• No results found

The relationship between the homeownership rate and volatility

N/A
N/A
Protected

Academic year: 2021

Share "The relationship between the homeownership rate and volatility"

Copied!
51
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

           

Faculty  of  Economics  and  Business  

MSc  Business  Economics:  Real  Estate  Finance  &  Finance  

 

 

 

 

 

 

The  Relationship  between  the  Homeownership  

Rate  and  Volatility    

 

 

July  2014  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Master  Thesis     Pepijn  Holst  (10672753)   Supervisor:  Marc  Francke   Second  Reader:    

(2)

 

 

 

 

This  paper  explores  the  relationship  between  the  homeownership  rate  and  volatility  by   examining  panel  data  on  five  OECD  countries  between  1970  and  2012.  To  determine  the   sign  and  causality  of  the  relationship  a  VAR  analysis  is  performed.  Next,  an  explanation   for  this  phenomenon   is   sought   by   exploring   the   influence   of   housing   policy  factors   on   volatility   in   an   OLS   regression.   The   VAR   results   uncover   the   presence   of   an   inflexion   point   in   the   relationship,   where   the   sign   of   the   causal   relationship   changes   from   negative   to   positive,   and   the   strength   of   the   causality   is   minimised.   This   occurs   at   a   homeownership   rate   of   approximately   50%.   Here,   the   housing   market   volatility   is   minimised.   Furthermore,   the   OLS   regression   results   demonstrate   that   the   effect   of   housing   policies   is   smallest   at   the   inflexion   point,   and   is   an   increasing   function   of   the   distance  from  the  50%  rate.    

           

(3)

Table  of  Contents  

1.  Introduction  ...  4  

1.1  Focus  Area  ...  4  

1.2  First  Hypothesis  ...  6  

1.3  Second  Hypothesis  ...  6  

1.4  Validity  of  Research  ...  7  

1.5  Methodological  Approach  ...  8  

1.6  Structure  of  Paper  ...  8  

2.  Literature  Discussion  ...  9  

2.1  Volatility  Literature  ...  9  

2.2  Tenure  Literature  ...  12  

2.3  Volatility  and  Tenure  Literature  ...  17  

2.4  Key  Articles  ...  18  

3.  Methodology  ...  19  

3.1  ARMA  and  GARCH  Model  ...  19  

3.2  VAR  Model  ...  20   3.3  OLS  Model  ...  21   4.  Data  ...  23   4.1  Data  Collection  ...  23   4.2  Data  Transformation  ...  24   4.3  Descriptive  Statistics  ...  25   5.  Results  ...  28   5.1  Volatility  Creation  ...  28   5.2  Causality  Results  ...  29  

5.3  Housing  Policy  Results  ...  32  

5.4  Model  Verification  ...  34  

5.4.1  VAR  Verification  ...  34  

5.4.2  OLS  Verification  ...  36  

6.  Robustness  Checks  ...  37  

6.1  Volatility  Series  Robustness  ...  37  

6.2  VAR  and  OLS  Robustness  ...  38  

7.  Conclusion  ...  40  

7.1  Hypotheses  ...  40  

7.2  Implications  of  Findings  ...  40  

7.3  Limitations  and  Further  Research  ...  41  

8.  List  of  References  ...  43  

9.  Appendix  ...  47    

(4)

1.  Introduction    

The  majority  of  developed  nations  promote  homeownership.  This  is  because  it  permits   wealth   accumulation,   in   addition   to   increasing   community   engagement   and   enabling   children  to  perform  better  at  school  (Andrews & Sanchez, 2011).  In  the  United  States,  the   government   has   pushed   lenders   to   extend   mortgages   to   low-­‐income   households (Wallison, 2010).   In   the   Netherlands,   the   extensive   tax   deductibility   of   mortgage   payments  has  produced  the  highest  wedge  between  the  prevailing  interest  rate  and  the   after-­‐tax  cost  of  financing  of  any  OECD  country  (Andrews & Sanchez, 2011).  

 

The  promotion  of  homeownership  does  not  come  without  costs.  The  prime  example  of   this  was  the  Financial  Crisis,  which  resulted  from  the  inability  of  American  households   to  service  their  mortgages.  An  “unsophisticated  population  of  new  homebuyers”  (Gelain, Lansing, & Mendicino, 2013)   had   been   issued   mortgages   based   only   on   the   expected   appreciation  of  the  value  of  their  house.  When  house  prices  stopped  rising,  the  ensuing   mortgage   defaults   sent   shockwaves   through   the   financial   system.   Subsequently,   the   Case-­‐Shiller  Home  Price  Index  tumbled  30%  between  the  fall  of  2007  and  early  2009.     Switzerland  on  the  other  hand  did  not  follow  suit  in  promoting  homeownership;  it  kept   lending   standards   tight   and   continued   providing   substantial   rent   subsidies   (Bourassa, Hoesli, & Scognamiglio, 2009).  The  crash  of  the  Swiss  housing  market  was  inexistent.     1.1  Focus  Area  

The  relationship  between  the  promotion  of  homeownership  and  the  risk  present  in  the   housing   market   forms   the   topic   of   this   paper.   By   comparing   the   approach   to   housing   adopted  by  the  Swiss  government  to  other  OECD  economies,  the  findings  of  this  paper   provide  food  for  thought  for  macroeconomic  planning.    

 

The   idea   is   that   the   relationship   between   homeownership   rates   (HOR)   and   volatility   (VLTY)  does  not  follow  a  linear  pattern,  but  presents  an  inflexion  point.  When  around   50%   of   households   are   homeowners,   the   causality   between   the   homeownership   rate   and  volatility  switches  from  negative  to  positive,  and  the  trade-­‐off  between  the  benefits   of  homeownership  and  volatility  is  optimised.  This  relationship  is  displayed  in  Graph  I.      

(5)

Graph  I:  the  relationship  between  ownership  rates  and  volatility  

   

To   the   left   of   the   inflexion   point   the   benefits   of   increasing   the   homeownership   rate   outweigh   the   cost   of   volatility   due   to   the   strength   of   the   ‘liquidity   effect’.   The   latter   occurs   when   an   increase   in   the   amount   of   people   searching   for   a   house   dampens   volatility.   As   more   people   trade   in   the   housing   market,   so   demand   will   be   better   satisfied.  For  a  seller,  the  probability  of  finding  a  buyer  has  increased,  which  means  the   asking  price  is  less  likely  to  be  altered.  This  effect  is  widely  reported  in  stock  markets.        

