Housing equity withdrawal and consumption. Dynamics in Estonia 2002-2011

Full text

RE B 20 12
Vol. 4, No. 1
19
Abstract
e linkages between housing wealth and household consumption are contentious. is
paper uses a recently introduced statistical concept, housing equity withdrawal (HEW), to
investigate the linkages in the case of Estonia. HEW is dened as net borrowing by the
household sector, which is secured on housing equity but not invested in housing assets.
HEW is thus a direct measure of the cash ow from housing assets available to the household
sector. e HEW series computed for Estonia are much more volatile than similar series in
countries such as the UK and the USA. e volatility is related to rapidly changing nancing
conditions and real estate prices, but also to consumption aspirations. Econometric analysis
conrms a close correlation between housing equity withdrawal and consumption, but the
relation appears to dier between the boom in 2002-2006 and the downturn period 2007-
2011.
JEL classification codes: D12, E21, R21, R31
Keywords: Housing equity withdrawal, household consumption, household saving, house prices,
mortgage market
Housing Equity Withdrawal and Consumption
Dynamics in Estonia 2002-2011*
* The authors would li ke to thank two a nonymous referees for u seful comments and Tõnu Mer tsina from Stat istics
Estonia , Andres Juss from the Estonian Land Board a nd Kaspar Oja from Eesti Pank for providing the
underlying data and usef ul discussions of their contents. Kuk k and Staehr ack nowledge support f rom Base
Financing grant no. B617A and Target Financing grant no. SF0140059s12. The views expressed are those of the
authors and not necessarily those of the Minist ry of Fina nce of the Republic of Estonia or Eesti Pa nk.
RESE ARCH IN ECON OMICS A ND BUSINE SS: CENTRAL AND E ASTERN EUROPE
Madis Aben
Ministry of Finance, Estonia
Address: Suur-Ameerika 1, 15006 Tallinn, Estonia
Phone: +372 6113506, e-mail: madis.aben@n.ee
Merike Kukk
Tallinn University of Technology
Address: Akadeemia tee 3-478, 12618 Tallinn, Estonia
Phone: +372 6204069, e-mail: merike.kukk@tseba.ttu.ee
Karsten Staehr
Tallinn University of Technology
Eesti Pank
Address: Akadeemia tee 3-486, 12618 Tallinn, Estonia
Phone: +372 6204062, e-mail: karsten.staehr@tseba.ttu.ee

REB 2012
Vol. 4, N o. 1
20
ABEN • KUKK • STAEH R
1. Introduction
e linkages between housing wealth and household consumption are contentious in both
the theoretical and the empirical literature. Some theories imply that an increase in housing
wealth due to higher real estate prices leads to increased consumption, while other theories
suggest that this is not the case. Empirical studies of housing wealth and consumption have
oen provided rather inconclusive results and typically do not reveal the exact mechanisms
or linkages between housing wealth and consumption. To attain a deeper understanding of
the linkages, the statistical concept of Housing Equity Withdrawal (HEW) was introduced
inthe UKinthe beginningofthe 1990s(Westaway, 1993; Holmes,1993). e amountof
housing equity withdrawal (HEW) is calculated as the household sector’s net borrowing that
is secured on housing but not invested in housing. HEW measures the liquid funds or cash
ow attained by the household sector from otherwise illiquid housing assets and is thus a
means to establish a direct link between, on the one hand, housing assets and nancial
intermediation and, on the other hand, the consumption and saving of the household sector.
is paper investigates linkages between housing wealth and household consumption in
Estonia based on HEW data. e analysis comprises two steps. e rst step entails the
computation of annual and quarterly HEW data for the period 2002–2011.1 Two dierent
versions of HEW are computed based on dierent data sources. e main components of
HEW are examined and some features specic to Estonia are investigated. e second step
entails the estimation of consumption functions in which real household consumption is
explained by real income and real HEW. e estimations are undertaken using the Engle-
Granger two-stage methodology as this method was found to provide the most reliable and
robust results. e analyses consider structural breaks around 2007 when the preceding
boom was replaced by a severe downturn.
e choice of the sample country is of signicance for several reasons. First, households
in Estonia own a lot of housing assets in t he form of both dwellings and land. is widespread
ownership dates back to the restitution and privatisation processes which took place
throughout the 1990s. As a result, home-ownership in Estonia is among the highest in
Europe. e private sector owns 96% of all dwellings in Estonia (ES 2011). Second, Estonia
has experienced rapid developments of nancial markets, particularly in association with
the admission of Estonia to the European Union in 2004 (Brixiova et al., 2010). ird,
Estonia has, like other countries from Central and Eastern Europe, experienced larger
business cycle uctuations than typically seen in West European countries (Becker et al.,
2010). e ndings for Estonia may arguably have lessons for other CEE countries with
similar developments.
e rest of the paper is organised as follows. Section 2 discusses the theoretical and
empirical background for the paper. Section 3 introduces the concept of HEW and describes
dierent activities of the household sector, which make up HEW. Section 4 presents two
HEW time series for Estonia for the period 2002–2011 and discusses factors that explain the
1 It is the rst time such data have been made available for a country from Central and Eastern Europe. Due to a
comprehensive register of all real estate contracts held in the Land Board of Estonia, it is possible to compute
HEW series for Estonia with a delay of only approximately one month.
“Housing wealth isn’t wealth”
Mervin King, Governor of the Bank
of England, 1997 (Buiter, 2010, p. 1)

RE B 20 12
Vol. 4, No. 1
21
ABEN • KUKK • STAEHR
dynamics. Section 5 considers the relation between HEW and household consumption
based on graphical and econometric analyses. Finally, Section 6 concludes the paper.
2. Housing Wealth, HEW and Household Consumption
e eect of changes i n real estate prices on household consumption is subject to considerable
controversy. e theoretical starting point is the models of intertemporal smoothing,
pioneered by Friedman (1957) and Modigliani (1966). According to these models of rational
choice, the individual household smooths consumption across all future periods in its
remaining lifetime. e implication is that the household will consume a fraction of its
discounted lifetime wealth. e upshot is that there are two main channels through which
changes in the value of housing assets can aect consumption, viz. a wealth eect and a
credit or collateral eect.
e wealth channel posits that households view their housing asset as any other form of
wealth. Households aim to smooth consumption, and an increase in house prices therefore
leads households to consume a fraction of the wealth increase. Buiter (2010), among others,
questions this argument by noticing that households that own a housing asset likely possess
the asset in order to consume the services of the housing asset in the future. An increase in
the price of the housing asset also implies an increase in the price of housing services in the
future, and households will therefore save the increased housing wealth in order to pay for
the higher housing services in the future. In this case, an increase in housing prices will have
no noticeable eect on consumption. is is a rationalisation of Mervin King’s argument
that “housing wealth isn’t wealth” (Buiter, 2010).
e credit or collateral channel is based on the assumption that many households are credit
constrained due to lack of collateral. If housing wealth increases, households can provide more
collateral, and this may allow otherwise credit-constrained households to borrow and increase
consumption in the short run. e credit or collateral channel suggests a close relationship
between housing wealth, nancial intermediation and consumption (Muellbauer, 2008).
Numerous empirical studies estimate the wealth eect from housing and stock market
developments on household consumption, using either aggregate or micro data (Paiella,
2009). Most studies nd a statistically signicant positive relationship between housing
wealth and household consumption, but the underlying reason for the result (pure wealth
eect or collateral eect) is typically not investigated. Muellbauer & Murphy (1990) argue
that nancial liberalisation in the early 1980s allowed UK households to use more valuable
housing assets as collateral to nance consumption. Aron et al. (2011) nd that increased
consumption in the UK and the USA is related to increased asset prices, but also the
liberalisation of nancial markets initiated in the 1980s, which made it possible for
households to liquidise their more valuable housing assets.
For Estonia the propensity to consume out of housing wealth has been found to be
modest based on estimations for the period 1997–2005: an increase of 100 EUR in housing
wealth increases consumption by 0.4 EUR in the short term and by 1 EUR in the long term
(Paabut and Kattai, 2007). e results capture primarily the eect before the rapid changes
in housing and mortgage markets that began around 2004, aer the accession of Estonia to
the EU. In a later study Sonje et al. (2012) estimate the eect of housi ng wealth on consumption
in four Central and Eastern European countries, including Estonia. ey nd a stronger

