Value of land and dwellings: Treatment in French National Accounts - - PDF document

1

Value of land and dwellings: Treatment in French National Accounts OECD Working Party on National Accounts

Table of contents

Preliminary notes ..................................................................................................................................... 1 Abstract .................................................................................................................................................... 1 I – Layout ................................................................................................................................................. 2 II – Introduction ........................................................................................................................................ 3 III – Data Sources .................................................................................................................................... 3 A - Data on GFCF by institutional sector and comparison with Housing Satellite Accounts for the whole economy .................................................................................................................................... 3 B - Data on capital stock ...................................................................................................................... 4 C - Data on prices ................................................................................................................................ 5 IV – Modified PIM .................................................................................................................................... 6 A - Initialization: 1988 ........................................................................................................................... 6 B - Other years, for dwellings............................................................................................................... 7 C - Choice for r (n) ............................................................................................................................... 7 D - Iteration for dwellings ..................................................................................................................... 8 E - Iteration for land after 1988 ............................................................................................................ 9 F - Results ............................................................................................................................................ 9 V – Financing of dwellings ..................................................................................................................... 11 A - Data sources ................................................................................................................................ 11 B - Financing dwellings for households ............................................................................................. 12 VI – Conclusion: Suggested improvements .......................................................................................... 13 VII – Bibliography .................................................................................................................................. 13

Preliminary notes

• The following note is a copy of the speech prepared for the OECD Working Party on National

Accounts (October 1-4, 2013). The presentation slides are available here.

• Throughout the presentation, the words dwellings or buildings will indifferently be used to talk

about residential buildings. The word housing will be improperly used to talk about both dwellings and land. The presentation only discusses the residential part of the real-estate wealth. There will be no reference about the non residential buildings, except if specifically mentionned.

Abstract

The national non financial wealth is mainly constituted by housing assets which have an important weight: about 60 % for the whole economy and about 90 % for the households in 2011. In the French

2 balance sheet, the treatment of housing wealth distinguishes the valuation of buildings from the valuation of lands underneath buildings, as suggested by the European System of Accounts (ESA). The net value of residential buildings is calculated using a perpetual inventory model. Like the other fixed assets, this method allows to evaluate the stocks by accumulating flows of investment in dwellings (GFCF of new dwellings and substantial repair and maintenance) with an estimated rate of depreciation. Data on housing stocks (dwellings + land) are also available in France. The perpetual inventory model is improved in order to integrate this additional information: the total value of the housing stock is provided every five years by a specific survey known as “National Housing Survey”. The valuation of housing wealth is based on two different price indexes: the buildings price follows the construction cost index and the total of housing stocks (buildings + land) is valued according to the housing price index. Eventually, with this method, the land price supports the main part of the housing prices’ evolution, especially recently. This reflects the fact that localization is the most important factor which determines the price of housings. The presentation will focus on the description of the methodology used in French national accounts to evaluate non financial housing wealth. A few points on expected improvements for the base-year 2010 will also be developed.

I – Layout

[Slides 1 & 2] The purpose of this presentation is to explain the method used in French national accounts to measure the value of possessed dwellings and of land attached to these dwellings. It is meant to be a technical presentation which sets forth the data sources and the methods associated with the building

• f this account. The latter have not significantly changed since the base-year 1995. At the end of the

presentation, several elements about the changes planned for the base-year 2010, to be published in late 2014, will also be given. The layout of this presentation is as follows: First, a brief introduction to recall the history of the building of the French balance sheet. Second, the details of the French data sources on housing: for Gross Fixed Capital Formation (GFCF), price indexes and stocks of dwellings and land series. Third, what is commonly refered to as the perpetual inventory method. And more specifically when used for that particular kind of asset which is housing, meaning dwellings plus land. This method enables us to build the series of dwelling capital stocks. And it also enables us to find the price index which is then used to value the land on which dwellings are built, the land underneath dwellings. Fourth, some results on the financing of housing for French households. Some of the time series presented in this section are included in the French national accounts but are not produced by Insee (French national institute of statistics and economic studies). The French central bank and the ministerial statistical service for sustainable development (which provides the satellite account for housing) publish more detailed information on the different types of financing used by households. Throughout the presentation, the consistency between the different data sources on housing will be

• studied. To conclude, a short presentation of the improvements that are implemented for the transition

to the base-year 2010 of national accounts.

