Accessible Unlicensed Requires Authentication Published by De Gruyter March 5, 2019

Housing Wealth Effects in Japan: Evidence Based on Household Micro Data

Masahiro Hori and Takeshi Niizeki

Abstract

Using micro data covering almost 500,000 Japanese households over the period 1983–2012, we examine to what extent household consumption responds to changes in housing wealth. To do so, we estimate the housing wealth of individual households by matching several official statistics. Employing cross-section and pseudo-panel-based regressions, we find that the marginal propensity to consume (MPC) out of housing wealth is approximately 0.0008–0.0013 for nondurable consumption and 0.0059–0.0082 for total consumption. We further find that the consumption response of older households is larger than that of younger households, which is consistent with the pure wealth effects hypothesis.

JEL Classification: D12; D31; E21

A Appendix: Estimation process for land and house values

As mentioned in Section 2.2, since the FIES does not contain information on households’ housing wealth, we estimated it by matching the FIES data with several official statistics. Housing wealth consists of two parts: a house (building) and the land on which it sits. In this appendix, we provide a detailed description of the estimation process employed to obtain land and house values, respectively.

Land values

We calculate the value of the piece of land a household owns by multiplying the price of residential land in the area where the household lives by the area of land owned by the household. Suppose that a household reports that the piece of land on which its house sits measures 200 m2. In theory, all we have to do is to multiply this with the price of residential land where the household lives. Unfortunately, the FIES does not contain such information on the price of land, and the exact location where the household lives is also unclear. To deal with this issue, we first determine the approximate address where a household lives (to the greatest level of precision possible using multiple official statistics) and use the price of residential land at the closest survey location in the LMVP as the relevant land price. More precisely, we employ the following strategy.

First, although the FIES does not provide the exact address where households live, it provides information on the prefecture number, city number, and two enumeration district (chosa-ku) numbers taken from the Population Census from which households are chosen randomly.[32] An enumeration district covers approximately 50 households and its maximum area is restricted to 1 km2. Figure 4 illustrates how we identify the approximate address where household X lives in City A in the case that the address of household X belongs to either the enumeration district number 1571 or 1572. Since again an enumeration district contains approximately 50 households and the maximum area of a district is restricted to 1 km2, we can pinpoint the address where household X lives to the extent that it must be one out of 100 possible households and to within 2 km2.

Figure 4: Example of how we identify the approximate address of household X.

Figure 4:

Example of how we identify the approximate address of household X.

Unfortunately, no further information is available to identify which of the 100 households was actually surveyed (household X). However, we need the address where household X (presumably) lives to estimate the value of the land that household X owns. To assign an address to household X, we employ the following algorithm. First, among the two enumeration district numbers, we choose the smaller number (in the case of Figure 4, it is 1571). Second, if the selected enumeration district (1571) contains several addresses as shown in Figure 4, we pick the address presented at the top of the list. In this way, we obtain the approximate address at which household X lives, that is, 1-1-1 City A, in this case. It should be kept in mind that our methodology of identifying the approximate address at which household X lives is not immune to measurement error in the sense that the actual address may be, say, 1-1-3 City A, of enumeration district number 1572. However, we believe that under the circumstances this is the best methodology to determine the approximate address of household X.

Having identified the approximate address at which household X lives (1-1-1 City A), we next find the closest survey location in the LMVP provided by the Japanese government. The LMVP annually reports land prices per square meter (of 5,000–20,000 residential sites) all over Japan as of January 1 and is widely used as a benchmark for land transactions. Suppose there are six survey locations in City A, which are represented by the small squares in Figure 5, with the price of residential land in each location shown in parentheses. In this case, the closest survey location from the approximate address where household X lives is the one located 700 m away, where the price of residential land is 150,000 yen (about 1,500 U.S. dollars) per square meter. We therefore assume that the price of residential land in 1-1-1 City A, where household X owns its home, is also 150,000 yen per square meter. Finally, multiplying the land price per square meter (150,000 yen) thus obtained by the land area that household X reports in the FIES (say, 200 m2) yields the value of land household X owns (30,000,000 yen).