REB 2012
Vol. 4, N o. 1
22
ABEN • KUKK • STAEH R
relationship between housing wealth and consumption in Estonia: if housing wealth
increases by 100 EUR, consumption increases by 4 EUR in the long term.
e importance of the developments in the nancial sector point to the need for a direct
measure of the household sector’s cash ow from housing assets. Housing equity withdrawal
is such a measure. It was introduced in the UK in the early 1990s when mortgage borrowing
bythehouseholdsectorexceededresidential investment(Westaway, 1993;Holmes, 1993).
HEW facilitates analyses of linkages between, on the one hand, house prices and nancial
development and, on the other hand, consumption and saving by the households.
Evidence on the relationship between HEW and household consumption diers across
countries. Boone et al. (2001) investigate the relationship between HEW and consumption
in the USA, the UK and Canada during the period of nancial market liberalisation. eir
ndings provide some support to the hypothesis that increasing housing equity withdrawal,
following the relaxing of credit conditions, is linked to increased consumption in those
three countries in the sample. Catte et al. (2004) compare the marginal propensity to
consume of housing wealth in OECD countries for the period 1990–2002. ey nd that in
countries with developed credit markets HEW explains consumption changes better than
house prices as HEW is a direct measure of the liquidising of housing wealth. ey nd that
HEW drives consumption and estimate that 89% is consumed in the United Kingdom, 63%
in Canada and Australia and 20% in the USA.
Benito (2009) nds that while HEW tracked consumption quite closely in the UK until
the end of the 1990s, the linkage has subsequently become weaker. Klyuev and Mills (2006)
report that HEW explains some short-run uctuations in consumption in the USA, the UK,
Australia and Canada, but there is no long-run eect. ey nd for the USA that HEW had
a short-term negative impact on household saving, in the order of 20 cents to a dollar, and
argued that HEW could explain part of the decrease in the saving rate since the mid-1990s.
Smith (2010) also nds that that there are only short-term eects of HEW on consumption
in the case of New Zealand.2
Catte et al. (2004) conclude that HEW can explain a large part of consumption changes
for countries in which three conditions prevail. First, nancial markets provide easy access
to mortgage nancing and to nancial products that facilitate equity withdrawal. Second, a
high rate of owner-occupation implies a wider distribution of housing wealth. And third,
low housing transaction costs and housing wealth exemption from capital gains taxes
encourage owners to perceive housing assets as more liquid.
In spite of the potential use of HEW in macroeconomic analyses, data has been only
produced for a relatively small number of developed countries. For an extended period of
time, quarterly HEW data have been published by the Bank of England. Detailed HEW
measures have also been produced for the USA (Greenspan and Kennedy, 2008), Australia
(Bloxham et al., 2010) and New Zealand (Smith, 2010).
e overall conclusion in the empirical literature is that HEW has substantial explanatory
power vis-à-vis household consumption, but the reaction of consumption to higher HEW
varies substantially across countries. Remarkably no studies have investigated possible
asymmetric reactions across dierent phases of the business cycle.
2 A number of studies have used microeconomic survey data to determine the allocation of resources from HEW
to inter alia consumption. Important studies include Benito and Power (2004) and Smith and Searle (2008) for
the UK, Hurst and Sta ord (2004) and Cooper (2010) for the USA, Schwartz et al. (2008) for Australia and van
Els et al. (2005) and Ebner (2010) for the Netherlands.

RE B 20 12
Vol. 4, No. 1
23
ABEN • KUKK • STAEHR
3. Types of Housing Equity Withdrawal and Injection
e amount of housing equity withdrawal (HEW) is calculated as the household sector’s net
borrowing that is secured on housing but not invested in housing. Housing equity is
withdrawn when lending secured on housing increases more than spending on housing
assets, which generates a cash ow that can be spent on consumption or investment in
nancial assets. Conversely, equity is injected into housing stock when spending on housing
assets exceeds lending secured on housing, which reduces the cash ow available for
consumption and nancial investments.
Although HEW, as it is dened and calculated in this paper, is an aggregate cash ow
measure for the household sector as a whole, it is instructive to specify household level
individual actions which constitute the aggregate gure. Withdrawals and injections can
stem from a large number of individual micro level activities (Klyuev and Mills, 2006).
A withdrawal takes place when a household:
• sellsrealestatewithoutbuyingnewone(lasttimesales);3
• tradesdowntocheaperreal estate,while reducing themortgagebylessthan theprice
dierence;
• whenmoving,increasesitsmortgagebymorethanthedierenceinhouseprices;
• takes out a second mortgage or renances an existing one with higher principle
(remortgaging)withoutmovingproperties;
• increasesmortgage-backedconsumercredit.
An injection takes place when a household:
• makesadownpaymentonarst-timepurchaseofrealestate;
• makesamortisationandadditionalpaymentsonamortgage;
• remortgageswithalowerprincipal;
• purchasesasecondhomeandinvestmentpropertieswithcash;
• makeshomeimprovementsclassiedasinvestmentinhousingstock;
• reducesmortgage-backedconsumercredit.
e transactions of households with the nancial sector are quite straightforward – they
either increase or decrease the aggregate stock of mortgage backed loans for the whole
household sector. However, purchase and sale of real estate may take place with another
household or with an entity from another sector in the economy, and this distinction makes
a dierence form the point of view of aggregate H EW. In the former case, the buyer household
is an equity injector and the seller, correspondingly, is an equity withdrawer with a similar
amount. is real estate transaction has no impact on aggregate HEW, as the injection of
one household cancels out the withdrawal of the other. However, if a household buys real
estate from another sector (a new at from the enterprise sector, for instance), housing
equity is being injected by that transaction, and vice versa.
e aggregate or macro level HEW measure adds up all these dierent micro level
transactions from the household sector’s point of view. Hereby, if the household sector on
3 Households may move into rental accommodation or may have a spare at, house or area of la nd (stemming for
instance from a bequest or property restitution).