3

II – Introduction

[Slide 3] In the French national accounts, the first estimates of the housing balance sheet date back to the

• 1970s. The French national accounts already developed the main method still used today in most

countries: the perpetual inventory method (PIM). The Balance sheets are detailed since the European System of Accounts (ESA) of 1993. In France, the housing balance sheet is published annually, for the non-financial portion, starting with the base- year 1995. As noted in the ESA, the estimates of dwellings – which are production inputs – and of land underneath buildings - which are non-produced assets- need to be separated. The standard perpetual inventory method estimates the stock of fixed assets as a process of accumulation of past investments. This method is used in most countries that produce housing balance sheets on a regular basis. This method is convenient because it is mainly based on flows, which are easier to assess than stocks. To implement the method, we just have to make assumptions about the average service life of the assets and their survival prospective laws. Since the beginning of the French national accounts, building housing balance sheet required a specific method. For housing as opposed to other assets, we have real data sources on stocks which can therefore serve as a benchmark to check the consistency of the permanent inventory model. Consequenltly, the perpetual inventory method presented here is a specific one.

III – Data Sources

A - Data on GFCF by institutional sector and comparison with Housing Satellite Accounts for the whole economy

[Slide 4] Let’s first begin by presenting the different data sources on dwellings and land. The first step is to gather series of GFCF of residential buildings in current prices. GFCF of dwellings have two different components: the biggest part is the acquisition of new dwellings and substantial repair and maintenance, the smallest part is the costs associated with the transfer of ownership. In the French national accounts, notary fees represent the largest share of these costs, but we also include the costs

• f real estate promoters, real estate agents, architects and engineers, when there are some. These

GFCF series are available in the annual national flows accounts. The housing satellite account also produces data on GFCF of dwellings with both buildings and associated charges. Those time series are not exactly the same between both sources as you can see

• n the graph. The national accounts series are in blue and the housing satellite account in pink.

For new dwellings, timeseries are very close in levels and have almost the same annual growth rates. In fact, these first two series are built on the same fieldsource. The national accounts actually use the satellite account. The latter is based on home started and delayed work. But the national accounts data also use the annual business surveys which provide housing companies turnover depending on the status of the customer (individuals, governments, corporations). Finally, the national accounts also have to review the overall consistency of input-output balances. This is what leads to the small differences in levels between the two series for the past few years. Regarding the series of GFCF for costs associated with the transfer of ownership, the gap between the two series is larger and comes from a difference in the fieldsource considered. Indeed, the satellite account housing serie (which is the pink dotted line) consists mainly of Notary fees (transfer taxes, registration fees), while the national accounts serie covers a wider field including the fees of estate agents for example.

4 All GFCF data for housing are available by customer type and are turned into institutional sectors. To do this, we use either the breakdown of the national housing survey or that of the insurance account.

GFCF of dwellings

value in billions euros

20 40 60 80 100 120 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

dwellings in national accounts dwellings in housing account costs of ownership transfer in national accounts costs of ownership transfert in housing account Graph 1: GFCF of dwellings. Sources: Insee, housing satellite account.

