Back to Additional Materials

Alan M. Strout, for the
Davenport Historical Society

Note on Estimating Davenport Population, 1800, 1810, 1820 & 1830 [1]

Population estimates for the area that became the later Davenport, NY, are not available from published sources on a consistent basis for the years 1800-1830. In 1800 and 1810, Davenport was still a part of the adjacent towns of Kortright (Delaware County) and Maryland (Otsego County). Davenport was formed (“erected” in official language) from these two towns in 1817, but sometime between 1822 and 1835 the new town lost a significant part of its territory and population to what became Oneonta, NY, in Otsego County. Many years later, in 1878, more land was transferred from Davenport to the adjoining town of Meredith.

If one is to gauge accurately population trends over time, Davenport's population should be reported for a comparable land area. For the period 1800-1870, this area has been defined to include those parts of the Fitches Patent in Maryland and Kortright which became Davenport in 1817 minus Davenport's Wallace Patent land lost to Otsego County sometime after 1822. (The exact date—or indeed the legal mechanism—of the latter transfer was at the time and is today unclear.) A second series of overlapping population estimates for the area excluding land lost to Meredith in the 1878 transfer, will be used for 1870 and subsequent years.

This note describes the procedure used for making the 1800-30 estimates. The population loss in the 1878 annexation was better documented at the time and will not be further discussed in this note. (But see footnote “e” in Table B of the final population estimates shown below.)

What follows is relatively technical. It describes in some detail a methodology for examining family names from Federal decennial censuses for three different towns and then determining the possible residents on that part of the land that subsequently changed townships. The procedure is then to assign a probability estimate to each family, reflected the degree of certainty that the family household in fact lived on the land in question. The final step is to combine household sizes and probabilities to arrive at an estimated total “probable population” for the transferred land at a particular date.

The method relies on several assumptions and is greatly limited by the fact that the early town censuses included the first name of only the household head. Children's, spouse's, and other names were not shown. Without knowing the names of family members, it is hard to trace households from census to census as household heads change through death, marriage or other events.

We assume, for instance, that if the same household (having identical first and last names) appear in the 1810 census, say, of Kortright and the 1820 census of the newly formed town of Davenport, then that household in 1810 was most likely living in what subsequently became Davenport. The likelihood is strengthened if in 1820 there was no household with the same first and last name still living (in this example) in Kortright. The likelihood is weakened in the opposite case if a household with the same first and last name (or even the last name only) still resided in Kortright. But what if the 1820 household is headed by the namesake son of the 1810 household head, now living at a different location, and the 1810 household has by now dissolved or is now headed by a different family member?

What one must remember in judging the various “likelihoods” (probabilities) is that movement of families in the early days was frequent and difficult to track. The turnover of a town's population between two decennial censuses was high. Families would give up on one farm, move to another elsewhere, perhaps in a nearby town and often beyond. A household would be joined by relatives with the same last names settling nearby, or would see sons (sometimes with an identical first name) establishing new households in the vicinity.

Fortunately, we are not trying to trace individual families exactly but only whether a particular household in one year lived on a particular section of land that was later part of a different town. Thus if the “John Doe” household of one year becomes the “John Doe Jr.” household (where the “Jr.” does not appear in the census listing) even at a slightly different location in a subsequent year, we are unconcerned as long as the two locations are in the same section of transferred land.

There is one further clue that can strengthen or weaken the belief that a family lived in a certain area. The census takers recorded a number for each household showing the order in which the household was visited. Because of travel and time constraints, household along the same crude road or path, or in the same general vicinity, tended to be visited in sequence.

This of course was by no means always the case. Census enumerators could have begun different days in different places or gone back much later to visit families missed earlier. Nevertheless, if households can be identified with some certainty from census data as belonging in a particular location, there is a greater chance that other households visited about the same time (as indicated by the sequence of enumeration numbers) would have lived in the same vicinity.

This last clue leads to the definition of a “cluster” of households likely to have lived in the same general area. The cluster consists of a sequence of enumeration numbers accompanying or interspersed with of two or more households believed on other evidence to have been residing in the area of interest. Where to draw the boundaries for such a cluster is of course the question. We have answered with a fairly narrow definition. Assigned probabilities are fairly high (that is, between 0.8 and 1) for all households where census evidence supports the location. Households visited immediately before or following this group could also be in the same locality but have been assigned a much lower “likelihood,” namely a probability of 0.25 or 0.5. (This should become clearer in the discussion below of the Block II clusters in the attached spreadsheet table.)

