Aid  Intensity  and  Firm  Subsidiaries   Reaction  in  Hungary   Pálma  Mosberger  (Joint  Research  Center)   Zsuzsanna  Varga  (Hungarian  National  Economic  Ministry)  1   Introduction     This   study   examines   whether   firms   react   to   changes   in   investment   incentives   provided  by  the  government.  We  analyze  firms’  location  and  investment  choice  in  a   Central  Eastern  European  country,  Hungary,  based  on  subsidiary  level  administrative   tax   microdata.   In   mid-­‐2014,   in   accordance   with   new   European   Commission   regulations,  the  aid  intensity  ceiling  aligning  the  amount  of  regional  development  aid   (state  and  European  Union)  was  significantly  modified  in  Hungary.       The   study   is   based   on   the   municipality   business   tax   registry,   which   contains   administrative   microdata   on   all   firm   subsidiaries   for   the   period   of   2013-­‐2016.   As   the   municipality   dataset   does   not   contain   information   on   investment,   we   use   the   available   turnover   as   a   proxy.   We   estimate   how   the   revenue   changed   for   firms   located   in   regions   differently   affected   by   the   aid   intensity   change   by   using   a   difference   in   difference   (DID)   setup.   The   coefficient   of   interest   is   significant   positive,   but   small   in   magnitude,   suggesting   firms   reacted   to   the   aid   intensity   change   and   turnover  decreased  more  on  average  where  the  aid  intensity  was  reduced  more.  The   regression  results  measure  together  the  extensive  and  intensive  responses.     After   a   short   literature   review,   we   describe   the   policy   background   in   subchapter   2   and   the   data   in   subchapter   3.   Subchapter   4   contains   the   results   and   methodology,   and  then  we  conclude.                                                                                                               1  Opinions  expressed  in  this  paper  are  those  of  the  authors  and  do  not  necessarily  reflect  the  views  of   their  institutions.     Literature  review     Firms   decide   where   to   (re-­‐)   locate   production   in   order   to   optimize   profit.   A   firm’s   decision   on   where   to   locate   investment   and   open   a   new   subsidiary   is   affected   by   many   factors.   Dijk   and   Pellenbarg   (2000)   categorize   these   factors   as   'firm   internal'   factors  (e.g.,  quality  of  management,  organizational  goals,  ownership  structure,  fixed   capital   investment),   'location'   factors   (e.g.   lot   size   and   size   of   possible   expansion   space;  accessibility  by  road,  distance  to  customers  and  suppliers),  and  'firm  external'   factors   (e.g.,   government   policy,   regional   economic   structure,   technological   progress,   labor   market   etc.).     In   this   paper   we   focus   on   the   affect   of   government   policy  on  firms’  location  choice.     There   are   papers   analyzing   empirically   firms’   response   to   government   incentives.   Devereux  et.  al  (2003)  examines  the  determinants  of  new  plants  locations  in  Great   Britain.   They   find   that   Regional   Selective   Assistance   aid   grant   has   an   effect   in   attracting  plants  beside  the  regional  industrial  structure.  Bronzini  et.  al  (2008)  look   at   the   impact   of   investment   tax   credit   on   business   investment.   They   analyze   the   effect   of   an   Italian   reform   where   the   amount   of   tax   credit   differed   across   areas   of   eligibility,   and   the   amount   of   the   tax   deduction   decreased   with   the   level   of   local   development.   The   paper   provides   evidence   that   the   program   was   effective   stimulating  investment.  Bellak  and  Leibrecht  (2011)  study  Central  and  East  European   countries  and  show  evidence  that  tax-­‐lowering  strategies  have  an  important  impact   on   foreign   firms’   location   decision.     On   the   other   hand   Beyer   (2002)   looks   at   the   relation  between  tax  incentives  and  the  level  of  FDI  in  Eastern  European  transition   countries  and  find  no  significant  relationship.       