Is it common? Does it happen to everyone? Is it a big deal?
These are questions we genealogists ask ourselves when we fail to find an ancestor in a U.S. federal census despite our most strenuous efforts. When we do find a missing ancestor, noting how bad handwriting or poor indexing impeded us, we come to believe that such problems are always the culprit, and we were at fault for not overcoming these common impediments by searching more cleverly. But how often do we fail to find our ancestors in the census because they simply are not there?
Within the Jewish community of the Homestead District, the answer is: between 2% and 31% of the time.
Read on to see how I got to those numbers and, more importantly, why my analysis can help you become a better (or at least saner) researcher.
Determining the % undercount in every Homestead census
As part of my Homestead Hebrews project, I’ve been reviewing as many surviving sets of records as I can to find every Jewish person who ever lived in the Homestead District. For every person I have found, I have compiled a full set of records to trace as much of their life as I could— typical genealogical research, though at a larger scale.
After compiling a person’s records, I added a less common step: I reviewed those records to determine what years that person lived in Homestead. It helped that Homestead regularly published city directories from 1890-1945, but there were almost always gaps between records where I had to make inferences. With what information I could collect, I attached well-reasoned residence dates to as many people as possible.
From there, it was straightforward to compare the lists of Jewish people whom I believed lived in Homestead at the time of each census to the list of Jewish Homestead residents I found actually recorded in each census. Here is what I computed:
Jewish residents of Homestead skipped by the census
|Census||Number of Jewish residents in Homestead I identified||Number of these people in the corresponding census||% not in census|
These percentages are all over the place! The 1900 census of Homestead, with its 31% overall undercount, had serious deficiencies related to the town’s large immigrant population, which I wrote about previously. I will now examine the censuses from 1910-1940.
People the census missed and why
Looking more closely at the data, I found that one kind of person the census tended to miss are “singletons,” whether a young person working away from his family, an old widow/er whose grown kids lived elsewhere, a father whose family had not yet immigrated, or some other person who lived apart from their family members. In the pie charts below the table, these people comprise the blue wedges. This undercount is understandable; even a meticulous enumerator might not extract the names of all the secondary household members coming and going from a place.
Breakdown of people missed in each census
|Census year||Singletons||Families||Children||Couples||Total people|
(from 4 families)
But what about the disproportionately undercounted 1920 census? Here, the pie chart shows something else: entire missing families (orange) comprise most of the undercount, along with missing children from partially enumerated families (gray). Neither of these circumstances was significant in other years.
- One of those missing families is my own, 8 people including my grandfather. I had long assumed they were omitted because the enumerator came by while they were sitting shiva for the family’s youngest child, who had died of appendicitis just days before his Bar Mitzvah. (The enumerator was supposed to come back.)
- Another of those missing families was the Louis Schwartzes of Enumeration District (ED) 144, 9 people in total. They lived on one of the blocks skipped by Henry Silverstein, rogue enumerator extraordinaire (note the star).
- Also missing are their neighbors, the Weises of the adjacent ED 143, 11 people in total. Their house (first X from the left) was skipped by Anna Malone, who did spotty work.
There are six other missing families besides, and all but one of the missing families were large families of 5 or more individuals. These families, along with the four partially-enumerated families, were spread across the Homestead District, so I cannot blame a select number of shoddy enumerators. Something went awry on a large scale—in just the Homestead District? the whole county? the whole region?— because of hiring challenges? training challenges? the political atmosphere? the frigid weather?
I cannot propose to apply this specific range of undercount percentages more broadly beyond Homestead’s Jewish community. Odds are, though, that every ED you encounter omitted some residents. A recent study calculated that every census from 1850-1930 undercounted native-born white people by 3.6-6.6% (and other groups by more). The census bureau found that the 1940 census undercounted men of draft age by 3% (13% for black men of draft age), and the 1950 census 2.5% (11%). Real people made up that gap, and some of them are my family, and some of them are surely yours.
For genealogists: How understanding these results can improve your research
You have no way of knowing when you look at an enumeration district (ED) how good or bad it is. Every ED has its own characteristics—population, conditions, even enumerator— that affect its percentage of people counted. In other words, the quality of every ED is different. When you fail to find a person, there is simply no way to know if the problem is you, for not researching effectively enough, or the enumerator, for having skipped your person.