Figure 5: Example of how we find the closest location in the LMVP from the approximate address of household X (1-1-1 City A).Notes: The squares represent hypothetical survey locations in the LMVP, with the figures in parentheses showing the price of residential land per square meter in the survey location.

Figure 5:

Example of how we find the closest location in the LMVP from the approximate address of household X (1-1-1 City A).

Notes: The squares represent hypothetical survey locations in the LMVP, with the figures in parentheses showing the price of residential land per square meter in the survey location.

Estimating the value of land that households own presents two potential pitfalls. The first of these concerns the absolute size of any measurement error in land values, since measurement error in an explanatory variable biases coefficient estimates for that variable toward zero. In our approach, we tried our best to minimize such potential measurement error to the greatest extent possible employing the procedure just described rather than using the average land price at the municipality level. The second potential pitfall concerns the relative size of any measurement error in land values between households residing in urban areas and those residing in rural areas. If the measurement error in land values for households residing in urban areas, whose heads tend to be young, is larger than that for households residing in rural areas, whose heads tend to be older, our main finding that the MPC out of housing wealth is larger for older than for younger households could be spuriously driven by the larger measurement error in land values for younger households.[33]

Regarding this second potential pitfall, note that there may be two different types of measurement error in estimated land values. The first stems from the fact that we may have misidentified the address at which household X lives. Unfortunately, there is no way to gauge the absolute size of this type of measurement error, since the actual address where household X lives is unknown. However, it is clear that the size of enumeration districts containing about 50 households is larger in rural areas than in urban areas, meaning that the measurement error in land values is bound to be larger for households residing in rural areas.

The second type of potential measurement error in estimated land values springs from the distance between the approximate address at which household X lives and the closest survey location in the LMVP. In general, the closer the survey location in the LMVP is, the smaller the measurement error in estimated land values is likely to be. To check that the distance in our sample is sufficiently short and that the distance is longer in rural areas, Table 10 presents summary statistics of the distance between households’ approximate address and the closest survey location in the LMVP by prefecture. To construct Table 10, we restrict the sample to homeowners, since we are now focusing on the measurement error in the value of land households own. The left-hand side of Table 10 provides descriptive statistics for the full sample consisting of all homeowners for the period 1983–2012. As can be seen, there is substantial heterogeneity in the mean distance across prefectures. For instance, while the mean distance is less than 0.5 km in Tokyo (#13) and Osaka (#27), it is more than 2 km in some rural prefectures such as Shimane (#32) and Ehime (#38). The mean (median) distance for Japan as a whole is 1.09 (0.53) km. Next, the right-hand side of Table 10 presents descriptive statistics when households living in towns and villages are excluded.[34] The figures indicate that excluding households living in town and village shortens the prefectural mean distance. (The mean distance for Japan overall becomes 0.7 km). This implies that there is also heterogeneity within prefectures in the sense that households living in towns or villages increase the prefectural mean distance. In summary, we need to bear in mind that while the mean (median) distance for Japan as a whole is 1.09 (0.53) km, measurement error in land values may be severer for households in rural areas, since there are fewer survey locations in the LMVP close to households’ approximate address.

Table 10:

Distance from households’ approximate address to the closest survey location in the LMVP by prefecture (kilometers).