To   the   right   of   the   inflexion   point   the   cost   of   volatility   weighs   more   heavily   than   the   benefits   of   homeownership.   This   is   due   to   the   ‘default   effect’   increasing   in   strength.   When  the  homeownership  rate  is  high,  the  promotion  of  owning  will  start  affecting  risky   buyers,  those  without  the  normal  means  to  own  a  house  but  obtain  the  means  through   the   lowering   of   borrowing   requirements   and   favourable   housing   policies.   As   the   concentration  of  risky  buyers  increases,  so  does  the  probability  of  defaults.  This  induces   volatility,  as  the  supply  of  houses  will  experience  increased  fluctuation.    

Furthermore,  as  Belsky,  Retsinas  and  Duda  (2005)  found,  low-­‐income  households  (risky   buyers)   often   receive   subprime   loans,   to   compensate   investors   on   the   secondary   mortgage  market  for  their  increased  risk.  This  means  the  interest  rate  is  higher,  and  the  

30%   40%   50%   60%   70%   80%  

Volatility

 

(6)

debt  service  a  bigger  burden,  certainly  when   compared  to  the  level  of  income.  This  in   turn  makes  defaults  even  more  likely  than  in  the  case  of  non-­‐subprime  loans.    

In  addition,  the  authors  found  housing  expenditures  to  be  positively  related  to  income,   meaning  that  low-­‐income  households  spend  less  on  maintenance.  As  a  result,  not  only   does   the   supply   of   housing   fluctuate   more,   but   it   also   becomes   more   heterogeneous.   Together  these  effects  will  induce  volatility  in  the  housing  market.    

1.2  First  Hypothesis  

The  first  hypothesis  of  this  paper  states  that  countries  with  a  low  proportional  amount   of   homeowners   should   exhibit   negative   time-­‐causality   between   ownership   rates   and   volatility.   Those   with   a   high   rate   should   demonstrate   positive   time-­‐causality.   For   simplicity’s  sake,  the  inflexion  point  is  hypothesised  to  lie  at  a  rate  of  50%.  

 

!!: ↑ !"#!"#$%  !"% →  ↓ !"#$, ↑ !"#!"#$%  !"% →  ↑ !"#$    

!!: ↑ !"#!"#$%  !"% ↛  ↓ !"#$, ↑ !"#!"#$%  !"% ↛  ↑ !"#$  

 

The  countries  used  in  this  paper  were  chosen  because  their  homeownership  rates  differ   widely.  Switzerland  and  the  Netherlands  have  experienced  relatively  low  rates  (below   50%)  for  the  majority  of  the  chosen  time  period,  whereas  the  United  States,  Denmark,   and   the   United   Kingdom   present   higher   rates.   Due   to   limitations   in   the   availability   of   data,  the  homeownership  rate  used  for  each  country  is  the  average  homeownership  rate   over   the   chosen   time   period.   This   rate   is   used   for   comparative   purposes   only,   as   the   analysis  itself  is  based  on  a  series  of  homeownership  rates  and  volatility.    

1.3  Second  Hypothesis  

The  idea  of  an  inflexion  point  can  be  further  refined.  Notably,  within  the  OECD  countries   studied,   demographic   characteristics   drive   a   ‘natural’   demand   for   homeownership,   which   is   hypothesised   to   coincide   with   the   inflexion   point   in   each   country.   It   is   the   influence   of   housing   policies   that   shifts   the   rate,   by   either   promoting   or   restricting   access  to  funding  or  housing.  The  movement  away  from  the  inflexion  point,  through  the   creation  of  artificial  demand,  is  what  drives  volatility.    

   

(7)

The   second   hypothesis   of   this   paper   posits   that   the   promotion   or   restriction   of   homeownership  causes  volatility  in  the  housing  market.  In  fact,  the  further  away  from   the   inflexion   point   the   homeownership   rate   needs   to   be   held,   the   more   volatility   is   caused.  In  practical  terms,  this  means  that  volatility  caused  by  an  increase  in  the  debt-­‐ to-­‐income   ratio,   or   the   expenditure   of   the   government,   is   larger   when   it   drives   the   homeownership   rate   from   60   to   61%   than   when   it   drives   the   rate   from   50   to   51%.   Similarly,  more  volatility  is  caused  when  the  rate  is  reduced  from  40  to  39%  than  from   50  to  49%.  The  cost  (in  terms  of  volatility)  of  increasing  the  ownership  rate  increases   with  the  distance  from  the  inflexion  point.  

 

!!:   !!"!"#$!%!  !"# > !!"!"#$!%#  !"#   !!:     !!"!"#$!%!  !"# ≤ !!"!"#$!%#  !"#  

 

Here,  !"#  is   the   coefficient   on   the   housing   policy   measure,   where   volatility   is   the   dependent  variable.  The  housing  policy  measure  in  question  could  be  either  the  debt-­‐to-­‐ income   ratio   or   government   expenditure   as   a   percentage   of   GDP.   The   sub-­‐notation   of   ‘extreme’   or   ‘average’   refers   to   the   distance   to   the   inflexion   point   of   the   average   ownership  rate,  where  extreme  is  distant  and  average  is  close  to  the  hypothesised  point.     1.4  Validity  of  Research  

Studying   the   causal   relationships   between   volatility,   homeownership,   and   housing   policies   is   important   because   housing   forms   the   largest   part   of   an   individual’s   wealth   portfolio (Blanchflower & Oswald, 2013).   Furthermore,   in   the   aftermath   of   the   Crisis   governments  have  started  rethinking  their  housing  policies.  As  such,  maximum  loan-­‐to-­‐ value  ratios  have  been  lowered  in  most  countries,  and  the  tax  deductibility  of  mortgage   payments  will  slowly  disappear  in  the  Netherlands  (ABN Amro, 2012).  These  are  but  a   few   of   the   developments   currently   taking   place.   This   paper   adds   to   the   debate   about   housing  policies  by  shining  light  on  the  cost  of  homeownership.  By  comparing  different   countries,  it  is  able  to  demonstrate  what  rate  minimises  risk.    