REB 2012
Vol. 4, N o. 1
24
ABEN • KUKK • STAEH R
aggregate has increased its mortgage backed loan stock by more than it has acquired housing
assets during a certain period of time, housing equity has been withdrawn.
e term housing should be interpreted broadly in this context as all real estate
transactions by the household sector (including land, wit h or without dwellings) are included
in the computation of HEW. An alternative term would be real estate equity withdrawal, but
housing equity withdrawal is the conventional and most recognisable term.
4. HEW Results for Estonia
HEW can be computed from two broad components: the change in the household sector’s
stock of loans secured by housing assets and the household sector’s net investment in
housing. e rst component is quite straightforward and can easily be calculated using
nancial sector statistics.
Concerning the second component, data on the household sector’s net investment in
dwellings and net acquisition of land are needed. Data series on both items are calculated by
Statistics Estonia (SE), but there may be concerns about whether the data fully takes into
account all real estate transactions between the household sector and other sectors. We
therefore supplement the data from Statistics Estonia with data from the register of real
estate contracts of the Estonian Land Board (LB) in order to calculate an alternative HEW
series. e register should capture all real estate transactions between dierent sectors. As
we are interested in net real estate investments by the resident household sector, the sales of
real estate to the business, government and foreign sectors are subtracted from the gross
investment.
Tab le 1. HEW and its Components for Estonia (EUR in Millions)
  2002 20 03 2004 2005 200 6 2007 2008 200 9 2010 2011
1. HEW (SE) = 3 – 6 16 163 237 646 1,082 780 4-572 -43 8 -481
2. HEW (LB) = 3 – 9 ....376 776 1,3 90 1,032 193 -379 -371 -334
3. NASFL = 4 + 5 185 393 585 1,142 1, 841 1, 511 631 -120 -156 -115
4. New mortgage-backed
consumer credit 4 9 18 62 181 200 83 -3 9 -25 -21
5. New mortgage-backed
housing loans 181 384 567 1,080 1,6 60 1,311 548 -81 -131 -94
6. NAHA (SE) = 7 + 8 169 230 348 495 759 731 626 453 281 366
7. Housing investment (SE) 206 274 365 542 904 994 718 491 466 550
8. Net acquisition of non-
produced assets (SE) -38 -44 -17 -47 -145 -263 -92 -38 -18 4 -184 a
9. NAHA (LB) = 10+11+12–13 .. .. 209 366 451 479 438 259 215 219
10. Purchases from other
sectors (LB) .. .. 282 549 789 652 417 200 207 237
11. Home improvements (SE) 63 83 81 108 153 181 174 127 112 143
12. Transfer costs (SE) 27 39 40 59 101 75 61 33 37 42
13. Sales to other sectors (LB) .. .. 195 350 592 429 213 101 141 203
a Estimate.
Source:
Authors’ calculations, Statistics Estonia, Estonian Land Board

RE B 20 12
Vol. 4, No. 1
25
ABEN • KUKK • STAEHR
Table 1 presents two measures of HEW in Estonia as well as the source data necessary for
their calculation. Row 1 shows HEW (SE) based on data from Statistics Estonia and row 2
shows HEW (LB) based on data from the Estonian Land Board. e estimation methodology
is detailed in Appendix 1, while an overview of the computations is provided below. All rows
present cash ows during the period. e source data for HEW (LB) starts from 2003Q3, so
2004 is the rst full year for which this measure can be calculated. Quarterly HEW series are
reproduced in Appendix 2.
e two main components of HEW are the change in the household sector’s stock of
loans secured by housing assets (the net acquisition of secured nancial liabilities, NASFL)
and the household sector’s net investment in housing (the net acquisition of housing assets,
NAHA). As positive NASFL generates cash to households, while positive NAHA implies
spending on real estate, the latter has to be subtracted from the rst in order to obtain HEW.
e net acquisition of secured nancial liabilities (NASFL, row 3) is calculated from the
change of the loan stock secured on housing assets (both new mortgage backed housing
loans and mortgage backed consumer credit). e household sector’s net investment in
housing (NAHA) is computed using two sources; NAHA based on data from Statistics
Estonia (SE) is given in row 6, while NAHA based on data from the Estonian Land Board
(LB) is given in row 9. When SE is used as the data source, housing investment and net
purchases of land are summed. When LB is used as the data source, construction and
improvements undertaken by households and transaction costs are added to the net cash
ow from real estate transactions (row 10 minus row 13).
Figure 1 shows the two HEW measures based on data from, respectively, Statistics
Estonia and the Estonian Land Board. e two HEW series co-vary closely within the
common sample, but the LB measure gives a higher value than the SE measure in all periods.
As described in Appendix 1, the SE version of HEW tends to underestimate the sale of real
estateassetstoothersectors;theLBversionmeasurestheowsfromrealestatetransactions
between the household sector and other sectors more precisely.
Figure 1. HEW for Estonia Using Two Different Data Sources (EUR in Millions)
Notes:
n.a. indicates that data is not available.
Source:
Table 1, Statistics Estonia and Bank of Estonia, authors’ calculations.
-600
-400
-200
0
200
400
600
800
1000
1200
1400
1600
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
HEW (SE) HEW (LB)
n.a.n.a.

REB 2012
Vol. 4, N o. 1
26
ABEN • KUKK • STAEH R
e dynamics of the HEW series reveal that the household sector withdrew large amounts
of equity from the housing stock during the period 2004–2007 with a peak in 2006. A sharp
reversal occurred in 2008 and HEW subsequently turned negative, implying that the Estonian
household sector injected liquidity into housing assets during the period 2009–2011.
Figure 2 shows the two main components of HEW, i.e. the net investment in housing
assets, NAHA (LB), and the change in the stock of mortgage backed loans, NASFL. e two
components exhibit dierent dynamics during the period 2004–2011. Net investment in
housing assets was relatively stable and positive in all years, implying negative housing
equity withdrawal. During the downturn starting in 2008, the household sector continued
toinvestinthehousingstock;evenifpurchasesofnewdwellingsfromtheenterprisesector
were modest, own construction and repair still contributed to injections.
Figure 2. Equity Withdrawal from Net Acquisition of Housing Assets, NAHA (LB), and Net Acquisition
of Secured Financial Liabilities, NASFL (EUR in Millions)
Source:
Table 1, Statistics Estonia and Bank of Estonia, authors’ calculations.
Mortgage-backed borrowing has shown much more volatility. During the period 2004–
2008 the secured loan stock rose far more than the household sector’s net investment in
housing. e substantial increase of the stock of housing loans, compared to housing
investment, is due to the fact that about half of the real estate contracts were between two
households;ifsuchpurchasesarenancedbymortgageloans,thecashowfromthebank
ends up in the hands of the household sector.
During the period 2004–2008 the household loan stock increased rapidly as initially it
was at a low level: the ratio of debt to disposable income increased from 45% in 2004 to 91%
in 2008 (Eurostat, 2012). e growth followed looser credit conditions coinciding with large
capital inows (Brixiova et al., 2010). e share of mortgage-backed loans in the total
household loan stock increased from 55% in 2002 to 85% in 2007, indicating a tight
connection between developments in real estate and credit markets.
e household loan stock began to decrease from the beginning of 2009. e contraction
occurred due to the global nancial crisis and concerns about the sustainability of the stock of
debt accumulated by the household sector, while demand for borrowing shrank due to
increased income risk. In periods, new lending to the household sector virtually ceased. As
activity in the real estate market decreased, amortisation of mortgages exceeded the amount of
-500
0
500
1000
1500
2000
2004 2005 2006 2007 2008 2009 2010 2011
NAHA (LB)
NASFL
HEW (LB)