PrefectureFull sampleExcluding towns and villages
MeanMedianStd. dev.NMeanMedianStd. dev.N
(kilometers)(kilometers)
1Hokkaido1.000.551.9712,3520.750.550.7710,438
2Aomori1.310.583.536,1900.700.580.585,473
3Iwate1.700.573.785,9981.050.542.025,442
4Miyagi0.920.442.065,9280.530.420.505,435
5Akita0.940.611.545,7290.940.611.545,729
6Yamagata1.370.912.266,5240.960.900.765,881
7Fukushima1.510.575.396,7610.660.560.546,250
8Ibaraki1.250.533.236,4740.560.510.335,892
9Tochigi0.680.540.846,3790.590.530.375,819
10Gunma1.080.731.416,4550.870.710.665,799
11Saitama0.530.391.2011,6260.440.380.3211,112
12Chiba0.950.431.8010,1260.780.421.389,635
13Tokyo0.380.360.2221,0920.380.360.2221,092
14Kanagawa0.430.380.3416,5250.410.370.2315,382
15Niigata1.220.544.067,9720.690.540.577,395
16Toyama1.360.623.347,1150.870.600.866,553
17Ishikawa1.150.562.875,7240.670.540.825,129
18Fukui0.740.600.626,3690.740.600.626,369
19Yamanashi1.230.741.775,1860.870.710.814,749
20Nagao1.030.661.597,7390.770.620.587,199
21Gifu1.230.621.846,7821.090.611.546,312
22Shizuoka2.070.677.209,5430.860.650.888,390
23Aichi0.580.461.0612,9750.520.460.3412,373
24Mie1.470.773.566,9910.940.770.676,450
25Shiga0.780.541.445,4500.620.550.414,982
26Kyoto0.670.402.016,1010.480.400.345,497
27Osaka0.420.370.2915,6710.420.370.2915,671
28Hyogo0.740.351.6610,9350.450.330.4210,250
29Nara0.920.462.715,0310.510.450.504,352
30Wakayama1.020.661.506,6890.840.650.686,221
31Tottori1.130.592.645,2210.710.530.574,506
32Shimane2.880.687.714,6700.870.641.014,114
33Okayama1.280.772.265,8070.970.741.095,235
34Hiroshima0.750.541.277,2140.690.540.586,765
35Yamaguchi2.351.981.636,9712.351.981.636,971
36Tokushima1.610.663.404,9630.740.630.524,312
37Kagawa0.790.620.665,2350.790.620.665,235
38Ehime2.520.779.266,0750.810.750.465,529
39Kochi2.200.515.864,3640.540.490.353,887
40Fukuoka0.660.460.7510,9990.550.420.509,898
41Saga1.280.742.514,4980.820.700.703,992
42Nagasaki1.240.514.486,0590.520.490.305,594
43Kumamoto1.830.564.715,5760.680.500.575,016
44Oita2.010.656.025,6870.890.630.915,271
45Miyazaki0.740.570.875,6590.640.550.415,278
46Kagoshima2.130.656.295,5400.810.620.765,063
47Okinawa1.070.622.109,9000.710.590.628,801
Total1.090.533.183,58,8700.700.510.813,32,738

The two pieces of evidence presented above suggest that the measurement error in land values is larger for households residing in rural areas, whose heads tend to be older than those residing in urban areas, whose heads tend to be younger. However, note that it is also possible that the cross-sectional variation in land prices is smaller in rural areas than in urban areas. To examine this possibility, we calculate the coefficient of correlation between the prefectural means of land prices per square meter and their variation within each prefecture measured by the coefficient of variation over the period 1983–2012 and find that it is 0.52. Thus, we find that the cross-sectional variation of land prices per square meter is indeed larger in urban than in rural areas.

In sum, there are two competing factors affecting the relative sizes of the measurement error in land values for younger and older households: one making the measurement error larger for older households, and the other making it larger for younger households. Unfortunately, it is extremely difficult to test which factor dominates, since the measurement error in land values is essentially unobservable.

Therefore, since we cannot determine the relative sizes of the measurement error, we use an alternative strategy. Specifically, we restrict the sample to households for which the size of measurement error in land values is likely to be similar and conduct the same estimations as in Table 2, Table 3, and Table 4. If we still find that the MPC for older households is larger than that for younger households in this restricted sample, we conclude that our main finding is not driven by differences in the size of the measurement error in housing wealth between younger and older households. More specifically, we first drop households living in less-populated areas (towns or villages), which reduces the number of households by 26,132. We then further drop households residing in densely populated prefectures (Saitama, Tokyo, Chiba, Kyoto, Osaka, Hyogo, and Hiroshima), which reduces the number of households by another 80,022. The remaining sample contains households that live in areas with similar population densities. For convenience, we call these areas with medium population densities.