Furthermore,   volatility   in   the   housing   market   is   also   known   to   influence   volatility   in   consumption   (Gelain, Lansing, & Mendicino, 2013),   which   is   itself   a   prime   driver   of   economic   success.   Ultimately,   aiding   citizens   in   accumulating   wealth   is   encouraging  

(8)

them  to  consume.  As  such,  a  stable  housing  market  prompts  stable  consumption,  and  is   thus  an  important  consideration  for  any  modern  government.    

1.5  Methodological  Approach  

The   methodological   approach   adopted   in   this   paper   is   based   on   VAR   and   OLS   regressions.   The   former   is   used   to   uncover   the   presence   of   causality   between   the   homeownership   rate   and   volatility,   whereas   the   latter   shows   which   determinants   of   homeownership  are  important  in  explaining  volatility.    

A   contrast   is   made   across   five   OECD   economies,   namely   the   Netherlands,   the   United   States,  Switzerland,  Denmark,  and  the  United  Kingdom.  These  were  chosen  not  only  for   their   widely   different   ratios   of   homeownership   to   volatility,   but   also   because   the   necessary   data   for   each   country   is   publicly   available.   The   analysis   is   purposefully   focused   on   developed   economies,   as   there   are   a   host   of   other   factors   influencing   homeownership   rates   in   developing   economies,   beyond   demographic   and   housing   policy  factors.  These  make  it  hard  to  give  a  causal  interpretation  to  variables.    

The   volatility   that   is   analysed   in   this   paper   is   based   on   the   appreciation   of   average   national  house  prices.  It  is  constructed  by  inserting  the  residuals  from  an  ARMA  into  a   GARCH  model.  Clarifications  are  provided  in  the  methodology  section.    

The   time   period   analysed   covers   1970-­‐2012   in   the   VAR   model,   and   1990-­‐2012   in   the   OLS  model.  This  difference  is  due  to  the  dearth  of  demographic  variables.  

1.6  Structure  of  Paper  

This   paper   is   divided   into   seven   sections.   First,   the   existing   literature   discussing   volatility  and  tenure  decisions  is  analysed,  and  the  added  value  of  this  paper  is  clarified.   The  third  section  explains  the  methodological  approach  adopted  in  this  paper.  Next,  the   data   used   for   the   analysis   is   examined.   This   involves   a   discussion   of   sources   and   descriptive   statistics.   The   fifth   section   is   a   discussion   of   the   results,   while   the   sixth   section  provides  an  attempt  to  limit  the  downfalls  of  the  methodology  by  testing  for  the   robustness  of  results.  Finally,  section  seven  contains  the  conclusion.  

       

(9)

2.  Literature  Discussion    

In   light   of   the   Financial   Crisis   of   2007-­‐2009,   the   risk   of   promoting   homeownership   should   be   clear.   However,   though   the   determinants   of   volatility   and   homeownership   have   been   studied   in   detail,   there   has   never   been   an   attempt   to   empirically   uncover   what  the  relationship  between  both  variables  is,  and  whether  it  differs  across  countries.   As   this   paper   attempts   to   link   the   development   of   homeownership   rates   to   volatility   levels,  the  relevant  literature  covers  the  determinants  of  volatility  and  tenure  decisions.   Until  recently,  these  subjects  were  covered  separately.  However,  there  is  a  nascent  focus   on   the   direct   link   between   volatility   and   tenure   decisions,   which   constitutes   the   final   section  of  this  literature  discussion.    

2.1  Volatility  Literature    

There   is   an   extensive   literature   on   the   determinants   of   volatility   in   the   United   States.   Usually  consisting  of  equating  changes  in  macroeconomic  variables  to  specific  volatility   events,   current   findings   agree   on   the   majority   of   relevant   determinants.   Nonetheless   there   remain   fundamental   questions   regarding   the   volatility   of   house   prices.   Notably   missing  is  an  explanation  for  cross-­‐sectional  differences  in  long-­‐term  volatility  levels,  be   it  between  regions  or  countries.  The  United  States,  for  example,  has  a  housing  market   that   consistently   displays   a   higher   level   of   volatility   than   its   European   counterparts.   Homeownership  rates  could  go  some  way  in  explaining  this  phenomenon.    

 

The   literature   on   housing   market   volatility   has   uncovered   several   macroeconomic   determinants   that   are   relevant   to   isolate   the   relationship   between   aggregate   homeownership   rates   and   volatility   levels.   Dolde   and   Tirtiroglue   (2002)   find   personal   income   growth   to   be   the   most   significant   determinant,   in   addition   to   inflation   and   interest   rates.   Furthermore,   the   literature   finds   the   rate   of   house   price   appreciation   (Hossain   &   Latif,   2009)   (Miller   &   Peng,   2006),   the   population   growth   rate   (Miller   &   Peng,  2006),  and  the  change  in  the  prevailing  mortgage  rate  (Hossain  &  Latif,  2009)  to   explain  variations  in  housing  market  volatility.    

 

Dolde   and   Tirtiroglue   (2002)   use   repeat   transaction   data   between   1975   and   1993   for   four  regions  in  the  United  States.  They  identify  36  volatility  events,  defined  as  a  squared  

(10)

deviation   significantly   different   from   the   sample   mean.   The   average   level   of   volatility   observed   understates   the   one   used   in   this   paper   by   1%,   which   is   consistent   with   the   upward  trend  of  both  volatility  and  homeownership  in  the  United  States.    

 

Focusing  on  the  United  States  again,  Miller  and  Peng  (2006)  use  median  sales  price  data   from   1978   to   2002   for   316   metropolises.   They   control   for   both   metropolitan-­‐specific   and   time-­‐varying   factors   in   uncovering   volatility   determinants.   The   former   is   done   by   creating  subsamples  by  metropolitan  size  and  estimating  a  new  VAR  system  using  only   the  most  and  least  populated  metropolises.  The  latter  is  achieved  by  diving  the  sample   period  in  two.  Though  the  dynamic  interrelations  between  variables  are  present  in  both   small  and  large  metropolises,  their  effects  vary.  Volatility  reduces  population  growth  in   small  metropolitan  areas,  but  not  large  ones.  Furthermore,  volatility  has  a  smaller  effect   on  the  variables  in  less  populated  areas.  In  terms  of  time-­‐varying  factors,  the  dynamic   interrelations  between  variables  are  similar  in  both  period  subsamples,  meaning  there   are  no  regime  shifts  between  1978  and  2002.  As  a  result,  no  attempt  is  made  to  control   for  time-­‐varying  factors  in  this  paper.      