Our   paper   is   the   first   to   analyze   firms’   responses   to   changes   in   the   regional   development  aid  in  Hungary  based  on  municipality  administrative  tax  data.   Policy  Background     Hungary  offers  companies  located  within  its  borders  several  investment  incentives,   among   which   are   the   state   aid   subsidies   and   development   (investment)   tax   credit.   The  Regional  Aid  Map  is  aligned  with  the  maximum  amount  of  state  and  European   Union  aid  subsidies,  and  development  tax  credit  that  EU  Member  States  can  give  to   companies  to  promote  investment.  In  less  developed  regions  the  level  of  investment   is   lower   as   reported   in   Table   1,   which   contains   the   per   capita   GDP   in   HUF   and   as   percentage   of   the   average   (columns   1   and   2),   as   well   as   the   total   investment   by   region  (column  3)  and  per  capita  investment  as  percentage  of  the  average  (column   4).       Aid   intensity   is   calculated   as   the   ratio   of   the   total   amount   of   EU   and   state   aid   and   development   tax   credit   relative   to   the   present   value   cost   of   new   investment.   Depending  on  how  developed  a  region  is,  the  aid  intensity  ceiling  in  Hungary  varies   from   0   to   50   percent.   It   is   based   on   the   geographical   location   of   the   subsidiary’s   business  activity  and  investment,  not  on  where  the  firm’s  headquarters  are  located.   The  state  and  EU  aid  is  a  cash  flow  transferred  during  the  investment  period,  while   the  development  tax  credit  can  be  used  to  reduce  CIT  tax  payment  in  10  years  after   the   completion   of   the   investment.   The   ultimate   aim   of   the   EU   regional   aid   is   to   promote   development   of   the   less   advantaged   regions   of   Europe.   In   principle   regions   with   a   GDP   per   capita   below   75%   of   the   EU   average   are   eligible   for   regional   investment  aid.  Based  on  EU  legislation  Member  States  can  grant  state  aid  to  firms   to  support  investment  in  new  production  facilities  or  to  extend  or  modernize  existing   facilities  in  the  disadvantageous  regions  in  order  to  support  economic  development   and  employment.       We  would  have  liked  to  evaluate  the  development  tax  credit  in  this  study,  but  only   approximately   80-­‐100   firms   requested   this   tax   credit   during   the   last   years.2   So   instead,   the   study   captures   the   link   between   the   change   in   aid   intensity   and   subsidiary   behavior   via   the   state   aid   and   EU   aid   link.   During   the   period   of   2011-­‐2016   the  total  amount  of  development  state  aid  was  185  billion  forint.                                                                                                                   2  Based  on  statistics  of  the  Finance  Ministry.   Table  1:  Per  capita  GDP,  per  capita  GDP  as  percentage  of  the  average,  total  investment  by  Geographical  Area     Per*capita*GDP* Per*capita* Total* Per*capita* (in*thousand* GDP*as*%*of* investment*(in* investment*as*%* Geographical*area HUF) the*average million*HUF)* of*the*average 2013 Budapest 6*573 215% 934*398 130% Pest* 2*568 84% 400*275 79% Central*Transdanubia 2*735 89% 470*217 106% Western*Transdanubia 3*082 101% 606*018 148% Southern*Transdanubia 2*071 68% 263*640 69% Northern*Hungary 1*865 61% 335*460 68% Northern*Great*Plain 1*933 63% 550*520 89% Southern*Great*Plain 2*118 69% 475*580 89% Total*Hungary 3*057 100% 4*108*024 100%   Source:  Statistical  Office  (KSH)  Stadat  Table  6.1.1,  Table  6.3.1  and  Table  6.3.3.  The  geographical  area   of  investment  is  based  on  the  actual  location  of  investment.  2013.     Starting   in   mid-­‐2014,   there   were   significant   changes   to   the   Regional   Aid   Map   of   Hungary   in   response   to   the   obligatory   adoption   of   Commission   Regulation   (EU)   No   651/2014.   