My analysis leads me to recommend some specific search strategies so you can feel confident that you have searched as carefully as possible. This list is not comprehensive (there are a lot of guides out there for that), but hopefully one or more of these ideas will help you find someone you couldn’t before—or learn to make peace with their absence.
1. Some people are just more findable than others
Most census research guides fail to acknowledge that your person of interest simply may not be there. Even after 1850 for white people and 1870 for black, when enumerators were instructed to list everyone, people were always missed. And, if the person you’re looking for falls under the broad category of singletons, or if they were non-white or non-native born, the odds are greater that the enumerator missed them. Set your expectations accordingly.
Even if you find a possible match, you have fewer context clues to confirm it for a singleton. In the 1930 census did I find my bachelor Jacob Mervis in Los Angeles or my divorced Max Markowitz in Miami? These are common names, and the census entries do not record enough differentiating information.
Non-singletons have another advantage: because they lived in a household, you will find them in the census adjacent to the rest of their family. It’s hit-or-miss how well transcribed a person’s name was; perhaps a family member appears more clearly in the index. For example, I couldn’t find Israel Miller in the 1940 census because his name was mistranscribed as Cassel Miller, but his wife and kids’ names were transcribed correctly.
2. You will need to search by address far more often than you realize
Over the years since I initially assembled my dataset, I have been continually surprised to find people in censuses whom I had not spotted previously. Even in this most recent round of work to prepare this analysis, when I did one last review of my missing people to make sure they were truly missing, I found people I hadn’t caught before. The primary difference between my earlier efforts and my later work is that I transitioned from name-based to address-based searches.
If the underlying transcription has mistakes, even your cleverest name searches might fail, and if your ancestors, like many of Homestead’s Jews, had ethnic names unfamiliar to both the canvasser and the indexer, the transcription will be more of an obstacle than you may realize. I found many transcriptions so far off that they eluded wildcard or soundex searches: Lembersky was indexed as Gambersky, Keisler as Rusler, Fischman as Tisthman, Grinberg as Geissberg, Carpe as Kalpe, Leinwand as Lemmand, Lefkowitz as Latkowitz, Glick as Wilk, Tauber as Pauls, Forkosh as Zorkosa or Corkern, and Sadowsky as Donnelly. Some didn’t even resemble names, like when Sam Ruben became Lam Buchen or Izak Hertz became Spak. In one instance the Samuels family, with their comparatively easy name, were indexed under “San…” because the name was partially obscured by over-writing.
The net effect of all these mistakes is that if you only use name searches, you will fail to find people who really are there. For years I could not find Dr. David Reiter in the 1920 census because the enumerator mistook David’s name, and the indexer transcribed “Docta Ruter.” Dr. Reiter was at 613 Eighth Avenue, however, exactly where the other records said he would be. In 1910 Morris Lefkowitz was not in the census where the other records suggested, but the indexer’s rendition of his name as “Lefhonetz” proved the bigger obstacle. The only way I could find David and others of my most elusive people was to set aside name searches and read through the census pages covering the address where I thought they lived.
Searching by address for a person in a census involves several steps. First, I had to build a case for where each missing person lived during the census by bridging evidence from before and after that census. Some examples:
- City directories played a central role. In Homestead every census either had a directory from the same year (1900, 1910) or a pair of directories 1-2 years on either side (1918/21, 1929/31, 1939/41). Or, I could pair a directory with a record even closer to the census date. When I had matching addresses in the 1939 Homestead directory and an October 1940 draft card, for example, I had a good case for assuming that person lived at that same address at the time of the April 1940 census.
- Creative use of genealogical records filled in other gaps. Abraham David Levin served as a witness for a July 1939 naturalization and wrote down his address as part of the process. That address matched his address in the 1941 Homestead directory, confirming that he had not moved. Benjamin Mervis died less than a month after the 1920 census; his death certificate gave his last residence. Eliezer Seiavitch’s daughter listed his address in her ship manifest when she arrived a month after the 1910 census.