 

Others   authors   have   looked   beyond   macroeconomic   factors   for   determinants   of   volatility.   Their   results   are   not   relevant   to   isolate   the   effect   of   homeownership   on   volatility,  but  hint  at  the  relevance  of  this  relationship.    

Stein  (1995)  and  Gelain,  Lansing  and  Mendicino  (2013)  have  looked  at  the  link  between   financing   policies   and   the   volatility   of   house   prices.   Stein   analyses   a   theoretical   price-­‐ demand   equilibrium   model,   and   finds   there   are   multiple   equilibriums   present   when   down  payment  constraints  are  introduced.  This  is  due  to  the  fact  that  when  prices  rise   beyond  a  certain  point  they  relax  liquidity  constraints  for  homeowners  saddled  with  a   mortgage,   while   still   having   the   ‘normal’   effect   of   decreasing   demand   on   prospective   buyers.   A   situation   of   multiple   equilibriums   causes   prices   to   react   more   violently   to   changes   in   macroeconomic   determinants.   The   violence   of   this   movement   depends   on   the   amount   of   active   constrained   movers   relative   to   active   unconstrained   ones,   which   increases   as   prices   rise.  His   finding   suggests   that   countries   with   an   active   role   for  the   government  in  promoting  or  restricting  housing,  experience  more  volatility.  He  hereby   provides  ammunition  for  a  comparison  of  policies  and  volatility  levels  across  countries.      

(11)

Gelain,  Lansing  and  Mendicino  (2013)  analyse  the  policy  actions  that  are  most  effective   in  diminishing  excess  house  price  volatility,  which  is  defined  as  “volatility  that  cannot  be   explained   by   a   rational   response   to   fundamentals.”   They   find   that   a   debt-­‐to-­‐income   constraint  is  most  effective  in  this  regard,  as  it  does  not  increase  the  volatility  of  other   macroeconomic  series.  Next,  a  lower  loan-­‐to-­‐value  ratio  is  also  effective  in  dampening   house   price   volatility,   but   at   the   cost   of   a   small   increase   in   inflation   volatility.   Furthermore,  the  authors  find  that  a  debt-­‐to-­‐income  ratio  is  subject  to  less  speculative   distortions  than  a  loan-­‐to-­‐value  ratio,  because  the  latter  enables  the  issuance  of  debt  to   follow   movements   in   house   prices.   As   a   result,   this   paper   makes   use   of   the   debt-­‐to-­‐ income   ratio   as   a   proxy   for   housing   policies,   as   it   gives   the   best   approximation   of   the   effect  of  policies  on  the  mean  household.      

 

Reichart  (1990),  in  his  study  of  the  US  housing  market,  finds  that  similar  factors  affect   volatility   differently   in   different   regions;   mortgage   rates   are   most   important   in   New   England,  whereas  permanent  income  has  the  largest  effect  in  the  western  regions  of  the   United   States.   This   means   that   homeownership   rates   could   also   affect   volatility   differently   depending   on   the   region   being   studied,   not   to   mention   the   country.   The   author   thus   clears   the   way   for   a   comparison   of   the   effect   of   changes   in   the   homeownership  rate  on  volatility  in  different  countries.    

 

The  relevance  of  the  articles  by  Hossain  and  Latif  (2009)  and  Miller  and  Peng  (2006)  also   lies  in  their  use  of  rational  expectations  models  for  analysing  the  appreciation  of  home   values.  The  residuals  from  such  a  model  constitute  the  unpredictable  component  that  is   used  to  model  volatility.  In  addition,  because  the  volatility  series  created  with  this  model   is  used  for  comparative  purposes,  the  potential  underestimation  of  real  volatility  is  not   problematic  (Gelain, Lansing, & Mendicino, 2013).    

In   Hossain   and   Latif   (2009),   the   rational   expectations   model   is   constructed   using   an   ARMA   and   GARCH   technique.   Recreating   this   series   for   the   relevant   countries   enables   the   development   of   volatility   to   be   analysed   alongside   the   homeownership   rate   and   other  macroeconomic  variables  in  a  VAR  model.  Hereby  the  authors  provide  a  base  from   which  to  commence  the  methodological  study  of  this  paper.    

(12)

Flood  and  Hodrick  (1986)  look  at  the  ability  of  variance  bound  tests  to  model  volatility   during   rational   speculative   bubbles.   They   argue   against   the   common   view   that   these   rational  models  are  unable  to  process  a  bubble-­‐like  movement  in  a  time  series,  due  to   the  latter  appearing  as  irrational.  The  authors  find  that  the  design  of  a  rational  model,   like   the   one   used   in   this   paper,   “precludes   bubbles   as   a   reason   for   failure.”   In   other   words,  a  rational  model  can  handle  the  presence  of  a  bubble  movement  in  a  time  series   of  house  prices,  which  is  what  this  paper  attempts  to  do.  However,  they  caution  that  a   variance   bound   model   cannot   be   used   to   identify   the   presence   of   a   bubble,   as   the   expectations   of   agents   would   require   remodelling   (rationality   evaporates   during   a   speculative  bubble).    

Furthermore,   Brailsford   and   Faff   (1996)   analyse   the   applicability   of   ARCH   models   to   make   volatility   forecasts.   By   employing   forecasting   techniques   ranging   from   simple   (random   walk   model)   to   complex   (ARCH   family   models)   on   daily   Australian   stock   market  data,  they  are  able  to  rank  their  accuracy.  Their  findings  suggest  that  the  GARCH   (1;1)  model  ranks  second  to  the  Glosten,  Jagannathan,  and  Runkle  modified  GARCH  in   terms  of  forecasting  performance,  which  is  what  this  paper  attempts  to  use  the  GARCH   model  for.  Both  models  predict  around  60%  of  the  actual  observed  volatility  (all  models   under-­‐predict  observed  volatility).    

Flood   and   Hodrick   (1986)   and   Brailsford   and   Faff   (1996)   thus   confirm   the   use   of   an   ARMA  and  GARCH  combination  to  process  the  bubble  in  house  prices  that  was  present   in  this  paper’s  chosen  time  period,  and  subsequently  forecast  volatility  series.    

2.2  Tenure  Literature    

The  literature  on  tenure  decisions  also  has  its  roots  in  the  United  States.  However,  the   range   of   countries   studied   in   these   articles   is   far   wider   than   is   the   case   with   the   determinants  of  volatility.  Furthermore,  the  determinants  of  homeownership  have  been   studied  for  different  demographic  tranches.    