The   change   was   introduced   due   to   the   economic   development   of   the   Central   Hungary   region   over   the   last   funding   framework   period.   Enterprises   in   the   capital   city,   Budapest,   are   no   longer   eligible   for   either   the   development   tax   credit   or   a  state  aid  investment  subsidy.3  Also,  subsidiaries  in  only  86  of  the  183  towns  in  the   central  Pest  county  near  the  capital  are  eligible.  Aid  intensity  was  also  reduced  by  5   percentage   points   in   the   Western   and   Central   Transdanubia   regions.   Meanwhile   it   remained  at  50  percent  in  the  less  developed  4  regions.  Table  2  shows  aid  intensity   before  and  after  the  mid-­‐2014  change  for  each  region,  and  Figure  1  shows  where  the   regions   are   located.   There   are   4   different   size   intensity   changes   in   3   regions   (Budapest   and   Pest   county   in   the   Central   Hungary   region,   Western   Transdanubia   region  and  Central  Transdanubia  region)  affecting  nearly  440  thousand  subsidiaries.                                                                                                                 3  The  supplementary  regulation  that  independently  of  the  location,  small  enterprises   may   claim   an   additional   20   percentage   point   increase   of   aid   intensity   threshold,   and   medium   size   enterprises   an   additional   10   percentage   point   remained   still   in   force   also  after  the  2014  reform  (2014/165  5§).  As  there  is  no  information  on  the  size  of   enterprise   in   the   dataset   we   cannot   take   into   consideration   this   additional   size   dependent   factor   and   we   assume   that   size   category   (small,   medium   or   large)   changes  are  independent  of  the  geographical  aid  intensity  reform.   Table  2:  Aid  Intensity  Changes  in  Hungary  by  Geographical  Area  (Regional  Aid  Map)   Funding&intensity Nbr.&of& Before& After& Change&in&intensity& Geographical&area subsidiaries& 2014 2014 (percentage&point) in&2013 1 10% 0% B10 Budapest&(capital) 134&601 2 30% 0% B30 Pest&&(apart&from&86&towns) 58&288 3 30% 20% B10 4&towns&in&Pest& 3&799 4 30% 35% 5 82&towns&in&Pest& 49&881 5 30% 25% B5 Western&Transdanubia 96&446 6 40% 35% B5 Central&Transdanubia 94&920 7 50% 50% 0 Southern&Transdanubia,&Northern&Hungary,& 426&961 Northern&Great&Plain,&Southern&Great&Plain     Figure  1:  Hungarian  Regions       Data     Besides  the  national  corporate  income   tax,  Hungarian  firms  also  pay  local  business   taxes   (HIPA,   helyi   iparűzési   adó)   in   the   municipalities   where   subsidiaries   are   registered.  The  local  business  tax  form  has  to  be  filed  and  taxes  paid  by  May  31st  of   the  year  following  the  fiscal  year.  The  legislation  defines  how  firms  must  allocate  the   local   tax   base   between   subsidiaries   based   on   their   actual   business   activities   (e.g.   based  on  the  ratio  of  wages  paid  by  each  subsidiary  to  the  overall  wage  bill,  similarly   on   the   cost   ratio,   or   on   both).   The   corporate   income   tax   is   paid   at   company   level   for   the   whole   organization   including   all   subsidiaries;   hence   the   tax   form   only   contains   aggregated   firm   level   information.   The   local   business   tax   is   paid   to   local   municipalities   and   each   subsidiary   submits   separately   the   tax   form   including   subsidiary   level   information.   To   study   firm   subsidiary   location   choice   we   need   subsidiary  level  information  that  is  only  available  on  the  local  tax  forms.     The   local   business   tax   administrative   database   used   in   this   study   contains   yearly   anonymous  subsidiary  data  for  2013-­‐2016  collected  by  the  Hungarian  Treasury  and   received   by   the   Ministry   of   Finance.   It   is   an   administrative   database   containing   information  from  all  submitted  tax  returns,  though  the  number  of  variables  is  limited   including  yearly  subsidiary  level  information  on  postal  code,  county,  industry  code,   and   firm   ID.   