percentages have a number of serious methodological flaws.  This blog post from Judy Russell outlines several of the limitations of ethnicity percentages:

http://www.legalgenealogist.com/2016/05/01/those-percentages-revisited/

Comments

Thanks, Ray.  What methodological flaws specifically do you have in mind?  To my understanding, Judy Russell's blog post generally asserts that the ethnicity estimates are unreliable, but doesn't specifically suggest what the source of the alleged error may be.

---

There are a number of validity and reliability challenges that have been made against the ethnicity percentages.  Here is another blog post, this one from Roberta Estes, in which she goes into great detail on this:

https://dna-explained.com/2016/02/10/ethnicity-testing-a-conundrum/

Specifically, she focuses on three issues:

"1. In determining majority ethnicity at the continent level, these tests are quite accurate, but then you can determine the same thing by looking in the mirror.  I’m primarily of European heritage.  I can see that easily and don’t need a DNA test for that information.

2. When comparing between continental ethnicity, meaning sorting African from European from Asian from Native American, these tests are relatively accurate, meaning there is sometimes a little bit of overlap, but not much.  I’m between 4 and 5% Native American and African – which I can’t see in the mirror – but some of these tests can.

3.  When dealing with intra-continent ethnicity – meaning Europe in particular, comparing one country or region to another, these tests are not reliable and in some cases, appear to be outright wrong."

So - as she and several other DNA genealogists have stated, the ethnicity percentages are reliable and valid at the continent level.  If you have European ancestors, your ethnicity percentages will include European results.  If you have African ancestors, your ethnicity percentages will include African results.

The ethnicity percentages are neither valid nor reliable beyond the continent level.  The example that is frequently cited is the Ancestry DNA commercial in which a person exchanges lederhosen (Germany) for a kilt (Britain).  While this might be appealing for a commercial, the test results cannot be taken as evidence.

In your case, I would watch putting too much faith into an ethnicity test result that shows Scandinavian ancestry while you were expecting German.  These are both European, and beyond this, they are very close geographically.  The current  approach to ethnicity percentages simply does not offer consistent valid and reliable results on ethnic differences within a continent.

---

I suspect that is a substantial overstatement, and that AncestryDNA can pretty reliably distinguish, say, Sicilian ancestry from Norwegian ancestry, or Ethiopian ancestry from Nigerian ancestry.  But if you're talking about say, different related intracontinental regions, what causes the error/overlap?  Are there no SNPs that pretty reliably distinguish a Swede from a Bavarian from an Englishman?

---

Sure - you can find differences within Europe or Africa that will naturally be more accurate due to the simple fact that Sicily and Norway are very far apart geographically, etc.

But - people from Norwegian ancestry do not typically get Ancestry DNA results suggesting that they are Sicilian.  The much more common results for someone from Norwegian ancestry is to receive ethnicity results that state German or British ancestry.  As you describe it, related intracontinental regions.

You keep mentioning SNP testing that reliably shows intracontinent ethnicity differences (i.e. to distinguish a Swede from a Bavarian).  I have never seen any studies that show this level of distinction.  Have you?

The blog post that I linked from Roberta Estes goes into great detail on this.  For Northern Europe (Britian, Germany, Scandinavia, etc.), there has been substantial population intermixing over the centuries due to a vast range of political, economic, social and military influences.  The current referent populations for DNA ethnicity tests are all based on the tests of living people from these various regions.  The DNA tests of living people are an inadequate representation of this vast range of population intermixing.  Roberta goes into this flaw in great detail.

So - I guess it comes down to belief vs. science.  If you want to believe that these ethnicity percentages mean something, then go with it.  But - the science simply is not here to back to this up - especially when you are talking about Scandinavia vs. Germany.

---

To be clear, I am not "mentioning SNP testing that reliably shows intracontinent ethnicity differences (i.e. to distinguish a Swede from a Bavarian)."  I am asking whether, in the experience of users here, people have generally found such asserted distinctions to be reliable or unreliable. As far as scientific papers go, I have not found one that specifically claims to distinguish Swedes from Bavarians on the basis of SNPs, but I also haven't searched for such a paper.

---

Let me give you a personal example. I have a fairly unusual Y haplogroup, and it is, so far, unique to England. But my line split from from the lines of all the others who have tested somewhere between 2000-3000 years ago. Not much good for genealogy! The current thinking is that the precursor to this haplogroup came from Iberia in the Bronze Age. So what would that make my ethnicity? My Ancestry report says I'm mainly Irish and Scandinavian. My Dads family is solidly English as far back as we've traced. My Mums is similarly French. As others have intimated, these Ethnicity Estimates don't/can't discriminate within Europe.

---

If a bunch of relatives in America share a mutation with a bunch of relatives in Sweden, obviously that calls for explanation and suggests one.  But this is closer to the level of individual matching, not ethnicity testing.

Any mutation that's widespread in Sweden is likely to be one that came from somewhere else when the ice receded.  It's difficult for a local mutation to spread without a bottleneck, so paradoxically the most typical features of a population will usually be foreign ones that came through the bottleneck with the founders.

---

I don't know if Sykes is still doing the Ursula stuff, but generally people have given up linking yDNA and mtDNA with ethnicity because of the absurdity of focussing on a single line.

This can so easily produce silly results.  Think plantation owners' sons and slave girls.  How many black Vikings are out there?

---

One right here!