Using the sample containing only households residing in areas with medium population densities, we find that the difference in MPCs between younger and older households is still statistically significant in 6 out of the 10 specifications.[35] We therefore conclude that our main finding that the MPC for older households is larger than that for younger households is not driven by differences in the size of measurement error in land values between younger and older households.

House values

As explained in Section 2.2, to calculate the house (building) value, we match the FIES data with data on average construction costs (by type of building structure, municipality, and year) reported in the ARBC (1953–2012). We essentially calculate house values by multiplying the construction cost per square meter by the total floor space homeowners report. For example, suppose there is a household that lives in Shibuya, Tokyo, and owns a wooden house constructed in 1990 with a total floor area of 100 m2. According to the ARBC, the (nominal) construction cost of wooden houses per square meter in Shibuya in 1990 was about 181,000 yen (about 1,810 U.S. dollars). Thus, in this case, the nominal value of the house when it was constructed is calculated as 181,000 yen×100 m2 = 18,100,000 yen (about 181,000 U.S. dollars).

In addition, in contrast to land, the value of houses depreciates over time due to physical wear and tear. To incorporate physical depreciation into the calculation of house values, we estimate depreciation rates by type of building structure based on the statutory expected lifetime of buildings provided by the “Ministerial Ordinance Concerning the Durable Years of Depreciable Assets.” For instance, as shown in Table 11, the expected lifetime of wooden structures is 22 years. In this case, we estimate the depreciation rate for wooden structures so that the remaining value of houses is 10% of the original value 22 years after construction. In other words, we solve the following nonlinear equation for δ, 1δ22=0.1, and obtain δ0.099. In the example above, suppose the household is surveyed in 2000, so that the wooden house it owns is 10 years old. In this case, the nominal value of the house when the household was surveyed is calculated to be about 6,380,000 yen (about 63,800 U.S. dollars) 18,100,000 yen× (10.099)10. Finally, for the estimation, calculated nominal house values are converted into real values using prefectural CPIs.

Table 11:

Estimated depreciation rates by type of building structure.

Type of buildingWoodenBlock/OtherSteel-frame and/or reinforced concrete
Expected lifetime22 years38 years47 years
Estimated depreciation rate9.9%5.9%4.8%

  1. Notes: The data for the expected lifetime by type of building structure are taken from the “Ministerial Ordinance Concerning the Durable Years of Depreciable Assets.” The depreciation rates are estimated assuming that house values are 10% of the original value after the expected lifetime ends.

Note that the estimation process for house values is also not immune to measurement error, for the following two reasons. First, since we employ municipality average data for construction costs by type of structure, heterogeneity within a municipality is not controlled for. It is possible that some houses in City A are built using high-quality woods that are expensive and depreciate slowly over time, whereas other houses also in City A are built using low-quality woods that are cheap and depreciate quickly over time. Averaging the construction costs over a municipality eliminates this type of heterogeneity, leading to measurement error in house values. Second, information on whether a house has been renovated since it was built is not available in the FIES. Failure to take into account whether houses have been renovated undervalues the value of houses that have been renovated. We leave it for future research to control for this type of heterogeneity within a municipality when estimating house values.