Non-traditional records may better track people overlooked by more typical records. In my research, Homestead synagogue records are the best surviving documentation for a surprising number of people. Using directories and censuses, you’d think Benjamin Krotin, an older man living apart from his family stuck in Russia, didn’t appear in Homestead until 1929, but the synagogue records show that he was in Homestead as early as 1917.
If your person was a child or a woman (especially an unemployed, married woman), you may need to research an adult, male member of their household to find enough records to form an address hypothesis.
- Start by trying the address in a text search:
- If a town directory is available for that timeframe, use the address to find the missing person’s neighbors. Use the neighbors’ names to see if they turn up using a name search where your person did not.
- Search by putting the street name in the keyword field and the town in the location field to locate EDs where that street runs
- If you’re researching a city 1870-1950, use Steve Morse’s Unified Census ED Finder to determine which ED(s) to check.
- Note that a street on the boundary between two EDs may have the even-side of the street in one ED and odd in another.
- If you’re not researching a city, try one of these map-based approaches:
Compare the descriptions of the boundaries of each ED (available on Steve Morse’s site as shown and elsewhere) to a map to determine where the address falls in the ED.
- If the present-day street grid is unchanged from the past, Google maps will suffice
- If the street grid or street names have changed significantly, you’ll need to find a historical map. Try the Library of Congress listing of digitized Sanborn Fire Insurance Maps.
- Or find a map showing the town’s EDs. Here are some possible methods:
- Steve Morse’s one-step page for viewing ED maps works for 1940 and 1950. Ancestry built a nice tool, but for 1950 only.
- For 1900-1930, enumeration district maps are online here: United States Enumeration District Maps for the Twelfth through the Sixteenth US Censuses, 1900-1940. It’s quite laborious to navigate the images, but you will find the exact enumeration map for your town if you persevere.
At that point, it’s all down to your patience and focus as you read through the ED(s) looking for the address in question to see if your missing person is there.
3. Do check other trees (carefully)
The crowd-sourcing influence in genealogy websites’ hinting algorithms sometimes turns up gems. I found the missing census for Morris “Lefhonetz” years after I had given up because one day it showed up in his hints as the result of another researcher who knew more about his movements than I. Yes, you must always take other genealogists’ work with a grain of salt, but when you see a new hint or find your ancestor linked in someone else’s tree, review their work to see if it contains records you hadn’t been able to find otherwise.
4. Know when to stop
Full disclosure: none of people in the bulleted examples in suggestion #2 ever turned up in the census. But you know what? Because I feel confident about the other records I gathered, and because I tried all the strategies above, I’m ready to accept that these people, all of whom were singletons and immigrants, are simply not there. I documented in my research log for each person what I tried (so that I don’t mistakenly repeat the same work later), and I moved on.
In short, if you take one thing away from my analysis, remember that the census enumerators never caught everyone. If you’ve done reasonably exhaustive research and still fail to find a person in the census, rest assured that it is not uncommon, it happens to everyone, and it’s not a big deal. At some point you’ll need to accept missing people for what they are.
This post was updated in March 2023 to incorporate the 1950 census findings and to clarify counting methodology.
 The Homestead District includes, variously, Homestead, Munhall, West Homestead, Whitaker, Hays, Lincoln Place, and Mifflin Township. Whenever I refer to Homestead in this post, I always mean the Homestead District, not the single borough of Homestead.
 These numbers represent the people whom I believe to have lived in Homestead on the single day assigned to the census. This count includes children with one Jewish parent, even though in only very cases is there any documentation that these kids ever identified as Jews (5 such children in the 1920 census, 8 in 1930, 11 in 1940 (1.7%), and 22 in 1950 (5.1%)). I always counted the Jewish parent, even when there was evidence that the Jewish parent took on the partner’s faith. I did not count the non-Jewish parent. The 1950 census excludes 4 people who were entered twice.
These numbers are often low compared to the community’s contemporaneous estimates of its size (as recorded in various editions of the American Jewish Yearbook). My estimate is only as good as the surviving documentation. I hope to explore the issue of documented vs. undocumented people in a future post.
 1870 is the first census fully enumerating Black people in slave states; only free Black people were enumerated fully prior to that date. (This discussion omits the changing methodologies for enumerating Native Americans.)