It   also   contains   some   basic   variables   from   the   local   tax   form,   such   as   firm   level   turnover,   material   costs,   R&D   costs,   local   tax   base,   and   tax   to   be   paid.   For   firms  with  subsidiaries  the  municipality  tax  form  contains  a  variable  (called  weight)   on   how   the   subsidiary   tax   base   were   allocated   between   the   subsidiaries   based   on   actual  business  activity.  We  use  these  business  activity  ratios  to  compute  subsidiary   level  turnover  from  the  reported  firm  level  turnover  data.     Because   the   database   contains   only   parent   company   ID   and   no   subsidiary   IDs,   we   compute   the   year   of   subsidiary   entry   based   on   the   first   appearance   of   a   new   subsidiary   in   a   new   location.   As   the   database   contains   only   information   for   2013-­‐ 2016,  it  is  thus  only  possible  for  us  to  compute  firm  entry  starting  in  2014.  Subsidiary   exit  is  computed  similarly  based  on  the  last  appearance  in  the  data  (e.g.,  exit  as  of   the  beginning  of  2016  if  last  appearance  is  in  2015).  The  entry/exit  measures  are  not   perfect,  but  data  limitations  prevent  us  the  use  of  any  other  method  of  computation.     This   is   the   first   study   to   draw   on   the   local   business   tax   micro   database,   and   use   information   on   the   local   address   of   firm   subsidiaries.   While   the   national   corporate   income   tax   database   contains   balance   sheet   information   (capital   stock.   etc.)   and   would   make   it   possible   to   compute   investment,   but   the   two   datasets   cannot   be   linked  due  to  data  anonymity.       Methodology  and  Results     The  identification  is  based  on  the  2014  change  of  the  regional  aid  intensity  map.  The   capital  aid  intensity  change  affected  about  440  thousand  subsidiaries,  more  than  50   percent  of  all  subsidiaries.  In  the  Northwestern  regions  the  intensity  was  reduced  by   five  percentage  points.  In  the  capital  the  aid  intensity  was  eliminated.  In  the  central   region,   for   82   municipalities   the   aid   intensity   increased,   in   4   municipalities   it   decreased,   and   in   the   remaining   municipalities   it   was   eliminated.   (See   Table   2   and   Figure  1)       In  this  chapter  first  we  look  at  descriptive  statistics,  then  we  also  estimate  how  the   revenue  changed  for  firms  located  in  regions  differently  affected  by  the  aid  intensity   change   by   using   a   difference-­‐in-­‐difference   (DID)   setup.   As   the   municipality   dataset   does   not   contain   information   on   investment,   we   use   the   available   turnover   as   a   proxy  for  investment.     Descriptive statistics   In   this   subchapter   we   look   at   the   correlation   between   the   2014   change   in   aid   intensity   and   the   choice   of   where   to   locate   subsidiaries.   There   is   a   positive   correlation   between   subsidiary   entry   and   the   change   in   intensity,   and   a   negative   correlation  between  firm  exit  and  the  change,  suggesting  there  might  be  some  effect   of  the  reform  on  firm  location  choice.     Table   3   shows   information   for   the   seven   regions   in   Hungary   that   were   affected   differently   by   the   Regional   Aid   Map   change   introduced   in   mid-­‐2014.   The   top   section   depicts   the   changes   in   aid   intensity   for   each   region   and   the   number   of   subsidiary   entries  for  2014–2016.  The  last  three  columns  refer  to  firms  already  existing  before   the  study  period.       The  bottom  section  of  Table  3  reports  percentage  changes  in  firm  entry  with  respect   to   2014   in   each   region.   We   chose   2014   as   the   control   year   because   due   to   data   limitations  it  was  the  earliest  computable  entry  year  (as  data  is  available  only  from   2013).   The   change   in   the   Regional   Aid   Map   went   into   effect   in   mid-­‐2014.   