References

Atalay, K., S. Whelan, and J. Yakes. 2014. “House Prices, Wealth and Consumption: New Evidence from Australia and Canada.” Review of Income and Wealth 62 (1): 69–91.Search in Google Scholar

Attanasio, O. P., L. Blow, R. Hamilton, and A. Leicester. 2009. “Booms and Busts: Consumption, House Prices and Expectations.” Economica 76 (301): 20–50.10.1111/j.1468-0335.2008.00708.xSearch in Google Scholar

Attanasio, O. P., A. Kovacs, and K. Molnar, “Euler Equations, Subjective Expectations and Income Shocks,” University of Oxford, Department of Economics, Discussion Paper Series, No. 820, 2017.Search in Google Scholar

Attanasio, O. P., and G. Weber. 1994. “The UK Consumption Boom of the Late 1980s: Aggregate Implications of Microeconomic Evidence.” The Economic Journal 104 (427): 1269–302.10.2307/2235449Search in Google Scholar

Bhatia, K., and C. Mitchell. 2016. “Household-Specific Housing Capital Gains and Consumption: Evidence from Canadian Microdata.” Regional Science and Urban Economics 56: 19–33.10.1016/j.regsciurbeco.2015.10.004Search in Google Scholar

Browning, M., A. Deaton, and I. Margaret. 1985. “A Profitable Approach to Labour Supply and Commodity Demands over the Life-Cycle.” Econometrica 53 (3): 503–44.10.2307/1911653Search in Google Scholar

Browning, M., M. Gortz, and S. Leth-Petersen. 2013. “Housing Wealth and Consumption: A Micro Panel Study.” The Economic Journal 123 (568): 401–28.10.1111/ecoj.12017Search in Google Scholar

Campbell, J. Y., and J. F. Cocco. 2007. “How Do House Prices Affect Consumption? Evidence from Micro Data.” Journal of Monetary Economics 54: 591–621.10.1016/j.jmoneco.2005.10.016Search in Google Scholar

Case, E. K., M. J. Quigley, and J. R. Shiller. 2005. “Comparing Wealth Effects: The Stock Market versus the Housing Market.” Advances in Macroeconomics 5 (1): 1–32.Search in Google Scholar

Case, E. K., M. J. Quigley, and J. R. Shiller, “Wealth Effects Revisited 1978–2009,” NBER Working Paper Series, No. 16848, 2011.Search in Google Scholar

Christelis, D., D. Georgarakos, and T. Jappelli. 2015. “Wealth Shocks, Unemployment Shocks, and Consumption in the Wake of the Great Recession.” Journal of Monetary Economics 72: 21–41.10.1016/j.jmoneco.2015.01.003Search in Google Scholar

Cooper, D. 2013. “House Price Fluctuations: The Role of Housing Wealth as Borrowing Collateral.” The Review of Economics and Statistics 95 (4): 1183–97.10.1162/REST_a_00323Search in Google Scholar

Cristini, A., and A. Sevilla. 2014. “Do House Prices Affect Consumption? A Re-Assessment of the Wealth Hypothesis.” Economica 81: 601–25.10.1111/ecca.12098Search in Google Scholar

Deaton, A. 1985. “Panel Data from Time Series of Cross-Sections.” Journal of Econometrics 30: 109–26.10.1016/0304-4076(85)90134-4Search in Google Scholar

Dekle, R. 1994. “Market Value Estimates of Japanese Saving and Comparisons with the U.S.: Can the Capital Gains to Land Be Included in Saving?” Japan and the World Economy 6 (1): 27–44.10.1016/0922-1425(94)90035-3Search in Google Scholar

Disney, R., J. Gathergood, and A. Henley. 2010. “House Price Shocks, Negative Equity, and Household Consumption in the United Kingdom.” Journal of the European Economic Association 8 (6): 1179–207.Search in Google Scholar

Dvornak, N., and M. Kohler. 2007. “Housing Wealth, Stock Market Wealth and Consumption: A Panel Analysis for Australia.” The Economic Record 83 (261): 117–30.10.1111/j.1475-4932.2007.00388.xSearch in Google Scholar

Gan, J. 2010. “Housing Wealth and Consumption Growth: Evidence from a Large Panel of Households.” The Review of Financial Studies 23 (6): 2229–67.10.1093/rfs/hhp127Search in Google Scholar

Hayashi, F. 1985. “The Permanent Income Hypothesis and Consumption Durability: Analysis Based on Japanese Panel Data.” The Quarterly Economic Review 100 (4): 1083–113.Search in Google Scholar