The assigned “probabilities” for various combinations of circumstances are spelled out in Table A. The exact probabilities chosen for each set of circumstances and clues are based on “informed judgement,” that is to say the investigator's hunch or best guess. They are set forth in such detail in order that other investigators may arrive at their own judgements and, if desired, rework the longer tables at the end of this note using an alternative set of probability estimates. Even for the estimates reported here, some probabilities for individual households were adjusted based on the ­knowledge and judgement of Davenport Historian Emeritus, Mary S. Briggs.

Table A

Assumed Probability of a Household in a Kortright (1800 or 1810) or Maryland (1810) Federal Census Residing in the Part of Kortright or Maryland Transferred in 1817 to the New Town of Davenport

(y = yes; x = no, none)

Table A tells us, for example, that if the same family name and first name appear in both the 1810 Kortright census and the 1820 Davenport Census (“y” in columns 2 and 3), that household will be given a probability of 1.0 if, and only if, two further conditions are met. These are that (a) the household is part of a “cluster” believed from census evidence to have resided in that part of

Kortright which subsequently became Davenport (“y” in column 6) and (b) there was no family with the same first and last name reported in the 1820 Kortright census (“x” in column 11). In the case of the 1810 Maryland census (where first names are replaced by initials), the 1810 Maryland initial corresponds to an 1820 Davenport first name (“y” in column 4 instead of 3), then the assigned probability is reduced to 0.95.

If the same family does not appear to be part of an identified Davenport-associated cluster, that is, it does not appear closely associated (judging by its enumeration number) with families believed to be in an area destined for Davenport (“x” in column 7), then its assigned probability falls to 0.90. Finally for this particular example, if the family is part of a designated cluster but a household with the same last name is found in the 1820 Kortright census listing (“y” in column 10), then the assigned probability falls further to 0.8. (To see how these examples are reflected in Table A, observe the lines immediately opposite the probabilities 1.0, 0.95, 0.9 and 0.8.)

The probabilities and circumstances from Table A were used in estimating the Kortright (1800 and 1810) and Maryland (1810 only) origin of Davenport's 1820 population. The method was not applied to Maryland's 1800 population because Maryland in 1800 was still part of the much more extensive town of Cherry Valley. Unearthing the household names and enumeration numbers Cherry Valley residents in 1800 and matching them with 1810 Maryland and 1820 Davenport residents would have been possible but was beyond the capacity of the investigation.

A slightly different method was used for identifying 1820 Davenport households subsequently lost to Oneonta. In this case we have several published reminiscences giving the names of “early Oneonta” settlers across the Susquehanna and near the mouth of the Charlotte River. These were used to assign probabilities of 1.0 to 20 households in 1820 and 20, not necessarily the same, in 1830. (While the land transfer to Otsego County may in fact have occurred prior to 1830, the 1830 census takers were not informed and all of the residents in question were counted in the 1830 Davenport census.) There was less clear evidence for another group of 15-18 households, and these were assigned probabilities of 0.5. For this latter group, in other words, the assumption was that about one-half became Oneonta residents and the other half lived on land that remained in Davenport.

The method for identifying names and estimating “likelihoods” for those living on land lost to Oneonta is described more fully in “Davenport's Population in the Early Years.” Assigned probabilities are shown in Table A of this paper.

Table B, below, shows the end result of the several calculations, and the footnotes provide further details. As noted earlier, two overlapping population series are shown. That for 1800-1870 gives Davenport's population for the former Kortright and Maryland areas minus that subsequently lost to Oneonta. The 1870-2000 population series also excludes those persons on land lost to the town of Meredith in 1878.

To encourage other researchers to examine and hopefully improve upon these estimates, the names of the possible Davenport-to-Oneonta transferees are shown in the paper cited above. The much longer table at the end of this note lists all households in Kortright (1800 and 1810) and Maryland (1810) living on land likely to be included in the future town of Davenport (including that land in the Wallace Patent subsequently annexed by Oneonta).