B - Data on capital stock

[Slide 5] The special feature of the method used here to build the housing wealth comes from the fact that we have other sources, especially on housing stocks. The perpetual inventory model is improved in order to integrate this additional information. First, the total value of the housing stock (dwellings + land) is known approximatively every five years, provided by a specific survey carried out by Insee and known as “National Housing Survey (NHS)”. This survey gives detailed information about the national park of dwellings (it is based among nearly 43,000 units in France), including a direct estimate of the stock in current prices since 1973. However, this datasource does not split buildings from land underneath. This specific datasource is used to initialize the PIM and also to allow a benchmark for the results, especially for the recent period. Regarding the datasources avalaible for land, which is a non produced asset, there is not much information in dataflows. We can gather the flows associated to the exchanges of land between institutional sectors (when government bought a land to households for example) or we also have some information on land transactions in case of home started (i.e. when it cames simultaneous with the start of the construction of a building). But, these data flows represent only a small part of the land acquisition. However, another data source can be used. Indeed, we have a stronger idea of the areas of land underneath residential buildings. This data is provided by another survey, called Ter- Uti, which counts annually the surfaces of each type of area throughout the French territory: agricultural land, land underneath buildings or civil engineering works, forests, etc. This is what leads to build an index of volume for land stocks.

5 These surveys provide data for the whole economy. So as to turn the information into institutional sectors: government, households, financial and non- financial companies, we use the same breakdown than for the GFCF, for the whole housing stock as well as for land areas.

C - Data on prices

[Slides 6 & 7] To assess GFCF series, housing stocks and land areas, there are several price data sources. 1988 is the starting point of the PIM. Before 1988, we have a largest number of empirical information. Indeed, 3 price indexes are available, including a price index for land underneath buildings. After 1988, 2 price indexes are still used in the model. The index on construction costs is used to value GFCF and housing stocks. It is a quarterly index, produced by Insee since 1953. It measures the changes in prices of home started, through the

• bservation of the real-estate market. This index excludes the price of substantial repair and

maintenance and the costs of ownership transfers, yet in the GFCF fieldsource. To value the whole housing stock (dwellings plus land), we use the housing price index or notary index, also provided on a quarterly basis by Insee since 1996. This index reflects another aspect of the real-estate market as it tracks the price changes in sales transactions of existing dwellings (aged over 5 years old). Prior to 1996, a composite index of housing prices is computed in order to backcast the notary index. Due to the lack of source on land price after 1988, the model is build to allow the assessment of the land price, as a result from the estimates of housing stocks on the one hand and dwelling ones on the

• ther hand.

Price indexes

2000 = 100

50 100 150 200 250 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

House price index construction cost index

Graph 2: Price indexes for total housing stocks and for dwelling stocks. Source: Insee. On the graph above, are represented the two major price indexes that are used in the model since

• 1959. In black is the index on housing prices used to value the housing stock (dwellings plus land) and

in blue is the index of construction costs used to value the stock of dwellings only.

6 The sharp rise in property prices since the early 2000s is very noticeable from the housing price. The latter price is mainly based on the existing dwellings transactions. Indeed, this index rose from 100 in 2000 to 200 in 2007, which represents an increase of 100% in 8 years. In contrast, the price for the construction of new dwellings (represented by the cost of construction index in blue) increases at a much slower pace, rising by 30% only. Between 2007 and 2009, the housing price (which is the black dotted line) decreased by 8% and the index of construction costs stagnated. Thus, during the 2008 crisis, the lack of correlation between the two price indices was strong. After the crisis, the housing price increases again, by 9% between 2009 and 2012. The changes in the cost of construction are more similar during this period of uncertain recovery, the index rising by 8%. By definition, the price of land that will be computed will therefore even more sharply reflect rising real- estate prices as you will see later in the presentation.

IV – Modified PIM

A - Initialization: 1988

[Slide 8] Diagram 1: PIM’s Initialization. This diagram shows the modified perpetual inventory method used to build the stocks in dwellings and to determine the price of land under the buildings. 1988 is the year of the initialization of the model. Indeed, for this year, we have both survey data regarding the total housing stock value (via the national housing survey) and all price indexes, including the price index for land underneath buildings. We also have the surface of land that is developed via the Ter-Uti (LUCAS) survey.