If   firms   reacted   in   2014,   then   choosing   this   as   the   base   year   would   bias   the   result   in   the   direction  of  us  not  finding  an  effect.         Table  3:    Aid  Intensity  and  Subsidiary  Entry  by  Region,  Percent   N&of&subsidiary&entry&if&firm&existed& Funding&intensity N&of&subsidiary&entry already&in&2013 Change&in& Before&2014 After&2014 percentage& 2014 2015 2016 2014 2015 2016 points Budapest&(capital) 10% 0% >10 40597 32192 99778 12153 7769 6825 Pest 30% 0% >30 9753 6164 14440 2125 1396 3663 Pest&(4&towns) 30% 20% >10 565 425 814 119 92 272 Pest&(82&towns) 30% 35% 5 7561 5570 10872 1439 1078 2741 3&counties 30% 25% >5 14633 10970 18188 3847 2706 5219 3&counties 40% 35% >5 15034 11779 19308 3630 2806 5349 12&counties 50% 50% 0 63240 52396 75480 13370 11137 19446 Change&bw&N&of&yearly&entry Change&bw&N&of&yearly&entry 2014>2015 2014>2016 2014>2015 2014>2016 Budapest&(capital) >0,21 1,46 >0,36 >0,44 Pest >0,37 0,48 >0,34 0,72 Pest&(4&towns) >0,25 0,44 >0,23 1,29 Pest&(82&towns) >0,26 0,44 >0,25 0,90 3&counties >0,25 0,24 >0,30 0,36 3&counties >0,22 0,28 >0,23 0,47 12&counties >0,17 0,19 >0,17 0,45 Correlation&bw&change&in&intensity&and&change&in&N&of&firms&entry All 0,74 >0,21 0,61 >0,01 All&except&Budapest 0,80 >0,50 0,70 >0,13 Only&Pest 0,85 %0,93 0,80 0,39     The  bottom  block  reports  the  correlation  between  change  in  intensity  and  change  in   subsidiary  entries.  “All”  refers  to  all  regions.  The  correlation  for  2014–2015  is  large   and   positive,   i.e.,   the   larger   the   decrease   in   intensity,   the   fewer   subsidiaries   are   entering   the   region.   For   2014–2016   the   correlation   is   negative   because   a   large   number   of   subsidiaries   opened   in   Budapest.   The   next   line   excludes   the   capital   to   check  the  possibility  that  other  dimensions  might  affect  the  choice  of  location  for  a   new   investment   besides   aid   intensity   (e.g.,   a   more   highly   skilled   work   force   near   larger   cities   or   access   to   airports   and   motorways).   Probably   investors   choose   the   capital  regardless  of  the  new  rules  on  aid  intensity.       The   last   line   shows   the   correlation   between   entry   and   intensity   changes   between   the  towns  located  only  in  the  county  of  Pest  but  affected  differently  by  the  reform.  It   may   be   that   towns   within   Pest   are   more   similar   in   the   eyes   of   investors   and   difference   in   entry   choice   is   more   determined   by   the   change   in   aid   intensity.   The   correlation  is  positive  and  strong  for  both  2015  and  2016.     Table  4:  Change  in  Aid  Intensity  and  Changes  in  Subsidiary  Exits  by  Region     N&of&subsidiary&exit&if&firm&existed& Funding&intensity N&of&subsidiary&exit already&in&2013 Change&in& Before&2014 After&2014 percentage& 2014 2015 2016 2014 2015 2016 points Budapest&(capital) 10% 0% >10 28444 21062 23421 17295 15776 15481 Pest 30% 0% >30 7628 11978 11735 7115 9489 9184 Pest&(4&towns) 30% 20% >10 446 529 1182 386 402 942 Pest&(82&towns) 30% 35% 5 6122 7400 8893 5950 5724 6645 3&counties 30% 25% >5 10786 11157 16373 10337 8814 12306 3&counties 40% 35% >5 11404 11413 18164 11301 9043 13590 12&counties 50% 50% 0 49870 49146 75843 44367 38127 56089 Change&bw&N&of&yearly&exit Change&bw&N&of&yearly&exit 2014>2015 2014>2016 2014>2015 2014>2016 Budapest&(capital) >0,26 >0,18 >0,09 >0,10 Pest 0,57 0,54 0,33 0,29 Pest&(4&towns) 0,19 1,65 0,04 1,44 Pest&(82&towns) 0,21 0,45 >0,04 0,12 3&counties 0,03 0,52 >0,15 0,19 3&counties 0,00 0,59 >0,20 0,20 12&counties >0,01 0,52 >0,14 0,26 Correlation&bw&change&in&intensity&and&change&in&N&of&firms&exit All >0,57 >0,06 >0,82 >0,14 All&except&Budapest >0,80 >0,14 >0,84 >0,19 Only&Pest !0,88 0,02 !0,97 !0,04     Table   4   gives   the   same   information   as   in   Table   3   for   exit   of   subsidiaries   from   regions   affected  by  the  Regional  Aid  Map  change.  The  strong  negative  correlation  shows  that   where   aid   intensity   decreased   more,   the   more   subsidiaries   exited,   which   suggests   that  firm  location  choice  might  depend  on  investment  aid.       