Hori, M., and S. Shimizutani. 2004. “Asset Holding and Consumption: Evidence from Japanese Panel Data in the 1990s.” Seoul Journal of Economics 17 (21): 153–79.Search in Google Scholar

Iwamoto, K., T. Niizeki, J. Hamaaki, M. Hori, S. Maeda, and K. Murata. 2015. “Estimation of Real Assets Holdings by Japanese Households Based on the Microdata from the Family Income and Expenditure Survey (In Japanese).” Economic Analysis (Keizai Bunseki) 189: 63–95.Search in Google Scholar

Jappelli, T. 1990. “Who Is Credit Constrained in the U.S. Economy?” The Quarterly Journal of Economics 105 (1): 219–34.10.2307/2937826Search in Google Scholar

Ludwig, A., and T. Slok. 2004. “The Relationship between Stock Prices, House Prices and Consumption in OECD Countries.” Topics in Macroeconomics 4 (1): 1–26.Search in Google Scholar

Mian, A., K. Rao, and A. Sufi. 2013. “Household Balance Sheets, Consumption, and the Economic Slump.” The Quarterly Journal of Economics 128 (4): 1687–726.10.1093/qje/qjt020Search in Google Scholar

Naoi, M. 2014a. “Housing Wealth, Reference-Dependent Preferences and Household Consumption (In Japanese).” Keio Journal of Economics 106 (4): 473–87.Search in Google Scholar

Naoi, M. 2014b. “Housing Wealth and Household Consumption Behavior (In Japanese),” in Economic Analysis of Cities, Regions, and Real Estate Markets, edited by K. Sumita, M. Naoi, and Y. Okumura,, 3– 26. Tokyo: Keio University PressSearch in Google Scholar

Ogawa, K., S. Kitasaka, H. Yamaoka, and Y. Iwata. 1996. “An Empirical Re-Evaluation of Wealth Effect in Japanese Household Behavior.” Japan and the World Economy 8 (4): 423–42.10.1016/0922-1425(96)00218-6Search in Google Scholar

Paiella, M. 2007. “Does Wealth Affect Consumption? Evidence from Italy.” Journal of Macroeconomics 29: 189–205.10.1016/j.jmacro.2005.06.001Search in Google Scholar

Paiella, M., and L. Pistaferri. 2017. “Decomposing the Wealth Effect on Consumption.” The Review of Economics and Statistics 99 (4): 710–21.10.1162/REST_a_00629Search in Google Scholar

Phang, S. 2004. “House Prices and Aggregate Consumption: Do They Move Together? Evidence from Singapore.” Journal of Housing Economics 13: 101–19.10.1016/j.jhe.2004.04.003Search in Google Scholar

Pistaferri, L. 2016. “Why Has Consumption Remained Moderate after the Great Recession?” paper prepared for Boston Fed Conference.Search in Google Scholar

Skinner, J. 1989. “Housing Wealth and Aggregate Saving.” Regional Science and Urban Economics 19: 305–24.10.1016/0166-0462(89)90008-2Search in Google Scholar

Slacalek, J. 2009. “What Drives Personal Consumption? the Role of Housing and Financial Wealth.” The B.E. Journal of Macroeconomics (Topics) 9 (1): 1–35.Search in Google Scholar

Takayama, N., F. Funaoka, F. Ohtake, M. Sekiguchi, and T. Shibuya. 1989. “Household Wealth and Saving Rates in Japan (In Japanese).” Economic Analysis (Keizai Bunseki) 116: 1–69.Search in Google Scholar

Windsor, C., J. P. Jaaskela, and R. Finlay. 2015. “Housing Wealth Effects: Evidence from an Australian Panel.” Economica 82 (327): 552–77.10.1111/ecca.12137Search in Google Scholar

Published Online: 2019-03-05

© 2019 Walter de Gruyter GmbH, Berlin/Boston