RE B 20 12
Vol. 4, No. 1
27
ABEN • KUKK • STAEHR
new loans. Consequently, in addition to investment to housing assets, mortgage-backed loans
contributed to the net injection of equity into housing stock during the period 2009–2011.
e LB data allow a disaggregation of the ows from dierent types of real estate: ats,
land with buildings and land without buildings. It follows from Figure 3 that the household
sector bought more ats from other sectors than vice versa during the whole sample period.
Figure 3. Composition of Net Flows of Transactions Based on LB Dataset (EUR in Millions)
Notes:
Values of real estate contracts only; transaction fees and own construction and repairs are not included.
Source:
Table 1, Statistics Estonia and Bank of Estonia, authors’ calculations.
During 2004–2011, the household sector sold more land without buildings than it bought
fromothersectors;netlandsalescontributedtoaboutonetenthofoverallHEW.Someof
the real estate stock that was obtained by restitution during 1990s could be sold on favourable
conditions during the economic boom, consequently generating cash to the household
sector.4 In the process of restitution around 33% of the land (incl. residential property) has
been distributed among a large proportion of households, though many of them did not live
on the property (Giovarelli and Bledsoe, 2001). Signicant resources have been withdrawn
from land property especially during the economic boom period and the explanation can be
ownership of “excess land” that could be easily liquidised during the vigorous economic
growth period, accompanied by increasing real estate prices.
e development of real estate prices contributed to the volatility of HEW series. Before
thecrisis realestatepricesincreasedveryrapidly; atthe endof2007 theaverage pricesof
ats and land without buildings were 2–3 times higher than at the beginning of 2003. e
reversal of prices was pronounced and very rapid until the stabilisation of prices in 2009
(Estonian Land Board, 2012).
To put the Estonian results into an international context, Figure 4 presents HEW (as a
share of household disposable income and in millions of GBP) for the UK and Estonia. e
UK data are from the Bank of England statistical database.5 e overall dynamics of HEW
4 e widespread ownership of real estate made it possible for the households to sell their spare la nd to real estate
developers, who sold the properties back to households, but with dwellings on them.
5 Bank of England Statistics, table LPQB3VH (Quarterly percentage of total sterling housing equity withdrawal
(previouslycalledmortgage equity withdrawal)byindividuals(inpercent)seasonally adjusted);http://ww w.
bankofengland.co.uk/mfsd/iadb/fromshowcolumns.asp?travel=nix&searchtext=housing+equity+withdrawal
&point.x=16&point.y=8.
-200
-100
0
100
200
300
400
500
2004 2005 2006 2007 2008 2009 2010 2011
Net purchase of flats
Net purchase of land without buildings
Net purchase of land with buildings
Net purchase, all

REB 2012
Vol. 4, N o. 1
28
ABEN • KUKK • STAEH R
in Estonia is very similar to that of the UK, but the volatility in Estonia is much higher. In
the UK, HEW as a share of household disposable income reached 5.1% in 2006, while in
Estonia the ratio was 21% or 16%, depending on the calculation method, in the same year.
Such high ratios of HEW to disposable income have not been reported for any other country.
During 2009–2011, the dierence between the HEW measures for the UK and Estonia is
somewhat smaller than observed during the years of rapid economic growth. In 2011 HEW
was -3.5% of disposable income for the UK, while HEW based on Land Board data was -5.5%
of disposable income for Estonia.
Figure 4. HEW as a Percentage of Household Disposable Income, UK and Estonia, 2004–2011
Notes:
The Variable for Estonia is HEW (LB).
Source:
Table 1, Statistics Estonia and Bank of Estonia, authors’ calculations.
6. The Relationship between HEW and Consumption in Estonia
is section discusses the linkages between HEW and household consumption in Estonia.
Figure 5 shows the two measures of HEW as well as household saving as percentages of
household disposable income. e household saving rate is the reverse mirror of the average
consumption propensity of the household sector. e gure illustrates that the household
sector attained substantial liquid funds from HEW during the growth period 2003–2007, a
period in which consumption consistently exceeded household disposable income, resulting
in a negative saving rate. HEW amounted to 15–20% of disposable income in 2006, and in
the same year household consumption outstripped disposable income by 6%.
e expectations of households and the resulting consumption aspirations may be an
important factor for the volatile HEW dynamics in Estonia. e integration into the
European economy and the convergence process was accompanied by a rapid increase in
disposable income until 2006. Improved condence and expectations of rapidly increasing
income also boosted consumption aspirations (Becker et al., 2010). e liquidising of real
estate assets comprised an opportunity to obtain liquid funds for consumption. e global
nancial crisis and the deteriorating outlook may have led households to reconsider their
income prospects and postpone consumption in order to consolidate their nances and pay
back housing loans, which ceteris paribus would reduce HEW.
-5
0
5
10
15
20
25
2004 2005 2006 2007 2008 2009 2010 2011
UK Estonia

RE B 20 12
Vol. 4, No. 1
29
ABEN • KUKK • STAEHR
Figure 5. HEW and Household Saving as Percentage of Household Disposable Income, 2002–2011
Source:
Table 1, Statistics Estonia and Bank of Estonia, authors’ calculations.
Figure 5 suggests a close correlation between HEW and household saving or, conversely,
between HEW a nd household consumption. However, such co-variation may be coincidental
or spurious. We will use econometric modelling to provide a more detailed analysis of the
connection between HEW and consumption.
e modest number of observations implies that we need to specify very parsimonious
consumption models. e an alysis includes quar terly data of four variables, v iz. consumption,
household disposable income and the two measures of housing equity withdrawal. When
HEWfromStatisticsEstoniaisused,thesampleis2002Q1–2011Q4,intotal40observations;
when HEW from the Land Board is used, the sample is 2003Q3–2011Q4, in total 34
observations. e series are deated using the quarterly average of the monthly HICP price
index. e series HEW (LB) and HEW (SE) are from Appendix 2, quarterly household
consumption and the monthly HICP index are from Eurostat (2012) and household
disposable income is from the Bank of Estonia.6 e following notation is used for the series:
real consumption is RCONS, real disposable household income is RINC and real housing
equity withdrawal is RHEWLB when based on data from the Land Board and RHEWSE
when based data from Statistic Estonia.
e series RCONS, RINC, RHEWSE and RHEWLB exhibit substantial persistence.
Augmented Dickey-Fuller tests suggest that the four series represent borderline cases
between integration of order one and integration of order two. Unit root tests typically
possess little power in small samples, which may explain the borderline results. In the
following we will treat the series as integrated of order one and, consequently, look for
cointegration between the variables depicting consumption, income and housing equity
withdrawal.
Preliminary investigation using the Johansen methodology showed the presence of one
co-integrating vector. Moreover, estimation of dynamic adjustment indicated that the
6 ese data series are provided by the Bank of Estonia. Statistics Estonia provides data on household disposable
income from the national accounts but only on an annual basis and the Bank of Estonia therefore computes a
quarterly series for the purpose of macroeconomic modelling. e quarterly data are computed from the same
components as used for the annual data, but some components are interpolated.
-10
-5
0
5
10
15
20
25
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
HEW (LB)
HEW (SE)
Household saving