Table B

Adjusted Davenport Population Totals, 1800-2000

Source: Federal decennial Censuses of Population except where indicated by footnote.

a Pre-Davenport. Kortright inhabitants in the area of the future Davenport estimated from the Kortright census, but Fitches Patent inhabitants guessed at about 50, based on 1810 estimate of 95. (Fitches Patent in 1800 was part of Cherry Valley.) Estimate excludes Wallace Patent residents (approximately 72) in lots later transferred to Oneonta.

b Pre-Davenport. Estimates based on 1810 Kortright and Maryland censuses. Excludes Wallace Patent residents (approximately 163) in lots later transferred to Oneonta.

c From first Davenport census but excludes Wallace Patent residents (approximately 174) in lots later transferred to Oneonta.

d Excludes Wallace Patent residents (approximately 180) in lots transferred to Oneonta, all of which were included in the 1830 Davenport census.

e Excludes 150 residents from the Houghtalling Hollow area in south-west Davenport annexed by Meredith in 1878. The 1870 alternative estimate is consistent with years following while the census 1870 total is consistent with earlier years.

The longer spreadsheet table or worksheet (Table D) at the end of the present note contains three blocks of names. The first, Block I, consists of Davenport census names from the 1820 census and Kortright names from the 1800 and 1810 censuses. First and last names of each household head, arranged alphabetically, are shown along with the number of family members and the estimated probability of having lived on land allocated to the future town of Davenport. The block will be examined in more detail below.

The second block is similar to the first except that 1810 Maryland households are shown and are arranged by the 1810 census enumeration number instead of alphabetically. For Maryland in 1810, only the initials of the first names are available. The final block of names contain all Davenport 1820 households not found in the table's first two sections.

This final table is not intended as a finished product for publication. It is merely a rearrangement of the analyst's original spreadsheet and may at first seem confusing to the reader. The place to begin an examination is with the list of last names in column C. This list originated with Davenport's first, 1820 census names. For these known Davenport households, the first name of the household head appears in column B and the 1820 census enumeration number is in column A.

The next step of the investigation leading to the compilation of Block I was to examine the Kortright 1810 households for identical or similar names to those found in 1820 Davenport. Thus the Jacob Banner household of Davenport in 1820 (census number 169) was exactly matched by a family of that name, No. 109, in the 1810 Kortright census. Rather that repeating the first name, Jabob, the analyst chose merely to note “same” in the first-name column for 1810, column F of the spreadsheet. “Same” was also used in a later step when the identical first name for the household head was found in the 1800 Kortright census. 1800 first names appear in column L.

Next, the 1820 Kortright census was examined to learn of any other Banner families still living in that town. There were none in this case (recorded with an “x” in column D), and the assumption was then made of a “100%” likelihood that the 1810 Jacob Banner family of Kortright was either the same or the direct antecedent of the 1820 Jacob Banner family in the new town of Davenport. As already noted, a further assumption was made that the household did not move physically between the two dates, or if it did move it remained nearby on eventual Davenport land. This cannot be proven and leaves a small margin for potential error.

The assumption of “100% likelihood” for the 1810 Jacob Banner household shows up as the number “1” in the 1810 probability column, column H. When multiplied by the 1810 household size, 2 persons as shown in column G, the result of 2.0 is our estimate of the probable number of persons contributed by the Banner household to the 1810 Kortright total of those living on land that later became Davenport.

A later step was to repeat the search for Banners in the 1800 Kortright census. There was a Banner family, but no known Jacob Banner (who could have been alive but not yet a household head in 1800 Kortright, and therefore unlisted in the census). For the 1820 and 1810 Jacob Banner household, a question mark is shown in the 1800 column K of the worksheet, indicating that a Jacob Banner could have lived in 1800 Kortright. If there had been a Jacob Banner household in 1800, the census number for that household would be given in column K. If no Banner at all had been found in 1800 Kortright, column K would contain an “x.”

Continuing with the Banner family example, the worksheet shows two other Banner households, headed by a John and a Wilhelmus, living in 1820 Davenport. Judging by the census enumeration numbers in column A, John may have lived close to or even next door to Jacob while Wilhelmus lived further away. There was a Wilhelmus Banner reported in 1810 Kortright with a household size of 5. For the same reasons as for Jacob, it was assumed that Wilhelmus Banner in 1810 lived with 100% certainty in what became Davenport in 1817, hence the probability of “1” in column H.