The  problem  with  the  existing  literature  is  that  is  uses  almost  solely  cross-­‐sectional  data   sets,  based  on  survey  results.  As  such,  even  though  mortgage  policies  have  been  studied   extensively,   there   is   no   coverage   of   the   effect   of   gradual   loosening   of   financing   standards,  for  example,  in  the  run-­‐up  to  the  Crisis  on  tenure  decisions.  In  addition,  there   is  no  possibility  to  test  for  non-­‐linear  relationships,  as  the  use  of  two  surveys  forms  the  

(13)

maximum  number  of  time  snapshots  (see (Andrews & Sanchez, 2011) (Chambers, Garriga, & Schlagenhauf, 2009) (Linneman & Wachter, 1989) (Gabriel & Rosenthal, 2005)).    

 

Starting   off   at   the   cross-­‐country   level,   Fisher   and   Jaffe   (2003)   attempt   to   explain   what   factors  determine  the  homeownership  rate  across  a  sample  of  106  countries.    They  find   that   the   rate   of   urbanisation   and   government   consumption   have   significant   negative   effects  on  the  homeownership  rate,  while  the  proportion  of  people  between  the  ages  of   15  and  64  and  the  presence  of  a  mandatory  financing  system  have  a  significant  positive   impact.  The  problem  with  the  cross-­‐panel  dataset  that  the  authors  use  is  that  it  cannot   explain   how   the   impact   of   their   determinants   varies   across   countries,   only   what   the   average   effect   is.   In   particular   it   cannot   explain   differences   in   the   effect   of   varying   mandatory  finance  schemes.      

 

Next,   Andrews   and   Sanchez   (2011)   focus   on   a   range   of   OECD   countries,   excluding   ‘transitory’  economies  that  might  induce  bias  in  the  significance  of  variables.  They  posit   that  the  evolution  of  homeownership  rates  is  due  to  a  mix  of  demographic  and  public   policy  influences.  The  former  contains  factors  such  as  age  of  household  head,  household   size,   and   real   disposable   income,   amongst   others.   In   terms   of   public   policy   influences,   they   analyse   the   relaxation   of   down-­‐payment   constraints,   mortgage   interest   deductibility,   and   rent   regulations.   They   find   that   changes   in   household   characteristic   account  for  75%  of  the  increase  of  the  homeownership  rate  in  the  UK,  but  only  33%  in   Switzerland  and  the  US.  This  result  agrees  with  the  second  hypothesis  of  this  paper,  as   Switzerland   and   the   US   are   both   expected   to   depend   more   on   housing   policies   than   countries  closer  to  the  inflexion  point.  Furthermore,  population  ageing  has  had  a  large   effect  in  both  Switzerland  and  Denmark.  In  the  latter,  changes  in  income  are  also  found   to   be   significant,   unlike   other   European   countries.   In   addition,   demographic   changes   cannot   explain   changes   in   their   entirety,   leaving   a   substantial   role   for   public   policy   measures   in   numerous   OECD   economies.   However,   the   authors   are   unable   to   demonstrate   where   the   impact   of   housing   policies   has   had   the   biggest   impact   on   the   homeownership  rate.    

 

At  the  single  country  level,  Chambers,  Garriga  and  Schlagenhauf  (2009)  employ  a  similar   hypothesis.   They   relate   the   boom   in   the   US   homeownership   rate   between   1994   and  

(14)

2005   to   development   in   demographics   and   mortgage   innovations.   They   do   this   by   estimating   the   contribution   of   each   factor   in   1994,   and   projecting   this   contribution   forward  to  2005  while  holding  other  factors  constant.  This  enables  them  to  calculate  a   hypothetical  homeownership  rate,  and  subsequently  compare  this  to  the  actual  rate  to   determine  the  influence  of  each  variable.  They  find  that  the  change  in  demographics  is   responsible  for  16  to  31%  of  the  increase  in  the  aggregate  homeownership  rate,  while   mortgage   market   innovations   account   for   between   56   and   70%,   depending   on   the   factors  held  constant.  The  former  is  especially  important  in  explaining  the  development   of  the  homeownership  rate  for  older  cohorts,  whereas  new  mortgage  products  mainly   help  younger  households  acquire  a  home.  Thus  it  would  seem  that  the  United  States  has   relied   largely   on   financing   innovations   to   drive   up   its   ownership   rate;   a   result   agrees   with  Andrews  and  Sanchez  (2011)  and  again  confirms  this  paper’s  second  hypothesis.    

Bourassa   and   Hoesli   (2006)   attempt   to   explain   the   low   ownership   rate   in   Switzerland.   They  analyse  3588  households  in  1998,  and  model  the  tenure  decision  as  a  function  of   the   cost   of   owning   and   renting,   the   borrowing   constraint   gap,   household   after-­‐tax   income,  and  a  host  of  demographic  characteristics.  The  first  variable  takes  into  account   the  extensive  rent  subsidies  provided  by  the  government,  the  high  level  of  rent  security,   the  taxation  of  imputed  rent,  high  house  prices,  and  the  stringent  underwriting  criteria.   The  outcome  points  to  the  importance  of  high  house  prices  relative  to  incomes,  which  is   only  indirectly  a  result  of  national  mortgage  policies.  The  most  important  explanatory   variable  however  is  the  set  of  demographic  variables,  such  as  the  age  of  the  household   head,  his  marital  status,  and  the  presence  of  children.  As  such,  their  study  disagrees  with   later   findings   (Andrews & Sanchez, 2011)   about   the   importance   of   non-­‐demographic   factors   in   Switzerland.   It   is   important   however   to   consider   the   different   timeframes   used  by  both  studies.  Whereas  Bourassa  and  Hoesli  (2006)  look  at  households  in  1998,   Andrews   and   Sanchez   (2011)   analyse   a   development   between   1994   and   2004.   It   thus   appears  the  importance  of  mortgage  market  innovations  has  increased  over  the  years  in   Switzerland;  an  observation  going  against  the  second  hypothesis  posited  in  this  paper.      