REB 2012
Vol. 4, N o. 1
30
ABEN • KUKK • STAEH R
adjustment mainly took place via changes in consumption. We will not present the results
here as the small number of observations in the dataset combined with the estimation of
many coecients implies that the results using the Johansen methodology lack robustness.
In many cases changes in the sample length and changes in the lag structure aect the
estimated coecients and standard errors considerably.
We will instead use the more robust Engle-Granger two-stage methodology (Engle and
Granger,1987;Veerbek, 2012,Sec. 9.2).e Engle-Granger methodology entailsthatthe
long-term cointegrated relation and the short-term adjustment relation are estimated in two
separate stages, where the estimation of the short-term adjustment relation in the second
stage is conditional on cointegration being detected in the rst stage.
e rst stage entails estimation of a “long-term” relation in which the level of the
dependent variable is regressed on the levels of the explanatory variables using ordinary
least squares. To rule out spurious correlation, the variables in the long-term relation must
be cointegrated, which entails that the residuals must be stationary (residual-based
cointegration test). e augmented Dickey-Fuller test is a unit root test with the null
hypothesis that the residual contains a unit root. In case of cointegration, the adjustment
can be estimated in a second stage in which changes in the dependent variable are modelled
as a function of the lagged residual from the rst stage as well as lagged changes in the
dependent variable and current and lagged changes in the explanatory variables. e
estimated coecient of the lagged residual provides information on the extent and speed of
error correction taking place through the dependent variable.
In the rst stage, real household consumption RCONS is the dependent variable in all
cases, while the real income RINC and RHEW (where RHEW is either RHEWLB or
RHEWSE) are the explanatory variables. e rst-stage estimation is shown in eq. (1), where
the index t denotes the quarter and takes all values within the sample.
Quarterly dummies are included in all regressions to account for seasonality in data but
not shown. e residuals are denoted ε(t). e coecients α0 and α1 are estimated using
OLS. e standard errors of the rst stage estimation follow a non-standard distribution due
to the variables exhibiti ng unit roots, and it is therefore not possible to ascertain t he statistical
signicance of the estimated coecients.
Cointegration requires that the residuals ε(t) are stationary. is is tested using an
augmented Dickey-Fuller test with the null hypothesis that the residuals contain a unit root.
e residuals ε(t) do not follow a standard Dickey-Fuller distribution since they result from
an estimated equation, but tabulated critical values are readily available. In case of
cointegration, the residuals ε(t) can be considered deviations from a long-term “equilibrium”
relation.
In case of cointegration, the short-term adjustment equation can be estimated using
OLS, as all variables, including the residuals ε(t), are stationary. e short-term equation is
givenineq.(2);theoperatorΔdenotesquarter-on-quarterchangeandξ(t) is the residual in
this case.
RCONS(t)=Constant+α0RINC(t)+α1R HEW(t)+ε(t) (1)
4 4 4
ΔRCONS(t)=Constant+
Σ
βiΔRCONS(t-i)+
Σ
δiΔRINC(t-i)+
Σ
γiΔRHEW(t-i)+λε(t-1)+ξ(t) (2)
i=1i=1i=1

RE B 20 12
Vol. 4, No. 1
31
ABEN • KUKK • STAEHR
Quarterly dummies are included but not shown. e short-term adjustment coecients
βi, δi, γiand λ areto be estimated.ecoecient λ depicts theshort-termadjustment to
lagged deviations from the long-term “equilibrium” relation estimated in the rst stage. e
coecient is expected to be negative and statistically signicant in which case deviations
from the long-term relation are corrected through short-term changes in consumption. In
caseofanegativeλ,thewholetermλε(t – 1) is called the error correction term.
Table 2 shows the results of the Engle-Granger estimations, specically the full results of
the rst stage, the results of the cointegration test (in the second last line) and the estimated
adjustment or error correction coecient of the second stage (in the last line). Detailed results
for the estimation of the short-term adjustment of the second stage are shown in Appendix 3.
Tab le 2. Estimation of Long-Term Relation and Error Correction
(1) (2) (3) (4)
RINC 0.849
(0.048)
0.899
(0.031) .. ..
RINC (pre-crisis) .. .. 0.950
(0.038)
1.193
(0.043)
RINC (post-crisis) .. .. 1.116
(0 .175)
1.627
(0 .110)
RHEWLB 0.747
(0.054) .. .. ..
RHEWSE .. 0.840
(0.052) .. ..
RHEWSE (pre-crisis) .. .. 0.592
(0 .108) ..
RHEWSE (post-crisis) .. .. 0.981
(0.252) ..
Constant 199.7
(74 . 85)
145.6
(49.75) .. ..
Constant (pre-crisis) .. .. 99.5
(46.6)
-179.7
(6 7.9)
Constant (post-crisis) .. .. -208.8
(307.0)
-1124 . 8
(193. 3)
R20.952 0.970 0.980 0.980
DW 0.898 1.13 6 1.575 1.575
Time 2003Q3–2011Q4 2002Q1–2011Q4 2002Q1–2011Q4 2002Q1–2011Q4
Observations 34 40 40 40
H0: Unit root in residualsa -3. 032 -3.860 -4.895 -4.064
Adjustment coefficient .. -0.621
(0 .18 1)
-0.689
(0. 246)
-0.418
(0.167)
a Asymptotic critical values for ADF unit root test with three variables and the null hypothesis of no
cointegration: -4.29 at 1% level, -3.74% at 5% level and -3.45 at 10% level (Veerbek 2012, p. 345).
Notes: OLS estimation. The dependent variables are RCONS in the long-term relation and the quarterly
change ΔRCONS in the short-term adjustment relation. Quarterly dummies are included in all estima-
tions, but the results have not been reported. Standard errors are shown in brackets. The standard
errors in the long-term relation follow a non-standard distribution so the statistical significance of the
estimated coefficients cannot be ascertained.
Source: Authors’ calculations

REB 2012
Vol. 4, N o. 1
32
ABEN • KUKK • STAEH R
Column (1) in Table 2 shows the results when the variable RHEWLB, based on data from
the Land Board, is used as the measure of housing equity withdrawal. e estimated
coecient to the income measure is 0.85 and thus close to one, while the est imated coecient
to the HEW measure is 0.75. e lag structure of the dierenced variables in the augmented
Dickey-Fuller test is found using the Akaike information criterion. e test statistic is -3.032
which implies that the null hypothesis of a unit root cannot be rejected. e number of
observations is low, which may explain that the hypothesis of a unit root in the residual
cannot be rejected. Due to the absence of cointegration, no short-term adjustment equation
has been estimated in this case.
To attain more observations, RHEWSE based on data from Statistics Estonia is used as
the measure of housing equity withdrawal. e results are shown in Column (2). e
estimated coecient of the income variable is largely unchanged, while the coecient of
RHEWSE is slightly higher than the one for RHEWLB in Column (1). is result seems
reasonable given that RHEWSE exhibit less variability than RHEWLB. e hypothesis of a
unit root can be rejected at the 5% level, which suggests that the three variables RCONS,
RINC and RHEWSE are cointegrated and the residual therefore can be interpreted as an
error correction term depicting deviations from a long-term “equilibrium” relation between
the variables.
e speed with which deviations from the long-term relation are closed can be found
from the short-term adjustment relation. e adjustment is modelled in eq. (2): the change
in consumption is regressed on the lagged error correction term as well as the change in
consumption (three lags), the change in income (current and three lags) and the change in
housing equity withdrawal (current and three lags). e coecient to the lagged error
correction term is estimated at -0.621, which implies that deviations from the long-term
relation are essentially eliminated within a couple of quarters. e result is qualitatively
similar if statistically insignicant coecients are removed from the adjustment regression
using a general-to-specic methodology.
In sum, the results in Column (2) suggest that real consumption is closely related to real
income and real HEW in the longer term. An increase in income of 100 EUR is associated
with an increase in consumption of 89 EUR, while an increase in HEW of 100 EUR is
associatedwithanincreaseinconsumptionof84EUR.elatterresultimpliesthat5/6of
liquidised housing assets are consumed over time. Deviations from the long-term relation
are eliminated very fast.7
Estonia entered recession in the fourth quarter of 2007 and subsequently experienced a
pronounced downturn. Column (3) shows the results when the explanatory variables from
Column (2) are interacted with a dummy for the pre-crisis period (2002Q1–2007Q3) and a
dummy for the post-crisis period (2007Q4–2011Q4), thus allowing dierent estimated
eects for the two periods. e marginal eects of income are rather similar across the two
periods and are estimated to be around one. e marginal eects of housing equity
withdrawal, however, vary considerably across the two periods. e marginal eect of
RHEWSE is around 0.6 in the pre-crisis period when the housing equity withdrawal was
7 is paper focuses on the applicability of the HEW variable in models of consumption. Some experimentation
with the inclusion of house price indices instead of the HEW provided unsatisfactory resu lts (not shown).
Consumpt ion, income and house pri ces were not cointegrate d as the null hy pothesis of a unit ro ot in the residual
from the  rst stage regression could not be reje cted even at the 10% level. Moreove r, the coecient of t he income
term changed markedly and t he regression was not robust to even minor sample changes.

