The data we have assembled may be regarded as a convenience sample, since we have used all the data we could assemble given the resources at our disposal. As a result, in our sample, the proportion of residential properties in different price ranges varies between time periods, as shown by Figure 2.3.
As the general level of residential property prices rose, many properties shifted from a lower to a higher price range. Accordingly, the proportion of properties in our sample that sold for less than 25,000 kronor declined from the late 19th century up to the First World War and the proportion in the upper price ranges increased. However, there also seem to be compositional changes between time periods in our sample of properties that cannot be explained by the evolution of the overall level of property prices. During the First World War, the annual average price of properties in our sample increased much more than the overall level of prices as measured by our price index. This was due to a disproportionate increase in these years in the share of properties that sold for more than 100,000 kr. The proportion in this price range in our sample is also larger in the 1950s. In the inter-war period, on the other hand, the average price in our sample fell even though the general level of residential property prices rose, because in these years the proportion in the lowest price ranges grew. A probable explanation for this is the extension of the geographical areas from which our properties are sampled. Properties outside downtown Göteborg were gen- erally smaller. For example, in neighbourhoods such as Änggården the new proper- ties consisted primarily of owner-occupied one- and two-family houses.
Figure 2.3. Proportion of residential properties in different taxation classes, centred moving five-year averages, 1875–1957
Sources: See Table A2.2. 1875 1876 1877 51,9179642 22,8771403 15,7532556 9,45163984 1878 46,7560728 24,8198377 16,1865191 12,2375704 1879 43,3976768 25,4380499 17,8573629 13,3069104 1880 43,7185324 24,7517754 17,2958656 14,2338267 1881 45,1743882 23,8797033 16,1859556 14,7599528 1882 51,2891423 20,830523 15,4810376 12,399297 1883 55,559736 18,363887 15,4994453 10,5769317 1884 52,2660852 19,7924584 15,8169057 12,1245507 1885 51,3474712 19,8622946 15,5751651 13,2150691 1886 52,8157443 18,4614914 15,6105065 13,1122579 1887 50,3676568 19,5106717 16,5154245 13,6062469 1888 48,1725349 22,8216473 16,3507904 12,6550274 1889 46,3614238 18,3327584 20,9730126 14,3328052 1890 46,6561519 20,3128332 21,6247353 11,4062797 1891 41,4926407 20,9898268 23,9005051 13,6170274 1892 39,836963 21,8249917 23,7774281 14,5606172 1893 40,3632788 20,285518 24,2116387 15,1395645 1894 45,9288702 23,4123997 19,9836817 10,6750484 1895 43,5046278 21,5942179 19,9079241 14,9932302 1896 40,0241083 24,8756032 19,4634796 15,6368089 1897 39,3464526 22,8182161 22,7754455 15,0598858 1898 35,8201369 24,6076898 23,5912349 15,9809384 1899 30,2661397 29,3996487 23,3902064 16,9440053 0 10 20 30 40 50 60 70 1875 1885 1895 1905 1915 1925 1935 1945 1955
Percentage share for properties with taxation values less than 25,000 kr., 3-year centred moving average
Percentage share for properties with taxation values between 25,000 kr. and 50,000 kr., 3-year centred moving average Percentage share for properties with taxation values between 50,000 kr. and 100,000 kr., 3-year centred moving average Percentage share for properties with taxation values larger than 100,000 kr., 3-year centred moving average
Does the change between periods in the composition of types of residential property affect our estimated price index? It might if the evolution of the sales price appraisal ratio differed between the various price ranges. The sales price appraisal ratio was often higher in the lower price ranges, as shown by Figure 2.4. However, what mat- ters most for the overall price index is the rate at which sales price appraisal ratios changed in the various price ranges. To get an idea of the extent to which a changing composition of properties in the various price ranges affects our estimated overall price index, Figure 2.5 presents our equal-weighted overall sales price appraisal ratio index together with an average of the sales price appraisal ratio index series for the different ranges of taxation values depicted in Figure 2.3. In the latter, the weights of the various taxation classes in the overall index are always the same, whereas in the former the composition of taxation classes varies over time. As can be seen, there is a difference between the series but it is small before the 1950s. Prices for properties in the upper price ranges rose more slowly than in the lower price ranges. Consequently, the larger proportion of properties in the upper price ranges tends to lower our equal- weighted index for the 1950s.
Figure 2.4. Sales price appraisal ratios for residential properties in different taxation classes, 1875–1957
Sources: See Table A2.2.
Sales price appSales price appSales price appSales price appraisal ratio for properties with taxation valu
1875 1,37 1,18 1,27 1,24 1876 1,31 1,14 1,42 1,19 1877 1,3 1,31 1,09 1,08 1878 1,27 1,15 1,11 1,25 1879 1,11 0,97 0,96 1,04 1880 1,15 1,01 0,98 1,03 1881 1,05 1,03 1,11 1,07 1882 1,02 1,09 1,07 0,96 1883 1,12 1,02 1,12 1,05 1884 1,17 1,04 1,03 1,18 1885 1,03 0,97 0,94 0,96 1886 1,1 0,95 1,04 1,08 1887 0,93 1,08 1,02 1,04 1888 0,88 0,98 0,98 0,86 1889 0,94 1,04 0,95 1,11 1890 1,09 0,97 1,01 1,12 1891 0,98 1,12 0,98 0,96 1892 0,95 1,07 1 1,03 1893 1,05 0,92 0,9 1 1894 0,93 0,93 0,95 0,95 1895 0,91 0,97 0,93 0,93 1896 1,02 0,87 0,97 1,03 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 1875 1885 1895 1905 1915 1925 1935 1945 1955 Sales price appraisal ratio for properties with taxation values less than 25,000 kr.