There was no John Banner household recorded for Kortright in 1810 but there was a Frederick Banner (household size of 3), living close to or adjacent to Jacob Banner (from the census identification numbers in column E). This same household name also appeared in the 1800 Kortright census. The Frederick Banner data could by rights have been reported on a separate line of the spreadsheet table. To save space and since the Frederick Banner household was not found in either the Kortright or the Davenport in 1820, the Frederick Banner information was merged with the 1820 data on John Banner. This is not meant to imply a necessary connection between the two households.

Although it is almost certain that the Frederick Banner household of 1800 and 1810 lived on land that was to become Davenport, the household fails to meet one Table A criterion specified for “100 %” certainty. This was that there was no Frederick Banner household listed as 1820 residents of Davenport. This is to say that there is a (very) small probability that a Frederick Banner still existed but was now located elsewhere and that the earlier Frederick Banner family had not in fact lived on land that became Davenport.

Other criteria were met. One was that the Frederick Banner household was part of a cluster of sequentially-numbered households believed with varying degrees of certainty to have been living in what became Davenport. (The notion of a “cluster” will be illustrated further below in the discussion of Block II.) Another criterion was that there was no Banner family reported from Kortright in 1820. Because in this case only two of the three criteria were met, the Frederick Banner household was assigned, perhaps arbitrarily but rules are rules, a probability of 0.95 in both 1800 and 1810.

(Other researchers, as already urged, are free to make their own judgements and to bring other evidence—property deeds, church and school records, burial data, wills, etc—to bear in modifying these probabilities. One strong clue, for example, would be if a Frederick Banner of the right age died between 1810 and 1820 and was buried in a Davenport cemetery. The change of one household probability from 0.95 to 1.0 in this example will not greatly change the final results, but a number of such changes could be more important.)

Let us digress briefly from this rather tiresome and laborious introduction to the worksheets and analytical method to point out how Table D can be used to make other informed guesses about the early Banner family of Kortright and Davenport. A logical interpretation of the census data is that one of the early (1800) households in what later was to become Davenport was headed by Frederick Banner. In that Banner family of eight, we know from the reported census data that there was the household head and a female, presumably Frederick's wife. The older male was of age 45 or more and the female, between 26 and 45. In addition, the household contained three younger women, probably daughters, and three younger males, almost certainly the sons Wilhelmus, Jacob and John. We can guess that Wilhelmus was the oldest boy since in 1810 he had a larger family than Jacob and that John was the youngest since he did not become a household head until after 1810.

Another point to note about the spreadsheet table is the large number of households that do not have an 1820 census identification number in column A. These are families who showed up as part of a Davenport-probable cluster of names in either 1800 or 1810 but who had presumably moved elsewhere or had dissolved by 1820. All of these households in earlier years have a “Davenport probability” less than 1.0 because there was no 1820 Davenport household headed by that particular person.

Thus, there was a Uriah Adams household in both 1800 and 1810 Kortright that was listed by the census within a cluster of others identified with the later Davenport. There were no Adams household listed in either the 1820 Kortright or Davenport censuses, and the earlier Adams household is thus given a probability of 0.9.

A neighbor of Frederick Banner in 1810 (but not in 1800) was Ammy Cleveland. There was also a Curtis Cleveland household in both 1800 and 1810, but the location, judging by the census enumeration I.D. in both years, was not in a Davenport-related cluster. There were no Clevelands reported in the 1820 Davenport census, but Curtis Cleveland's name appeared in the 1820 Kortright census, thus eliminating the likelihood of that household having been living within Davenport's later boundaries. Ammy Cleveland's household, however, was assigned a Davenport probability of 0.8, solely because of its apparent proximity to Frederick Banner and other prospective Davenport residents in 1810.