RE B 20 12
Vol. 4, No. 1
33
ABEN • KUKK • STAEHR
positive, while it is around 1 for the post-crisis period when housing equity withdrawal was
negative. e upshot is that while the household sector withdrew liquidity from housing
assets, a little more than half of the HEW was consumed, but when it was injecting liquidity
into housing assets it coincided with a one-to-one compression of consumption. e latter
result may be related to tight credit conditions and a need to deleverage the balance sheets of
households by decreasing the high debt levels (Meriküll, 2012).
Column (4) repeats the estimation from Column (3) but leaves out the interacted HEW
variables. e result is interesting as the estimated coecient of the income variable increases
and the estimated coecient for the post-crisis period of 1.63 is substantially larger than the
corresponding coecient estimate in the model in which HEW is included. e upshot is
that by omitting the HEW measure, the coecients of the remaining variables, in this case
the income variable, may become biased. e falling consumption in the crisis period is not
the result of an extreme overreaction to the falling income, but in large part the result of
substantial housing equity injections.
Overall the results in this section suggest that the liquidising of housing assets plays an
important role for consumption in Estonia although the eect has varied across the business
cycle. e results for Estonia are in line with ndings for some other countries, cf. the
literature survey in Section 2. e cross-country study by Catte et al. (2004), for instance,
estimates that 89% of HEW was consumed in the United Kingdom, 63% in Canada and
Australia, and 20% in the United States. ere are no studies that have investigated a possible
asymmetric reaction of consumption to HEW in the business cycle.
7. Final Comments
ere is substantial disagreement in the academic and policy-oriented literature about the
size of the eect of housing wealth on household consumption and the channels through
which the eect takes place. In any case housing wealth must be liquidised before it can be
translated into a consumption response. Household equity withdrawal depicts the liquid
funds or cash ows generated by the household sector from otherwise illiquid housing
assets.
is paper provides data on housing equity withdrawal in Estonia for the period 2002–
2011 and assesses the impact of HEW on household consumption. e data show that the
amount of housing equity withdrawal was substantial during the economic boom in 2004–
2007, in particular in 2006 when HEW amounted to 15–20% of household disposable
income. From 2008 the household sector injected cash into housing assets as the global
nancial crisis led the banking sector to curtail lending; the housing equity injection
amounted to around 5% of income during the period 2008–2011. e results reect that the
HEW series is very volatile for Estonia. e volatility comes mainly from one of the two
HEW components, viz. mortgage-backed loa ns, while the net investment in housing exhibits
a more stable trajectory.
Catte et al. (2004) argue that HEW is very important in countries with developed
mortgagemarkets;wend thatHEWexhibitssubstantial variation in Estonia, a country
that experienced fast changes in nancial markets during the sample period. A number of
factors may help explain the very volatile development of HEW in Estonia during the decade
of 2002–2011. First, the nancial sector has undergone rapid changes, initially with a rapid

REB 2012
Vol. 4, N o. 1
34
ABEN • KUKK • STAEH R
expansion of the loan stock and subsequently with a retrenchment of credit provisioning.
Second, real estate prices followed a strong pro-cyclical pattern and consequently changed
the scope of housing equity withdrawal across time. ird, the restitution and privatisation
of land and housing assets meant that many households possessed excessive real estate (not
backed with mortgage) which could be liquidised as soon as favourable conditions appeared.
Fourth, sentiments have changed markedly over the decade with follow-on eects on the
consumption, saving and portfolio preferences of the households.
e analysis showed a substantial correlation between HEW and consumption during
theperiod2002–2011;thedynamicsof HEWis animportantcomponentofconsumption
behaviour. Econometric analysis, however, reveals that the linkage between HEW and
consumption shis over time. During the rst part of the sample, 2002–2007, Estonia
experienced an economic boom. In this period HEW was positive and attained very high
values, but only approximately half of the cash generated was carried into consumption.
During the second part of the sample, 2008–2011, Estonia experienced a deep economic
crisis. In this period HEW was negative, and the housing equity injection appears to have
been entirely nanced by lower consumption. e dierent household behaviour across the
business cycle might reect dierences in credit conditions in Estonia and the need of
households to adjust their balance sheets.
Several directions for further research may be suggested. First, further empirical
evidence is needed to provide a thorough understanding of the use of HEW in nancial
management at the household level. One issue of particular importance is how HEW is
divided between changes in consumption and nancial assets. Second, a comparison of
developments in HEW across European countries would undoubtedly produce additional
information on the linkages between housing assets, consumption and business cycle
developments. It would be particularly interesting to ascertain to which extent the linkage
between HEW and consumption diers across the developed countries in Western Europe
and the transition countries in Eastern Europe. ird, restitution and privatisation of land
and housing has taken place in all CEE countries (Giovarelli and Bledsoe, 2001). e
implications of the privatisation on the balance sheets, consumption and saving of the
household sector warrant further investigation. Finally, it may be instructive to estimate
consumption models wit h a richer set of explanatory variables, including variables capturing
housing prices, interest rates and consumer condence. Such complex modelling will likely
yield the best results if a large number of observations are available.
References
Aron, J., Duca, J.V., Muellbauer, J., Murata, K., and Murphy, A. 2011. Credit, Housing
Collateral, and Consumption: evidence from Japan, the U.K., and the U.S. Review of
Income and Wealth. Availableat:http://dx.doi.org/10.1111/j.1475-4991.2011.00466.x.
Bank of England. 2011. Explanatory Notes – e Bank’s Estimate of Housing Equity
Withdrawal. Availa ble at: http://ww w.ba nkofengland.co.uk /mfsd/iadb/notesiadb/hew_
notes.htm.
Bank of Estonia.I nt e r n e tD a t a b as e . Av a i l a b le a t : ht t p : // w w w . e es t i p a n k . ee / p ub / en / d ok u m e n d id /
statist ika/ [Statistica l Indicators, Financial S ector Statistics, 3.3. Lo ans, 3.3.3 Stock and
number of loans granted to households by type of loan, currency and collateral]