7 So, on the one hand, we can calculate the value of the land that was developed in 1988 by multiplying the surface of land by the price index for land. And on the other hand, we have the amount of total housing value (dwellings plus land). One only has to subtract the cost of land from the total housing value to determine the dwellings stock

• value. To obtain the volume of the dwellings stock in 1988, the value is deflated using the cost of

construction index. The dwellings stock at constant prices in 1988 will serve as a fixed starting point to initialize the accumulation of GFCF in the model.

B - Other years, for dwellings

[Slide 9] At constant price: K (n) = K (n-1) + GFCF (n) – CFC (n) With r (n) = CFC (n) / K (n-1) (4): K (n) = K (n-1) + GFCF (n) – r (n) x K (n-1)

• Before 1988

K (n-1) = [K (n) - GFCF (n)] / [1 – r (n) ] (1)

• After 1988

K (n) = K (n-1) x (1 – r(n) ) + GFCF (n) (1) Legend: K: net capital stock of dwellings GFCF: gross fixed capital formation of dwellings CFC: consumption of fixed capital of dwellings r: rate of CFC To calculate the stock of dwellings before and after 1988, we rely on the classical equation of the perpetual inventory model, which expresses the capital of a year as a function of the capital of the preceding year added to the current year GFCF, reduced to reflect any used capital for the year (consumption of fixed capital). In this equation, we know the GFCF, we compute the stock by iteration since we know the first value for 1988, but the CFC (consumption of fixed capital) is not known. Therefore, we can express r (n), the damping coefficient of the CFC, as the ratio of the year n CFC divided by the capital for the year n-1 (see equation number 4) We can then rewrite the accounting equation before and after 1988, as a function of r. See equation number 1 on the slide.

C - Choice for r (n)

[Slide 10]

8 r (n) = a + b x T (3) where a = constant, b >0, T: turnover rate T = 5-year average of GFCF / 5-year average of capital stock (2) ⇒ r (n) = 0,01 + 0,02 x T (0,01 ; 0,02) : values matching with capital stock from NHS initial value of r = 0,015 and iteration To estimate the damping coefficient r, we assume a linear relationship between the damping coefficient and the dwelling capital turnover rate, noted T. This is reflected in equation number 3. The dwelling capital turnover rate is the ratio between the 5-year average of the GFCF and the 5-year average of the capital stock in dwellings, as reflected in equation number 2. The goal therefore is to determine the coefficients a and b in equation number 2. An empirical estimation was carried out to determine these two coefficients since the idea is to use all directly available information via surveys on stocks. Thus, at the end of the evaluation process, the addition of the value of stocks of dwellings and land must be consistent with the value of the total housing stock provided by the national housing survey. For France throughout the period, the study results of the coefficients a and b being equal respectively to 0.01 and 0.02. Thus, there is a variable capital depreciation rate over time while it is constant in time for all other fixed assets other than housing in the standard perpetual inventory method since there is no information on particular stocks of other assets. In practice, you still need a starting point to begin the iteration of the model. As with other assets, the amortization rate is set according to the average service life of non-residential buildings. In France, the average service life is estimated at 60 years, which means a depreciation rate of 1/60 or 0.015. This is the same average service life that is being used for non-residential buildings in the standard case.

D - Iteration for dwellings

[Slide 11]

Initial value of r = 0,015 GFCF of dwellings at constant price Net capital stock of dwellings at constant price (1) GFCF of dwellings at constant price Turnover rate T (2) Recalculated r (3) Net capital stock of dwellings at constant price (1) CFC at constant price (4) Iteration until the difference between the initial value of r and the recalculated r is negligible

Diagram 2: PIM’s Iteration. The diagram above details the iteration process. To start, we have both the initial value of the damping coefficient and GFCF of dwellings at constant price. Using the rewrited classical equation (the first one presented in section B) and starting from 1988, we can estimate one first serie of net capital stocks of dwellings at constant price.