In   terms   of   regional   determinants,   Eilbott   and   Binkowski   (1985)   analyse   the   tenure   decision   using   data   on   a   large   number   of   metropolitan   areas   covered   by   the   1970   US   Census.   Though   testing   for   both   supply-­‐   and   demand-­‐side   variables,   the   former   are  

(15)

found   irrelevant   in   explaining   differences   in   homeownership   rates,   and   are   thus   dropped  from  the  model.  The  authors  hereby  provide  ammunition  for  the  sole  focus  on   demand-­‐side  variables  in  this  paper.    

House   values   are   found   to   be   most   significant,   decreasing   the   ownership   rate   as   they   rise.  In  addition,  income  and  household  size  are  found  to  have  a  positive  relationship,   whereas   population   growth   and   the   percentage   of   people   under   35   years   of   age   are   found  to  bear  a  negative  relationship.  On  the  other  hand,  the  level  of  rent  is  consistently   found  to  be  insignificant  in  explaining  high  homeownership  rates  in  the  US.  This  finding   agrees   with   the   results   of   this   paper   about   the   decreasing   importance   of   government   expenditure  (proxy  for  sponsoring  of  the  rental  market)  when  the  ownership  rate  rises.      

Though  touched  on  previously,  there  are  also  a  host  of  articles  that  specifically  study  the   influence   of   public   policy   measures   on   tenancy   choices.   Malpezzi   (1996)   looks   at   the   effect  of  housing  regulation  on  house  prices  and  homeownership  rates.  He  constructs  a   variable   measuring   the   amount   of   regulation   in   56   US   cities   by   adding   together   seven   measures  reported  by  the  Wharton  regulatory  practices  data.  These  contain  measures   for   zoning   practices,   approval   times   and   the   quality   of   infrastructure.   Though   the   coefficient  on  the  regulation  variable  is  not  statistically  significant,  it  raises  house  values   and  rents,  but  the  former  by  more  than  the  latter.  Combined  these  effects  will  thus  have   a   negative   impact   on   the   homeownership   rate,   namely   a   reduction   of   10   percentage   points  when  switching  from  a  lightly  regulated  environment  to  a  heavily  regulated  one.   This  result  helps  explain  the  low  homeownership  rate  in  Switzerland.      

 

Next,  Linneman  and  Wachter  (1989)  study  the  effect  of  mortgage  underwriting  criteria   on  homeownership  rates.  Using  income,  the  relative  cost  of  ownership  versus  renting,   and   borrowing   constraints   on   household   data   covering   1975-­‐77   and   1981-­‐83,   the   authors   test   whether   income   and   wealth   constraints   reduce   the   probability   of   homeownership.  In  the  first  period,  income-­‐constrained  families  were  32%  less  likely  to   be  homeowners,  whereas  for  wealth-­‐constrained  families  this  was  61%.    

These  figures  are  smaller  for  the  second  period,  due  in  part  to  lower  interest  rates.  Thus,   income-­‐constrained  and  wealth-­‐constrained  families  were  19  and  21%  less  likely  to  be   homeowners,   respectively.   The   reduced   importance   of   the   income   constraint   in   the  

(16)

second  period  is  likely  to  be  a  result  of  the  widespread  use  of  adjustable-­‐rate  mortgages,   which  had  a  lower  rate  compared  to  fixed-­‐rate  instruments.    

Their  results,  namely  a  decrease  in  the  probability  for  risky  buyers  to  be  less  likely  to  be   homeownership,  points  not  only  to  a  slump  in  interest  rates,  but  also  the  relaxation  of   borrowing  constraints  in  the  United  States.    

 

There   is   also   a   range   of   articles   that   studies   the   determinants   of   homeownership   for   specific   demographics.   Chiuri   and   Jappelli   (2003)   look   at   the   distribution   of   homeownership   rates   across   age   groups   within   14   OECD   countries.   They   hypothesise   that  countries  with  tight  lending  standards  have  lower  occupancy  rates  amongst  young   individuals,  because  they  lack  collateral.  The  authors  use  the  down  payment  ratio  as  a   measure  of  finance  standards.  They  find  that  an  increase  in  the  down  payment  ratio  by   20%  lowers  the  owner-­‐occupancy  rate  by  the  same  amount  for  people  aged  26-­‐35,  while   it   lowers   it   by   15%   for   those   aged   36-­‐45.   This   result   is   maintained   even   when   controlling  for  income.  Though  they  confirm  their  hypothesis,  their  results  are  based  on   a  pair  of  surveys  from  different  time  periods  for  each  country.  Furthermore  they  do  not   control  for  other  factors  that  affect  the  homeownership  decision,  such  as  tax  incentives,   that  might  change  over  time.    

 

Gabriel   and   Rosenthal   (2005)   analyse   whether   the   Clinton   and   Bush   administrations   have   been   able   to   narrow   the   racial   gap   in   owner-­‐occupancy   rates.   Drawing   on   data   from   the   Consumer   Finance   Survey   between   1983   and   2001,   the   authors   decompose   racial   homeownership   rate   gaps   into   a   portion   that   can   be   explained   by   household   demographics   (other   than   race)   and   one   capturing   the   impact   of   credit   barriers.   They   find   that   the   increase   in   homeownership   is   primarily   a   result   of   changing   household   demographics,  which  was  responsible  for  14  out  of  the  26-­‐percentage  point  white-­‐black   gap   and   20   out   of   the   30-­‐percentage   point   white-­‐Hispanic   gap.   Furthermore   credit   barriers  consistently  account  for  5  percentage  points  of  the  gap  for  all  years,  meaning   mortgage   market   policies   designed   to   enhance   homeownership   amongst   minorities   have  had  limited  success.    

(17)

2.3  Volatility  and  Tenure  Literature    

The  focus  on  homeownership  decisions  and  volatility  is  interesting  because  there  are  a   host   of   similar   factors   that   influence   both   variables,   as   the   previous   two   sub-­‐sections   have   demonstrated.   Most   notably,   both   are   the   result   of   a   mix   of   demographic   and   housing  policy  factors.  This  would  suggest  that  movements  in  homeownership  rates  and   volatility   levels   are   not   independent.   As   a   result,   recent   literature   has   focused   on   uncovering  whether  a  relationship  exists,  and  whether  it  is  causal.    