RE B 20 12
Vol. 4, No. 1
35
ABEN • KUKK • STAEHR
Becker, T., Daianu, D., Darvas, Z., Gligorov, V., Landesmann, M., Petrovic, P., Pisani-Ferry, J.,
Rosati, D., Sapir, A., and Weder di Mauro, B. 2010. Whither Growth in Central and Eastern
Europe? Policy Lessons for an Integrated Europe. Bruegel Blueprint Series, No. 11.
Benito, A and Power, J. 2004. Housing Equity and Consumption: Insights from the Survey
of English Housing. Bank of England Quarterly Bulletin, Autumn, pages 302–309.
Benito, A. 2009. Who Withdraws Housing Equity and Why? Economica, Vol. 76, No. 301,
pp. 51-70.
Bloxham P., McGregor, D. and Rankin, E. 2010. Housing Turnover and First-home Buyers,
Reserve Bank Of Australia Quarterly Bulletin,June.Availableat:http://www.rba.gov.au/
publications/bulletin/2010/jun/1.html
Boone, L., Girouard, N. and Wanner, I. 2001. Financial Market Liberalization, Wealth and
Consumption. OECD Economics Department Working Paper, No. 308.
Brixiova, Z., Vartia, L., and Worgotter, A. 2010. Capital Flows and the Boom-bust cycle: e
Case of Estonia. Economic Systems, Vol. 34, No. 1, pp. 55-72.
Buiter, W.H. 2010. Housing Wealth isn’t Wealth. Economics E-Journal. Vol. 4, No. 2010-22,
Availableat:http://www.economics-ejournal.org/economics/journalarticles/2010-22.
Catte, P., Girouard, N., Price, R., and Andre, C. 2004. e Contribution of Housing Markets
to Cyclical Resistance. OECD Economic Studies, Vol. 38, No. 1.
Cooper, D. 2010. Did Easy Credit Lead to Overspending? Home Equity Borrowing and
Household Behavior in the 2000s. FRB Boston Public Policy Discussion Paper Series, No.
09-7.
Ebner, A. 2010. A Micro View on Home Equity Withdrawal and its Determinants. Evidence
from Dutch Households. Discussion Paper, No. 2010-2, Department of Economics,
University of Munich.
Engle, R. F. and Granger, C.W. 1987. Co-integration and Error Correction: Representation,
Estimation and Testing. Econometrica, Vol. 55, No. 2, pp. 251-276.
E S . 2011. Statistical Yearbook of Estonia. 2011. Tallinn: Statistics Estonia.
Estonian Land Board. 2012.RealProper tyPriceIndices.Availableat:http://w w w.maaa met.
ee/data/Real_property_price_indices___3.qt_20114.pdf
Eurostat. 2012. Statistics Database. Available at: http://epp.eurostat.ec.europa.eu/portal/
page/portal/eurostat/home/.
Friedman, M. 1957. A eory of the Consumption Function. Princeton: Princeton University
Press.
Giovarelli, R. and Bledsoe, D. 2001. Land Reform in Eastern Europe. Western CIS,
Transcaucuses, Balkans, and EU Accession Countries. Rural Development Institute,
Seattle.Availableat:p://p.fao.org/docrep/fao/007/AD878E/AD878E00.pdf.
Greenspan, A. and Kennedy, J. 2008. Sources and Uses of Equity Extracted from Homes.
Oxford Review of Economic Policy, Vol. 24, No.1, pp. 120-144.
Holmes, G.J. 1993. Housing Equity Withdrawal and the Average Propensity to Consume.
Applied Economics, Vol. 25, No. 10, pp. 1315-1322.
Hurst, E. and Stafford, F. (2004). Home is Where the Equity is: Mortgage Renancing and
Household Consumption. Journal of Money, Credit and Banking, Vol. 36, No. 6, pp. 985-
1014.
Klyuev, V. and Mills, P. 2006. Is Housing Wealth an ‘ATM’? e Relationship Between
Household Wealth, Home Equity Withdrawal, and Saving Rates. IMF Sta Paper, Vol.
54, No. 3, pp. 539-561.

REB 2012
Vol. 4, N o. 1
36
ABEN • KUKK • STAEH R
Läänemets, L. and Mert sina, T. 2009. Behaviour of Households in the Years of Fast Economic
Growth. Quarterly Bulletin of Statistics Estonia,No.4/09,pp.59-64,Availableat:http://
www.stat.ee/dok umend id/51830.
Meriküll, J. 2012. Households Borrowing During Creditless Recovery. Bank of Estonia
Working Papers, No 2.
Modigliani, F. 1966. e Life Cycle Hypothesis of Saving, the Demand for Wealth and the
Supply of Capital. Social Research, Vol. 33, No. 2, pp. 160-217.
Muellbauer, J. and Murphy, A. 1990. Is the UK Balance of Payments Sustainable? Economic
Policy, Vol. 5, No. 11, pp. 345-383.
Muellbauer, J. 2008. Housing, Credit and Consumer Expenditure. CEPR Discussion Paper
No. 6782.
Paabut, A. and Kattai, R. 2007.KinnisvaraväärtusekasvumõjueratarbimiseleEestis[e
Eect of Property Price Increases on Private Consumption in Estonia], Eesti Pank
Working Papers,No.5/2007.
Paiella, M. 2009. e Stock Market, Housing, and Consumer Spending: A Survey of
Evidence on Wealth Eect. Journal of Economic Surveys, Vol. 13, No. 5, pp. 947-973.
Schwartz, C., Lewis, C. and Norman, D. 2008. Factors Inuencing Housing Equity
Withdrawal: Evidence from a Microeconomic Survey. Economic Record, Vol. 84, No. 267,
pp. 421-433.
Smith, M. 2010. What do We Know About Equity Withdrawal by Households in New
Zealand? In: S.J. Smith and B.A. Searle (Eds.). e Blackwell Companion to the Economics
of Housing: e Housing Wealth of Nations, Ch. 8. Oxford: Wiley-Blackwell.
Smith, S.J. and Searle, B.A. 2008. Dematerialising Money? Observations on the Flow of
Wealth from Housing to Other ings. Housing Studies, Vol. 23, No. 1, pp. 41-43.
Sonje, A.A., Casni, A.C. and Vizek, M. 2012. Does Housing Wealth Aect Private
Consumption in European Post-transition Countries? Evidence from Linear and
reshold Models. Post-Communist Economies, Vol. 24, No. 1, pp. 73-85.
Van Els, P., van den End, W. and van Rooij, M. 2005. Financial Behaviour of Dutch
Households: Analysis of the DNB Household Survey 2003. BIS Papers, No. 22
(Investigating the Relationship Between Financial and Real Economy), pp. 21-40.
Veerbek, M. 2012. A Guide to Modern Econometrics, Chichester: John Wiley and Sons, 4th ed.
Westaway, P.F. 1993. Mortgage Equity Withdrawal: Causes and Consequences. NIESR
Discussion Paper, No. 59.

