10 BULLETIN 1224, U. S. DEPARTMENT OF AGRICULTURE. 
Not only are there large variations from group to group, but there 
are smaller variations from county to county within each group. But 
the variation in the county ratios is not as great, on the average, as 
the variation in the group ratios. That is, there is greater variability 
in these ratios between different regions of the country than between 
different counties in the same region. 
As stated above, average cash rents constitute the best obtainable 
measure of land income. However, when land income is measured 
in this way, the greatest variation is found in its relationship to the 
sale price of land. Are the relationships for the different areas as 
expressed by these ratios true ones? Is there as much real varia- 
tion from county to county as is shown in figure 3 and by Table 1 
in the Appendix? Are the variations from group to group repre- 
sentative of actual facts ? Are the variations real expressions of econ- 
omic forces, or are they to be explained by inaccuracies in the data ? 
It is essential to get an accurate measure of the relationship of land 
income to land value and then to discover the economic explanation 
of it. But before the data already presented can be considered as 
showing the true economic relationships it is necessary to find out 
whether or not the variations in the ratios are due (1) to inaccuracies 
in the data, or (2) to erratic fluctuations in 1920 in either rents or 
land values. After testing the data for accuracy it will be possible 
to discuss the economic causes that determine the ratios of rent to 
value and the variations in these ratios, if any exist, which are not 
accounted for by inaccuracies or erratic fluctuations in either cash 
rents or land values. 
ACCURACY AND REPRESENTATIVE CHARACTER OF THE DATA. 
ACCURACY OF CENSUS DATA ON LAND VALUES. 
In the first place, are the land values reliable ? The census valua- 
tions are based upon estimates made by the farm operators, and as 
such may be suspected of two kinds of error. (1) They may not 
accurately reflect the variations in productivity of land from county 
to county because of errors in making estimates. (2) The whole 
level of values may be biased above or below the actual average 
sale prices. An upward bias may occur because of the optimism 
of the farmers in 1919. This was the last year of the land boom, 
and the farms that sold at the highest price were the ones most 
talked about, so that farmers on the average might easily have 
stated that their farms were worth more than they would actually 
have sold for. On the other hand, they might have placed a lower 
value on their farms than the actual average sale price, fearing the 
effect of high values on their taxes. This bias might have been 
different in different areas and, if so, would account for the differences 
in the ratios of rent to value in these areas. 
BIAS IN CENSUS LAND VALUES. 
To test the census real-estate valuations for bias, the county 
average values of land and buildings as reported by the census were 
compared with actual county average sale prices. 6 These differences 
6 Sales data for 1920 were obtained from the tax commissions of Minnesota, Wisconsin, and Indiana for 
those states; from T. A. Polleys, tax commissioner of the Chicago and Northwestern Railway Co., for 
counties in Illinois, Iowa, Nebraska, and South Dakota; and from the annual report of the secretary oi 
state of Ohio. 