 

First,  Ortalo-­‐Magné  and  Rady  (2002)  analyse  the  relationship  between  homeownership   and   housing   market   volatility,   and   find   evidence   of   a   positive   relationship.   Enabling   homeownership   in   a   first   period   increases   the   variance   of   house   prices   in   the   second   period,  due  to  the  decrease  in  household  mobility.  However  their  analysis  is  constrained   to  two  periods  within  the  same  city.  As  such  they  ignore  cross-­‐country  differences  and   factors   that   influence   volatility   in   the   long   run.   Furthermore   their   analysis   holds   only   theoretical  value  as  it  fails  to  account  for  absolute  changes  in  the  ownership  rate  and  the   subsequent   impact   on   mobility   and   volatility.   Nonetheless   they   uncover   causality   between  both  variables,  flowing  from  homeownership  to  volatility,  which  supports  the   first  hypothesis  made  in  this  paper.    

 

On  the  other  hand,  Banks  et  al.  (2004)  attempt  to  make  a  comparison  across  the  United   States   and   the   United   Kingdom.   They   discuss   the   role   volatility   in   the   housing   market   has  on  the  decision  to  own.  They  find  that  the  higher  the  level  of  housing  uncertainty,   the  higher  the  demand  for  housing,  as  risk-­‐averse  individuals  seek  to  insure  themselves   from   the   risk   of   house   price   rises.   However,   volatility   is   analysed   as   a   given,   and   they   make  no  attempt  to  model  the  reversed  causal  relationship.    

 

As   a   result,   the   literature   confirms   the   existence   of   a   relationship   between   tenure   decisions  and  volatility.  The  causality  is  predicted  to  flow  in  both  directions,  but  only  the   effect  of  changes  in  the  homeownership  rate  on  volatility  will  be  analysed  in  this  paper,   leaving  the  reverse  causality  to  be  developed  in  future  research.    

(18)

2.4  Key  Articles  

The   three   key   papers   discussed   in   this   literature   review   are   Hossain   and   Latif   (2009),   Andrews  and  Sanchez  (2011),  and  Ortalo-­‐Magné  and  Rady  (2002).  The  former  provides   the   control   variables   and   models   that   are   used   to   analyse   the   causal   relationship   between  homeownership  rates  and  volatility,  and  judge  the  first  hypothesis.  The  second   OECD  article  introduces  the  idea  of  equating  homeownership  rates  to  demographic  and   policy   factors,   and   hereby   provides   a   base   from   which   to   test   the   second   hypothesis.   Finally,   the   third   article’s   relevance   lies   in   testing   precisely   the   relationship   between   ownership   rates   and   volatility,   and   hereby   providing   a   source   of   comparison   for   the   findings  of  this  paper.      

                                             

(19)

3.  Methodology    

The   methodological   approach   of   this   paper   employs   four   different   models   in   three   separate  processes.  The  first  of  these  creates  a  volatility  series  for  each  country  based  on   the  appreciation  of  house  prices  between  1970  and  2012.  The  second  uncovers  whether   there   is   a   positive   causal   relationship   between   homeownership   rates   and   the   level   of   volatility.  Finally,  the  third  process  uses  a  set  of  determinants  from  both  the  literature   and   the   hypotheses   presented   in   this   paper   to   analyse   the   significance   of   policy   measures  in  explaining  homeownership  rates  and  volatility  levels.    

3.1  ARMA  and  GARCH  Model  

To   construct   a   volatility   series,   use   is   made   of   an   autoregressive   moving   average   (ARMA)  and  generalised  autoregressive  conditional  heteroskedasticity  (GARCH)  model.   The  data  necessary  for  this  operation  is  quarterly  house  prices  between  1970  and  2012,   from   which   appreciation   rates   are   constructed.   As   the   variable   of   interest   is   an   appreciation   series   of   house   prices,   it   is   stationary.   This   means   the   ARMA   model   is   appropriate,   as   opposed   to   an   autoregressive   integrated   moving   average   (ARIMA)   model,  which  models  a  nonstationary  series.    

 

The  ARMA  model  assumes  the  development  of  house  price  appreciation  rates  (HPA)  is   due   to   a   series   of   unobserved   shocks   as   well   as   its   own   past   values.   This   means   it   is   based   on   a   rational   development   of   rates,   based   on   the   information   available.   This   is   made   clear   in   the   equation   below.   The   irrational   component   is   the   residual   from   the   model,  and  simultaneously  the  variable  of  interest  in  constructing  a  volatility  series.      

!"#!= ! !"#! !"!#$!%$&  !"#$%&'(!$"!!! + !!    

The  Bayesian  information  criterion  (BIC)  was  used  to  determine  the  number  of  lags  to   be  inserted  in  each  ARMA  model.    

 

To  create  a  variance  series  from  house  price  appreciation  rates,  the  residuals  from  the   ARMA   model   were   inserted   into   a   GARCH   (1;1)   model.   The   latter   is   applicable   when   modelling  series  containing  non-­‐uniform  volatility  clustering (Stock & Watson, 2011).  

(20)

3.2  VAR  Model    

The   VAR   model   is   used   to   uncover   whether   a   causal   relation   exists   between   homeownership  rates  and  volatility.  It  models  every  variable  as  being  dependent  on  its   own   lags   in   addition   to   those   of   the   other   variables.   As   such,   it   permits   a   dynamic   structure  that  is  stipulated  by  the  data  itself.  Furthermore  it  requires  specification  of  the   common  lag  of  variables.  These  should  affect  each  other  inter-­‐temporally.    

The   model   employs   the   determinants   of   volatility   as   put   forward   by   the   literature,   in   addition   to   the   homeownership   rate.   The   seven   variables   inserted   into   the   model   are   volatility   (VLTY),   GDP   growth   (GDPG),   inflation   growth   (CPIG),   population   growth   (POG),   mortgage   rate   change   (MRTC),   house   price   appreciation   (HPA),   and   the   homeownership   rate   (HOR).   All   variables   have   been   collected   over   the   longest   time   period  available  to  increase  the  robustness  of  the  results,  meaning  1970  until  2012  for   the   majority   of   variables.   Only   one   lag   of   variables   is   inserted   at   a   time,   to   avoid   multicollinearity  problems.  Furthermore,  the  variables  all  need  to  be  of  the  same  order   of   integration,   meaning   either   all   stationary,   or   non-­‐stationary   but   cointegrated.   The   first   option   is   applied   in   this   model,   as   some   of   the   variables   for   each   country   are   differenced   to   achieve   stationarity.   In   most   cases,   this   means   the   change   in   homeownership  rate  (HORC)  is  inserted  into  the  model  instead  of  the  homeownership   rate  itself.  The  resulting  estimation  of  the  VAR  model  is  the  following:  

 

!!" = !!"  ×  !!"!!+ !!"  