RE B 20 12
Vol. 4, No. 1
37
ABEN • KUKK • STAEHR
Appendix 1. Data and HEW Calculation Methodology
HEW is equal to the net acquisition of secured nancial liabilities (NASFL) minus the net
acquisition of housing assets (NAHA). As HEW is not regu larly calculated in most countries,
no single, precise and internationally agreed denition is available. e following
methodology follows the one used by the Bank of England (2011):
HEW= NetAcquisitionofSecuredFinancialLiabilities(NASFL)
– Net Acquisition of Housing Assets (NAHA)
e net acquisition of secured nancial liabilities can be calculated from the quarterly or
yearly change of the loan stock secured on housing assets. e source data can be obtained
from the Bank of Estonia.8 e change in the stock during the period includes both new
granted loans and the amortisation of existing ones.9 e NASFL measure can be either
positive or negative, as the loan stock can increase or decrease during any period.
NASFL=Netacquisitionofsecuredhousingloans
  +Netacquisitionofmortgage-backedconsumercredit
Concerning the net acquisition of housing assets (NAHA), we are interested in all real
estate transactions between the household sector and other sectors, because all such
transactions generate cash ows. One possibility to capture these ows is to use housing
investment of the household sector as computed by Statistics Estonia (SE) for the GDP
calculations. is measure includes purchases of new dwellings, construction of dwellings
and home improvements done by the household sector and all costs associated with the
transfer of ownership, e.g. dealer fees, lega l fees and state duties. It should in principle include
net purchases of existing dwellings from other sectors, but methodological requisites imply
that the sale of real estate by the household sector to other sectors may be underestimated
and the series may be subject to revision if additional data sources were used.10
In addition to dwellings, we also have to take into account transactions involving the
purchase and sale of land, which are not considered investments but rather acquisition of
non-produced non-nancial assets and are presented in the non-nancial accounts of
national accounts, row code K2 (Läänemets and Mertsina, 2009). us, using SE as a data
source, the second component of HEW is just the sum of housing investment and net
purchases of land.
NAHA(SE)=Housinginvestment(SE)
 +NetpurchaseoflandK2(SE)
8 SeeBa nkofEstonia,http://statistika.eestipan k.ee/?lng=en#t reemenu/nantssektor/147/650,Table“3.3.3Stock
and number of loans granted to households by ty pe of loan, currency and collateral”.
9 e source statistics do not allow a distinction between the two subcomponents, but it is not necessary for the
presentpurpose.Interestpaymentsdonotappearinthecalculations;interestpaymentsareconsiderednegative
capital income and are thus part of disposable income and do not inuence HEW.
10 is information was provided by Tõnu Mertsina from Statistics Estonia.

REB 2012
Vol. 4, N o. 1
38
ABEN • KUKK • STAEH R
Another possibility to capture the net cash ow from real estate transactions between the
household sector and other sectors is provided by the data on real estate contracts collected
by the Estonian Land Board (LB). As this source should include precise data from all
contracts, the net ow of funds for the household sector can be calculated. Because this data
source does not capture construction and improvements undertaken by households or
transaction costs, these components have to be taken from the dwelling data from SE and
added to the LB ows.
NAHA(LB)=Purchasesofdwellingsandlandfromothersectors(LB)
 +Homeimprovementsandownconstruction(SE)
 +Transactioncosts(SE)
– Sales of dwellings and land to other sectors (LB)
In conclusion, we can calculate two dierent HEW series using dierent data sources.
HEW(SE)=NASFL–NAHA(SE)
HEW(LB)=NASFL–NAHA(LB)
As explained above, the two measures will not fully coincide because of dierences in the
methodologies use d by SE a nd LB to calculate housing ass ets’ acquisition. Two ma in dierences
can be highlighted. First, the LB measure includes all real estate transactions between the
household sector and other sectors, while the SE measure may underestimate the sales of real
estate by the household sector to other sectors. is has the eect of increasing the LB measure
relative to the SE measure. Second, the LB measure likely underestimates transaction fees, as
only housing-related transaction fees are taken into account (from the SE dwelling investment
statistics) while fees related to land sales are omitted. Both of these factors widen the gap
between the two measures, but the rst is arguably more important than the latter in
quantitative terms. Hence, if the purpose of the use of HEW series is to track all monetary
ows between the household sector and other sectors, the LB measure would be the most
appropriate, while the longer series of HEW (SE) can be used as a reference.

RE B 20 12
Vol. 4, No. 1
39
ABEN • KUKK • STAEHR
Appendix 2. HEW for Estonia
Tab le B1. HEW for Estonia, Quarterly Data, No
Seasonal Adjustment (EUR in Millions)
Note:
HEW (SE) is based on data from Statistics
Estonia and HEW (LB) on data from the Es-
tonian Land Board. HEW (LB) only available
from 20 03Q3.
HEW (SE) HEW (LB)a
2002 Q1 -0.7 ..
Q2 6.8 ..
Q3 -0.9 ..
Q4 11.3 ..
2003 Q1 29.8 ..
Q2 49.4 ..
Q3 34.1 68.9
Q4 49.4 65.3
2004 Q1 19.7 66.7
Q2 62.5 97.4
Q3 72.6 97. 5
Q4 82.5 114 .7
2005 Q1 83.8 111. 6
Q2 163.2 196. 2
Q3 169.7 183.2
Q4 229.8 284.7
2006 Q1 212. 3 265.0
Q2 301.4 373.8
Q3 291. 8 360.9
Q4 276.0 390.0
2007 Q1 230.8 322.2
Q2 260.7 33 7.9
Q3 172.2 218.3
Q4 116 .2 153 . 4
2008 Q1 42.5 121.2
Q2 65.9 110 . 6
Q3 4.7 32.5
Q4 -10 8 . 6 -7 1 .7
2009 Q1 -123 . 5 -80.3
Q2 -159. 6 -1 05.0
Q3 -14 8 .4 -99.3
Q4 -141.0 -94.5
2010 Q1 -95.3 -94.4
Q2 -10 4 . 2 - 87.8
Q3 -111.7 -97.3
Q4 -126 . 3 -91.9
20 11 Q1 -110.5 -87. 8
Q2 -96.7 -61.7
Q3 -120 .1 -83.3
Q4 -153 . 5 -101.6

REB 2012
Vol. 4, N o. 1
40
ABEN • KUKK • STAEH R
Appendix 3. Short-Term Dynamics and Adjustment
Tab le C1. Estimation of Short-Term Adjustment
(2) (3) (4)
RCO NS (-1) 0.233
(0 . 212)
0.192
(0.230)
0.10 0
(0.222)
RC O NS (-2) 0.435
(0.205)
0.404
(0.221)
0.366
(0.225)
RCON S (-3) 0. 374
(0 .16 4 )
0.324
(0.172)
0.316
(0 .17 7)
RINC 0.714
(0 .17 8)
0.78 0
(0.195)
0.713
(0 .19 7)
RINC (-1) 0.278
(0 . 213)
0.221
(0. 244)
0. 317
(0.238)
RINC (-2) -0.425
(0.222)
-0.423
(0. 244)
-0.329
(0.239)
RINC (-3) -0.260
(0 .176)
-0.258
(0 .19 1)
-0.259
(0 .19 8 )
RHEWSE 0.633
(0 .19 1)
0.607
(0.204)
0.481
(0.208)
RHE WSE (-1) 0.287
(0.227)
0.393
(0.233)
0.666
(0 . 215 )
RH EWSE (-2) -0.354
(0.333)
-0. 217
(0.346)
-0.054
(0.339)
RH E WSE (-3) -1. 0 28
(0 . 311)
-0.800
(0.314)
-0.735
(0.319)
Residual from
first stage (-1)
-0.622
(0 .18 1)
-0.689
(0. 246)
-0.418
(0.167)
R20.940 0.932 0.940
DW 1.938 1.974 1.938
Time 2003Q–2011Q4 20 03Q1–2011Q4 2003Q1–2011Q4
Observations 36 36 36
Notes:
OLS estimation. The dependent variable is quarterly change ΔRCONS. The column numbers corre-
spond to those in Table 2. A constant and quarterly dummies are included in all estimations, but the
results are not reported. Standard errors are shown in brackets.