9 Thus, we can use the equation noted 2 in order to estimate a serie of dwelling capital turnover rate (T) for the whole period. The net capital stocks of dwellings are re-assessed until the difference between the whole-period- average of the recalculated r (using equation 3) and the initial starting point is negligible. Rewriting equation 4, the serie of CFC at constant price can also be computed throughout the iteration.

E - Iteration for land after 1988

[Slide 12] Dwellings Land Total Volume

Estimated by PIM (1) (2)= value in year n * volume index known (from surface survey n+1/n) (3 =1 + 2)

Value

Estimated by PIM (4) (6 = 5 - 4) (5 = 3*House price index)

Price

cost of construction idex (7 = 6 / 2) House price index

year n+1 since 1989 Diagram 3: Iteration for Land. We already have obtained all the information to measure the stocks of dwellings at constant price with the iteration of the PIM. The data used for prices is the cost of construction index. This is shown in the first column of the table above. In order to have a stock of housing, we have to compute the net value of land underneath dwellings: the building of this serie is indicated in the middle column. To do that, calculations use the information available on the total of housing. First, at constant price, we can estimate year by year the volume of land by multiplying the index of volume of areas by the value in the year before, again using 1988 as a starting point because we have all the datas at that point. Then, we can sum the volumes of dwellings and land in order to derive a total of housing stocks in

• volume. This is the cell in the top right of the table. As we have the housing price index, we can

compute the total of housing stocks in value. Third, we calculate the latter minus the serie of dwellings stocks in value which leads us to an estimation of the net value of land, as we can see in the central cell of the table. Finally, we also can deduct the last unknown serie of the model which is the price of land, by dividing the value of land by their volume. Again, we must iterate these calculations every year to obtain the entire period.

F - Results

[Slides 13 & 14]

10

Price indexes for dwellings and land

2000 = 100

50 100 150 200 250 300 350 400 450 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

House price index construction cost index Land price

Graph 3: Price indexes for total housing stocks, for dwelling stocks and for land. Source: Insee. On the graph above, are represented the results for the model in terms of price indexes since 1959. In black is still drawn the index on housing prices used to value the housing stock (dwellings plus land) and in blue the index of construction costs used to value the stock of dwellings only. The new price index on the graph, the red dotted line, is the land price, which is derived from the iteration of the PIM. We can observe that the geometrical average of red and blue curves probably gives the black one, which is consistent with the fact that the housing price should represent the fixation of the two other prices. We also can see that the land price supports the main part of the housing prices’ evolution, especially

• recently. This reflects the fact that localization is the most important factor which determines the price
• f housings. Indeed, dwelling price is only determined by the costs at the construction time.

The rise in land price since the early 2000s is sharper than those of the housing price. The former index rose from 100 in 2000 to 300 in 2007, which represents an increase of 200% in 10 years. The latter increases by 100% in 10 years, and the cost of construction index only by 30% in the last decade. Between 2007 and 2009, the land price (which is the red dotted line) decreased very deeply, by 15,5%. During the same period, the index of construction costs stagnated and the housing price decreased by 8%. After 2009, the three indexes has jumped by about 10%.

11

Housing net value

in billions euros

1000 2000 3000 4000 5000 6000 7000 8000 9000 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Dwellings Land Total from MIP Total NHS Graph 4: Housing net value for total housing stocks, for dwelling stocks and for land. Source: Insee. Regarding the results in terms of net value in billions euros since 1959, we can see that, since the mid-2000s, the net value in total housing is now balanced between the dwelling part and the land part. In black, the housing stock (dwellings plus land) rose sharply since 2000, mainly because of the rise of the land value (which is the red dotted line). The dwelling stock, in blue, increases at a slower pace. The green squares, represented on the graph, symbolize the data from the national housing survey, produced approximatively every 5 years. We can observe that the black curve follows quite well this empirical data, that lead us to validate the estimations of the PIM.