 

Here  Yit  is  a  vector  of  HORC,  GDPG,  CPIG,  POG,  MRTC,  HPA  and  VLTY  for  each  country.  Ait  

is   a   vector   of   coefficients,   and  !!"  is   a   vector   of   the   error   term   for   each   country.   This   model  is  performed  on  individual  lags,  up  to  a  maximum  of  six.    

 

The  two  variables  of  interest  are  the  homeownership  rate  and  volatility.  The  remainder   serve   as   control   variables.   Hossain   and   Latif   (2009)   have   employed   a   similar   model.   Based  on  their  findings,  GDP  growth  is  expected  to  have  a  negative  effect  on  volatility,   similarly   to   the   growth   in   inflation,   population,   and   the   appreciation   of   house   prices.   These   variables   all   dampen   the   level   of   volatility.   On   the   contrary,   the   changes   in   mortgage  rate  and  volatility  itself  both  show  a  positive  correlation  with  volatility.    

(21)

The   causality   of   the   relationship   between   homeownership   rates   and   volatility   will   be   analysed  with  a  Granger  causality  test.  It  poses  a  statistical  hypothesis  concerning  the   causal  power  of  lagged  values  of  homeownership  rates  on  volatility.  The  null  hypothesis   can  be  rejected  only  when  each  lagged  value  is  significant  according  to  a  t-­‐test,  and  the   lagged   variables   together   add   explanatory   power   according   to   a   F-­‐test.   When   this   happens,   the   homeownership   rate   is   said   to   Granger-­‐cause   volatility.   The   first   hypothesis  predicts  this  to  be  the  case  between  homeownership  rates  and  volatility.     3.3  OLS  Model    

The   ordinary   least   squares   (OLS)   model   is   used   to   determine   the   effect   of   housing   policies  on  volatility.  It  is  a  linear  regression  model  where  the  minimisation  of  the  sum   of  squared  errors  between  observations  and  predictions  will  determine  the  coefficients.      

 The   two   variables   of   interest   are   the   amount   of   government   expenditure   as   a   percentage   of   GDP   (GOVEXP)   and   the   debt-­‐to-­‐income   ratio   (D2INC).   To   analyse   their   unbiased   effect,   they   are   regressed   together   with   demographic   variables   and   control   variables  in  the  volatility  equation  displayed  below:  

 

!"#$! =   !!+ !!"!"#$%& + !!"!2!"# + !!"  !"#$%&!"ℎ!"#  &  !"#$%"&' !

!!! + !!  

 

According   to   the   second   hypothesis,   it   is   expected   that   the   absolute   size   of   the   coefficients  increase  as  the  homeownership  rate  distances  itself  from  the  inflexion  point.   This   means   that   the   five   values   for  !!  will   be   compared   at   first,   followed   by   the   five   values  for  !!.  Thus  the  hypothesis  will  be  tested  for  the  effect  of  both  rental  policies  and   borrowing  constraints.    

 

In   the   relevant   literature   the   GOVEXP   variable   has   a   negative   correlation   with   homeownership,  and  should  thus  also  affect  volatility  in  the  housing  market  (see  (Fisher & Jaffe, 2003) (Matznetter, 1994)).   As   it   promotes   renting,   the   decrease   in   the   homeownership  rate  should  have  a  positive  effect  on  volatility  to  the  left  of  the  inflexion   point,  whereas  this  effect  will  be  negative  to  the  right.    

The  opposite  is  expected  of  the  D2INC  variable,  as  a  higher  value  should  prompt  more   homeownership,   as   more   mortgage   debt   is   present   in   the   economy.   Consequently,   an  

(22)

increase  it  is  expected  to  increase  volatility  on  the  right  side  of  the  inflexion  point,  but   decrease  it  on  the  left.  This  is  partly  confirmed  in  the  literature,  where  an  increase  in  the   down-­‐payment  constraint  drives  US  volatility (Stein, 1995).  

 

The  demographic  variables  that  are  used  in  the  model  are  a  combination  of  the  number   of  households,  the  proportion  of  one-­‐person  households,  the  average  household  size,  the   rate   of   employment,   of   education,   and   of   migration.   In   terms   of   control   variables,   the   wealth  of  the  population  is  measured  through  either  GDP  per  capita  or  income.  Finally,   the  rate  of  inflation  and  the  rate  of  house  price  appreciation  complete  the  set.    

 

In   the   relevant   literature,   income   is   negatively   correlated   with   volatility,   and   inflation   tends   to   be   high   and   increasing   when   volatility   increases   (Dolde & Tirtiroglue, 2002).   Furthermore,   both   increases   in   population   and   employment   are   found   to   positively   affect  volatility  in  the  US (Reichert, 1990).  The  appreciation  of  house  prices  is  positively   correlated   with   volatility (Miller & Peng, 2006),   certainly   considering   volatility   that   is   constructed   using   past   rates   of   appreciation.   The   remainder   of   the   demographic   variables  that  are  not  covered  in  the  literature  shall  add  another  novel  dimension  to  the   findings  reported  in  section  5.    

                             

Referenties

GERELATEERDE DOCUMENTEN

'fabel 1 memperlihatkan jum1ah perkiraan produksi, jumlah perusahaan/bengkel yang membuat, jenis traktor yang diproduksi, asa1 desain dan tahap produksinya.. Jenis

Sub question 4: What will be the likely effects of Borneo’s soil, air and biodiversity changes on the palm oil industry at Borneo in the next 20 years.. As indicated by sub question

Zoals in het vorige hoofdstuk was vermeld, bepaald het dal tussen twee pieken de beste threshold waarde voor de Contour filter die vervolgens toegepast wordt om het gehele koraal

The purpose of this thesis is to provide an insight into the extent to which the Tribunal has provided victim witnesses a platform for truth telling, as one of the needs of

As ownership concentration (blockholder ownership) is high in Continental Europe, which is confirmed by Appendix A, corporate governance in Continental Europe

Since I hypothesized that the political risk has a positive relationship with the exchange rate volatility and a negative effect of exchange rate volatility on FDI inflows, a

This paper stands on the FDI host country point, tested how the relative exchange rate change, the relative company wealth in investor country, the relative Ownership

Next to that, due to a higher percentage of companies that are exposed when stock returns are lagged with one period these levels of exposure will be applied in the