Another digression, this one related to the difficulties of these sorts of investigations. In rechecking the Cleveland entries for the above illustration, it was discovered that the 1810 existence of the Curtis Cleveland household had not been picked up in the earlier compilation. It had been assumed that the Ammy Curtis household had followed that of Curtis Cleveland (even though the 1800 census I.D. number of the latter family should have raised a warning flag), and the 1800 Curtis Cleveland household had been assigned a Davenport-likely probability of 0.8. This earlier mistake has now been corrected, but it illustrates the problems of juggling names from several censuses for which name indexes (themselves containing occasional errors and misspellings) have been compiled, but for last names only. It is a warning than only another thorough checking could catch whatever further errors may exist in the present worksheets.

Finally in Block I of the spreadsheet table, the reader may note that instead of an 1820 Kortright census I.D. number for Ammy Cleveland, there is only a question mark in column D. This signifies that the name “Cleveland” but not “Ammy Cleveland” is found in that census. If there had been no Cleveland's at all in the 1820 Kortright census, the question mark would have given way to an “x.”

Turning to the Maryland township names in Block II of the same spreadsheet-based Table D, it will be observed that initials only are available in place of 1810 first names and that no 1800 data are shown. The 1800 material could with effort be compiled from the census records for the extensive township of Cherry Valley, but this chore has not yet been undertaken. Instead, a guess has been made that the 1810 population of that part of Fitches Patent that subsequently was incorporated in Davenport had grown over the previous ten years at about the same rate as had the Korteight population in areas that became Davenport. This suggested an 1800 Fitches Patent population of about 50. (In 1810, the probable total was perhaps 95.)

Block II, in addition, is arranged not alphabetically but by the 1810 census enumeration number. This arrangement permits a look at how clusters have been developed for Davenport-likely households. As in Block I, the first step was to begin with a listing of 1820 Davenport names. In the case of Davenport's Samuel Brown in 1820, for example, the town of Maryland in 1810 had an “S. Brown—deemed not likely to have been a future Davenport resident because Maryland had several Brown families in 1820, including at least one Stephen Brown. Other Davenport Browns are mentioned further down in the Block II listing, but again with no satisfactory match in 1810 Maryland.

It is not until we get to Maryland's 1810 census number 136 (shown in column D) that names begin to match more closely. That number is for an 11-person household headed by “C. Shaver.” There was no C. Shaver in Davenport ten years later, but there were three other Shaver households, headed by John, Peter and William. Furthermore, there were no Shavers listed in 1820 in the town of Maryland. Maryland's C. Shaver family, therefore, was assumed to have been antecedent to the Davenport Shavers and assigned a ninety-percent probability of having resided on land that later became Davenport. If there had been an exact match of first names, the probability would have been 1.0. A match between a first initial in 1810 and the 1820 Davenport first name would suggested an 1810 probability at 0.95.

The C. Shaver family was followed in the 1810 census listing by three other Maryland families (Strader or Strayder, Craft or Croft, and another Craft), all of whom matched 1820 Davenport households to some degree. The 1810 households numbered 136 through 139 were considered to compose a likely Davenport-bound “cluster” of neighbors. But what about other households coming before or after this group in the census enumerator's list? We have no clear evidence for these one way or another. The household names of the nearest potential “neighbors” did not appear on either the 1820 Davenport or Maryland censuses. We have therefore guessed at a fifty-fifty chance that the most closely adjacent families (numbers 135 and 140) were located on future Davenport land. (This is to say, we estimate that one of the two households lived on such land.) Household number 141, following the cluster, was assumed likely to have remained in Maryland since a family of that name appeared in the 1820 Maryland census. Preceding household numbers 133 and 134 (but not 132) could have been Davenport-situated as easily as number 135 except for their somewhat greater numerical (and hence locational?) distance from the No. 136-139 cluster.

In the case of the next identified cluster of three Bresee and one Spoor or Spurr family, 1810 numbers 151-154, the following adjacent number 155 (Dubois) was assigned a likelihood rating of 0.5 (no 1820 evidence one way or another). The last name of the preceding household (J. Gunn, census number 150), in contrast, turned up in the 1820 Kortright census. The 1810 Davenport probability for the adjoining J. Gunn household was therefore lowered to 0.25.

In some cases, households with 1820 Davenport last names were found in both the 1810 Kortright and Maryland censuses. Smith and Olmsted (sometimes Olmstead) are two examples. There were 10 Smith families in 1820 Davenport (Elijah, Harvey, Jeremiah, Jeremiah Jr., John, John J. John P., Lydia, Peter P. and Seymour). Six additional Smith households, including a Benjamin and a William, were listed in the 1820 Kortright census.

