As a genealogist with known German heritage, there has been a series of complications regarding the accuracy of my ethnicity DNA results given the complex nature of German DNA.
With the difficulty presented in identifying German DNA, or more broadly ethnicity in general, this has sparked my interest in identifying which of the DNA tests at 23andMe, AncestryDNA, MyHeritage, and FamilyTreeDNA gives the most accurate and detailed ethnicity results.
My Thought Process: Analysing the Companies
Considering that most of my ancestors (over 80%) originated from what is nowadays Germany, I began my analysis by comparing the current known locations of my ancestors (based on 25+ years of genealogical research) against the ethnicity predictions.
Kickstarting my analysis, I took into consideration the quality of ethnicity predictions and this is what I’ve found:
• The quality of ethnicity predictions was the lowest for FTDNA, followed by MyHeritage who could only score well based on their genetic groups (which are based on your DNA matches as well as your family tree).
• A comparison between AncestryDNA and 23andMe shows that they are far and away the best. Although Ancestry provides additional value with their communities, which are based on DNA matches and family trees, their ethnicity predictions are too far away from what the current genealogical research suggests.
• 23andMe presents its results somewhat high-level, but the real value is in the prediction of regions based on DNA data and the mapping of ethnicity predictions onto chromosomes. There was unfortunately only one downside: the test no longer determined my Italian heritage, as proven by DNA triangulation.
Now to deepen our analysis, let me bring you on a journey of my own genealogical research!
My Own Genealogical Research
While both my parents have shown traces of Scandinavian DNA, I have yet to find any evidence of it in my genealogical research. To understand this further, I have dived into how exactly ethnicities are predicted.
Prediction of ethnicities are commonly referred to as “Ethnicity Estimates’ in your DNA test. With many people turning to DNA testing to find out more about their ancestry and ethnicity, ethnicity predictions are largely based on two main factors:
• Samples of data collected from scientific databases (e.g., the 1000 Genomes Project) and customer data collected from customers who self-reported that all 4 grandparents originated in the same country/region
• Clustering of DNA segments inherited from countries or regions based on best fit
However, while these factors drive the bulk of ethnicity predictions, they have been showcasing several problems:
• A lack of diversity in scientific databases, especially on the African and Asian continents, which are marginally represented.
• Even though all four of my grandparents were born in the same town, it does not mean that their ethnicity, and therefore mine, comes from this city or even the same region. I know this from the fact that one of my 8 great-grandparents was born in Asiago, Italy and has no German ancestry.
• A lack of precision, e.g. Putting DNA segments into categories like "European" or "Northern European" doesn't really help identify the country, let alone region, these ancestors came from.
• Recombination events during meiosis result in a random inheritance pattern for DNA, so siblings can have very different ethnicity predictions, and some ethnicities may not be inherited by one or more siblings.
• Non- paternity events (NPEs) can lead to incorrect results, since the person testing is unaware of the ethnicity of the real birth parent(s). In hospitals, accidents like swapping babies may happen. Here's an example: https://www.washingtonpost.com/graphics/2017/lifestyle/she-thought-she-was-irish-until-a-dna-test-opened-a-100-year-old-mystery/
Despite the fact that the method of comparing ethnicity percentages between DNA-based estimates and genealogical resources isn't perfect, for the reasons aforementioned, in the absence of a better method, we will use it as our main quality assessment.
It is also important to keep in mind that DNA testing companies are constantly updating their predictions.
My DNA Test Results
23andMe offers different confidence intervals (speculative, standard, conservative), decreasing in detail (conservative is the least detailed). This article focuses solely on the speculative confidence level.
Split view between my parents (2014):
Looking at the 5 predicted regions of Germany (2021):
The following two Belgium regions (2021) have been identified:
Chromosome view (2021):
Ethnicity estimate and genetic groups (2021):
My Origin v3 (2021):
Comparison with my father (2021):
Comparison with my mother (2021):
How accurate are they really?
Now that you’ve had a sneak peek at how each DNA testing company performed in comparison to my genealogical research. For a fair comparison of the current state of their accuracy and detail level, we will use all of the 2021 predictions.
Summary of ethnicity predictions versus genealogical based ethnicity:
Ancestry DNA offers the greatest breakdown into countries, with six predicted countries. Their 55% of predicted German ancestry, or rather Germanic Europe, is the second lowest. In the case of Germany and France combined, it increases to 60%.
Since the genealogical research so far shows no evidence of British ethnicity, it could either be an NPE or, more likely, a misinterpretation or an inability to distinguish between Germans and British (due to the Saxon history of the British).
At first glance, the prediction of 100% German (French & German) does not seem very detailed. A great deal of 23andMe's value, however, is in the more detailed DNA-based regions that the company adds. Ancestry DNA and MyHeritage, on the other hand, base their communities (Ancestry) and genetic groups (MH) mainly on submitted family trees.
I only have a small family tree associated with my mom at 23andMe, so it doesn't have any location information (based on the last update December 2020). Most of my German ancestors came from those four regions. The link to the two Belgian regions makes sense through my Calefice ancestors (most likely from Liege).
Currently, 23andMe is the only DNA testing company that offers their customers a chromosome map of their ethnicity prediction. In this way, we can determine exactly where certain ethnicities are present on our chromosomes!
Update: FamilyTreeDNA just announced the same feature (as of September 2021).
The 97% figure for North and Western Europe is vague and disappointing.
However, it seems they are focusing more on combining DNA with their large number of family trees. The genetic group of people from Vicenza is correct. However, no DNA percentage is associated with this group or any of the other largely overlapping genetic groups. This approach, however, is a self-fulfilling prophecy, as NPE or incorrectly researched family trees cannot be used to determine ethnicities.
Family Tree DNA
Even though FTDNA's new v3 MyOrigins prediction (from a 100% Western & Central Europe prediction in 2014) shows a considerable improvement, the percentages are still very far off from both other ethnicity predictions and my own genealogical research. While the 38% for Central Europe is way too broad, the 33% for Britain is simply not supported by any genealogical evidence. It's probably way too high to have 18% Scandinavian DNA since it would mean that two of eight great-grandparents are more or less from Scandinavia.
FTDNA only recognizes Italian heritage, as does 23andMe's older prediction
Only FTDNA recognizes my Italian DNA, which is quite disappointing. It is even relatively close to the expected 12.5%. 23andMe identified my Italian heritage in 2014 and 2017 with 6.9% and 7.0%, but somehow stopped recognizing it in 2020 when they changed their algorithm.
FTDNA fails to tell me where my 10% Italian DNA comes from, which is where it goes wrong. It doesn't understand that it's from my father (whose predicted value of 0% was wrong), and with 0% from my mother, it came out of nowhere!
Verdict and Recommendation
FTDNA is ranked last. In addition to being the least helpful for your research, their predictions look completely off for a large portion of my DNA. Italian DNA “out of nowhere” is just the final nail in the coffin.
MyHeritage gets the 3rd place, building on the most detailed number of genetic groups but their pure ethnicity prediction isn’t giving you any hint of potential NPE and where you should look for those brick walls that you’re surely facing.
23andMe and Ancestry are far above their competition. Although AncestryDNA provides many fine details, some of these countries also seem incorrect. 55% of German DNA is too low, whereas 15% of Scandinavian DNA is too high.
Therefore, I recommend 23andMe for determining where exactly your German ancestors originated. When I compare their regional prediction to my existing genealogical research, it is the only one that is based on DNA and is very accurate. Only they recognized the Belgian DNA lead. Their currently sold DNA tests have shown the same detailed regional level for African and Asian countries, which is a huge improvement for those of African or Asian heritage.