Sales price appraisal ratio for properties with taxation values between 25,000 kr. and 50,000 kr. Sales price appraisal ratio for properties with taxation values between 50,000 kr. and 100,000 kr. Sales price appraisal ratio for properties with taxation values larger than 100,000 kr.
Figure 2.5. Price indexes for residential properties in Göteborg (1957=100), equal-weighted index and unweighted average of equal-weighted indexes for different taxation classes, 1875– 1957.
Sources: See Table A2.2.
Figure 2.6. Price indexes for residential properties in Göteborg 1875–1952 (1912=100), calculated by different methods.
Sources: See Tables A2.2 and A2.3. 74 75,9572123 79,7718102 81 82,8614707 87,4590065 75 66,0340281 71,7487323 78 74,1559409 84 79,5328534 82,3657904 40 90 140 190 240 290 1875 1885 1895 1905 1915 1925 1935 1945
Price index, sales appraisal ratio method Price index, repeated sales method (Bailey et.al) Price index, repeated sales method (Case & Shiller)
Residential price index, unweighteResidential price index, equal-weighted
1875 76,33626336 77,61300028 1876 74,02304326 78,34108825 1877 81,03210017 82,22442564 1878 79,07164614 82,40014206 1879 67,30892192 70,22508154 1880 74,81092603 77,66216825 1881 75,51668948 80,04143998 1882 74,10516258 77,70858323 1883 78,46678072 82,54228223 1884 82,06617433 83,96766391 1885 80,62641688 83,15742435 1886 83,85147356 89,16923826 1887 80,62641688 87,13867868 1888 74,0150507 81,1064324 1889 80,59416632 87,82235484 1890 85,52850303 91,17328449 1891 80,60239021 87,64625473 1892 81,41655577 87,44342225 1893 80,60239021 83,10853633 1894 78,74609274 84,34827219 0 50 100 150 200 250 300 350 400 1875 1885 1895 1905 1915 1925 1935 1945 1955 Residential price index, unweighted average of various taxation classes Residential price index, equal-weighted
A view of Göteborg from the Liseberg amusement park.
Source: Wikipedia.
Bangatan in Majorna, Göteborg, in the mid-1930s.
In conclusion, the changing composition of the types of property included in the sample has some effects on the price index series. However, the overall picture does not seem to be greatly affected by compositional changes, at least before the 1950s. 2.2.4.3. Price index series calculated by the repeated sales method
Since our database includes information on the name of each property, it has been possible to identify repeated sales of the same property. Slightly more than 3,000 of the observations were repeated sales, containing sales data on roughly 1,400 proper- ties. Since this dataset is based on identifying properties built on a particular piece of land, there is obviously room for mistakes in identifying repeated sales. It may, for example, be the case that a building had undergone extensive changes or that an entirely new building had been erected on the same piece of land between two sales, especially if many years passed between the sales. In order to minimize this risk and get rid of otherwise extreme observations, we excluded a pair of sales from the dataset if the sales price appraisal ratio is larger than 3 or smaller than 0.5. Furthermore, we excluded a pair of sales if the change in the taxation value between them was more than a three-fold increase or more than a 50 per cent decrease. Observations where the interval between two sales was longer than 25 years have also been excluded from the repeated sales dataset.
Two price index series calculated by the repeated sales regression method are pre- sented in Table A2.3. The first is calculated according to the original method of Bailey et al., the other according to the modification of this method proposed by Case and Shiller.31 For both series we have included an intercept in the regressions.
The two series are displayed in Figure 2.6 together with the series calculated accord- ing to the sales price appraisal ratio method. First we may note that the general con- tour of development is similar in all three series. A difference between the series is that the price rise is slightly faster up to the First World War according to the two repeated sales series, after which the sales price appraisal ratio series catches up with them. In the 1930s, prices are more or less unchanged according to the index calcu- lated by means of the sales price appraisal ratio method, while they decline according to the repeated sales series because of a larger fall in 1931–1933. This seems to be a sample issue, since calculating sales price appraisal ratios on the repeated sales sub-sample gives approximately the same fall in these years as in the series calculated according to the method of Bailey et al.
Another difference is that the repeated sales series gives more pronounced year-to- year fluctuations, especially from the late 1920s onwards, which may be related to the limited size of the sample used to estimate the repeated sales regressions. But it also has to do with methodological differences; the sales price appraisal ratio method yields less volatility even when calculated on the same sub-sample.
The series calculated according to Case and Shiller’s method tracks the series cal- culated by means of the Bailey et al. method closely until the 1930s, although the 31 Bailey et al. (1963); Case and Shiller (1989).
former method generally yields slightly lower index values. This difference widens in the 1930s and 1940s, which indicates that the intervals between sales that make up the log price ratios in the sample become longer over time.
Using different methods to calculate the price index numbers serves as a check on the series’ validity. The similarity of the results, even though the series are calculated on different samples, is reassuring. We prefer the series calculated by the sales price appraisal ratio method since it is based on a larger sample. As argued by Bourassas et al.,32 it should also be preferred for methodological reasons.
2.2.4.4. A comparison of the residential property price index for Göteborg with