V – Financing of dwellings

A - Data sources

[Slide 15] Briefly to finish the presentation, several features of the modes of financing of housing for households, which hold 70% of the whole net value in housing. In the financial data sources, collected by the French central bank, the flows of new housing loans are

• available. The housing satellite account also produces series on financing housing which are more

detailed and, most importantly, these are given excluding the repayement of loans, while the financial data sources represents net assets in order to be integrated in the financial accounts. In this latter source, the GFCF of dwellings for households is separated in 3 types of financing: the loans, the allowances and the own funds. The consistency of the series of loans between the two sources has been checked. Despite the difference in original datasource, if we rectify the financial data, both series are quite similar. Thus, the

12 following part will focus on the data from the housing sattelite account, even if the flows used in the national financial accounts are those from the central bank.

B - Financing dwellings for households

[Slide 16]

Financing dwellings for households

value in billions euros

20 40 60 80 100 120 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

GFCF in housing accounts

• wn funds

loans housing allowances

Graph 5: Financing dwellings for households. Source: Housing satellite account. In this graph, the black curve represents the GFCF in housing for french households and its breaking down between the own funds, the loans and the housing allowances. We may note that the changes

• f GFCF and of loans are quite uncorrelated. To this, one could argue several reasons.

First, conceptually, one can consider at least two explanations. As mentioned before, the loans represented here are net from repayments : thus, the loans are partly determined by the GFCF of the previous periods. Moreover, when the loans are due to renegotiations of old loans, they are also not related to the GFCF of the current year. Second, there are significant differences in the scope of both series of GFCF and loans. The loans data does not distinguish the financing of new dwellings and the financing of existing dwellings, while in the GFCF, only the new dwellings are counted. In addition, the loans data includes the financing of lands underneath buildings, which is not the case in the GFCF serie. And, finally, the GFCF serie is recorded at the time of the transfer of ownership while the loans are collected at the time of actual payment correponding to the completion of building works. Therefore, on the graph above, we can note that at the beginning of the financial crisis in 2007, the increase in GFCF continues as the loans start to decrease significantly. On the contrary, after 2009, there was a significant recovery of loans while GFCF in housing fell down. These opposite changes act on an opposite way on the financing capacity of households. Towards the financial accounts, the rise in loans after the crisis leads to a decrease in funding capacity for households while the decline of GFCF in housing results in an increase in funding capacity for households in the capital account. Finally, the changes in GFCF are mainly explained after the crisis by the own funds profile.

13

VI – Conclusion: Suggested improvements

[Slide 17] To conclude, one can present at least four improvements that are implemented for the transition to the base-year 2010, to be published in late 2014. First, the backdata series on GFCF of dwellings have been entirely rebuilt in order to make them more consistent between the different sources. Second, data of Ter- Uti survey which provides the main source on the areas of land will change with the building of the next account. Third, concerning the prices indexes of both the dwellings and land, several changes may be

• implemented. For the price of land, a new data source shall be exploited. For the price of dwellings,

we plan also use new price indexes in order to assess the dwellings stocks, in addition to the cost of construction index already used by the model. Fourth, in accordance with the demands of the new European system of accounts, the average service life of the costs associated to ownership transfers will be determined separatly from that of the residential buildings.

VII – Bibliography

• Economie et Statistique n°25, J. Mairesse, 1971, “ L’estimation du capital fixe productif ”
• Economie et Statistique n°114, Septembre 1979, “Les comptes de patrimoine”
• National Accounts: methodological note (base-year 2000)

http://www.insee.fr/fr/themes/comptes- nationaux/default.asp?page=base_2000/documentation/methodologie/resume_nb10.htm

• Housing Satellite Account: methodological note (mars 2011)

http://temis.documentation.developpement-durable.gouv.fr/documents/Temis/0023/Temis- 0023454/10386_2009_note_methodologique.pdf