Four Smith families were found in 1810 Maryland (E., H., and two J.s). None of the Maryland Smith names appeared in the 1820 Maryland census, but there was an Ichabod Smith (census number 85). In 1810 Kortright there were 13 Smith households, headed by Benjamin, Benijah, Elizabeth, Henry, James, Jeremiah, Jenner, John H., John J., Noah, two Peters (household numbers 204 and 345), and William.

Using membership in an 1810 cluster as further evidence, the thirteen Kortright Smith households from 1810 were assigned Davenport probabilities ranging from zero (Benjamin and William) to 0.5 (James and Noah) and on up to 0.95 (Peter) and 1.0 (Jeremiah and John J.). None of the Maryland Smith families of 1810 were assumed to live in the Fitches Patent area transferred in 1817 to Davenport.

This reveals another problem with the analysis when it comes to specific households. The James Smith household of 1810 Kortright, on the periphery of a cluster and thus assigned a probability of 0.5, could have easily been replaced by the J. Smith household of Maryland given a zero probability. What is hoped is that while the location of any one specific household may be in error to one degree or another, the probability estimate of the total population living in a broad area will be more or less accurate.

There were fewer Olmsteds than Smiths in 1810 Kortright and Maryland, only seven Olmsted families as opposed to seventeen Smiths. In the case of the Olmsteds, however, there were none reported in the 1820 censuses for those two towns. This implies that all Olmsted families either became part of the new Davenport, perhaps moved away or otherwise disappeared. 1820 Davenport was home to five Olmsted families, headed by Anson, Darius, John, and two Stephens. This was a drop of two from the combined 1810 Kortright-Maryland total, suggesting that one or two household had either disbanded or had moved on.

In 1810 Maryland there was an A. and a D. Olmsted, possibly standing for Anson and Darius. Maryland also one other D., an S. (Stephen?), and a W. Olmsted. Harvey and Noble Olmsted headed the two Kortright households. All of the five Maryland Olmsted households seemed to have lived close to one another (census numbers 184-186 plus 188, 189) and were given a Davenport probability of 0.95. Harvey Olmsted in 1810 lived, judging by census numbers, in the middle of a ten-household Davenport-associated cluster (nos. 153-162, including four Turner households). The Harvey Olmsted household was also assigned a Davenport probability of 0.95, following the rules of Table A. The Nobel Olmsted family, census number 500 and part of no Davenport-associated cluster, received a probability of zero.

Even if one of the six 1810 Olmsted households with a 0.95 probability had moved away or had disbanded by 1820 and thus did not reside in the 1820 Davenport, we are interested only in whether or not their 1810 residence was on land that became Davenport. In each case of these six families the probability seems highly likely, though of course not completely certain.

The third and final block of the attached spreadsheet table shows the names, in alphabetical order, of 1820 Davenport families that do not appear in Blocks I and II. These household names have been included for the sake of completeness.

Table C, following, gives the end results of the Kortright and Maryland household search and probability calculations. 910 residents of Kortright in 1810 are estimated to have lived in the area lost to Davenport in 1817. On a comparable land area basis, the Kortright population for 1810 would be reduced from 2993 to 2083. For Maryland, the Fitches Patent residents living in 1810 on what was to become Davenport territory totaled an estimated 95. The 1810 Maryland population for a comparable, 1810 to 1820 land area would therefore be reduced from 1103 to 1008.

Following Table C are the nine pages of the spreadsheet Table D discussed so laboriously above.

Alan M. Strout

For the Davenport Historical Society

January 2004

Table C

Estimate of Probable Pre-1820 Kortright and Maryland Residents

Residing in what Later Became Davenport

Table D

Names in 1820 Davenport NY Census Corresponding to Names in 1810 and 1800 Kortright Censuses and 1810 Maryland (Otsego Co.) Census, with Possible Antecedent Families in pre-1820 Kortright and Maryland Censuses

[Source: SuperCalc Spreadsheet compiled by Alan Strout from decennial Federal censuses]

I.   KORTRIGHT CENSUSES NAMES, ARRANGED ALPHABETICALLY