Difference-in-difference   We   estimate   how   the   revenue   changed   for   firms   located   in   regions   differently   affected  by  the  aid  intensity  change  by  using  a  difference  in  difference  (DID)  setup.   We  use  the  available  turnover  as  a  proxy  for  investment,  as  investment  information   is   not   available   in   the   municipality   dataset.   In   this   DID   setup   the   treatment   is   the   different   aid   intensity   changes   ranging   between   -­‐30   and   5   percentage   point   for   different  regions  after  the  2014  reform  (see  Table  2).  The  treatment  group  includes   firms   located   at   regions   of   the   country   where   the   aid   intensity   changed,   while   the   control  group  includes  firms  in  regions  where  did  not  change.     Simply   comparing   firms   in   the   treatment   group   before   and   after   the   reform   would   contain   the   effect   of   the   reform   and   also   the   additional   changes   in   the   macroeconomic   environment.   To   estimate   what   part   of   the   change   is   due   to   the   reform   and   what   part   would   have   been   realized,   we   compare   changes   in   the   treatment   group   to   changes   in   the   control   group   before   and   after   the   reform.   The   underlying   assumption   in   the   DID   estimation   is   that   firms   in   the   treatment   and   control   groups   were   similar   and   hence   in   the   absence   of   the   reform   would   have   behaved   similarly.   This   way   the   difference   between   the   change   in   the   treatment   minus   the   change   in   the   control   group,   i.e.   the   difference-­‐in-­‐differences   (DID),   identifies  the  effect  of  the  reform.         We   estimated   the   following   DID   regression   specification,   where   t   is   a   dummy   variable,   0   for   pre-­‐reform   years   and   1   for   after   reform   years.   D   captures   possible   differences   between   the   treatment   and   control   groups   prior   to   the   policy   change.   In   this  regression  specification  D  is  not  the  usual  0  -­‐  1  (not  treated  vs.  treated)  dummy   variable,  but  a  treatment  intensity  variable,  which  is  the  change  in  the  percentage  of   aid   intensity   (5,   0,   -­‐5,   -­‐10,   -­‐30).   The   δ   coefficient   of   Dt   is   the   main   coefficient   of   interest,  which  measures  the  effect  of  the  reform.  The  positive  estimated  coefficient   means   the   higher   is   the   decrease   in   aid   intensity,   the   higher   is   the   reduction   in   turnover.  If  it  is  positive  then  it  provides  evidence  that  increasing  the  aid  intensity  on   average   increases   the   turnover,   while   decreasing   the   aid   intensity   decreases.   X   includes  county  level  controls.  Subscript  i  refers  to  subsidiaries  and  j  for  time.       !! = + !! + !! + !! !! + + !     We  check  the  parallel  trend  assumption  based  on  municipality  level  aggregated  data   collected  by  the  Hungarian  Statistical  Office,  which  is  available  for  pre-­‐reform  years.   It   contains   information   on   the   number   of   employees,   personal   income   tax   and   municipality  business  tax  (see  Table  5).4     Table  5:  Municipality  level  statistics  for  pre-­‐reform  years  (parallel  trend)   Number'of'employees Treatment(group 2012 2013 2014 Budapest'(capital) 575'193 569'091 571'198 Pest 266'038 270'476 258'090 Pest'(4'towns) 10'114 10'398 7'111 Pest'(82'towns) 233'928 238'710 207'749 3'counties 369'199 371'875 504'060 3'counties 496'018 501'836 480'409 Control(group 12'counties 2'046'403 2'073'331 2'175'582 Personal'income'tax''(million'HUF) Treatment(group 2012 2013 2014 Budapest'(capital) 1'849'303 1'896'423 2'037'311 Pest 587'668 611'660 663'657 Pest'(4'towns) 21'118 22'302 24'324 Pest'(82'towns) 441'960 458'422 497'078 3'counties 689'703 717'018 780'205 3'counties 968'022 1'008'741 1'096'025 Control(group 12'counties 3'628'909 3'806'289 4'167'644 Local'municipality'business'tax'(million'HUF) Treatment(group 2012 2013 Budapest'(capital) 840425 995464 Pest 335094 367116 Pest'(4'towns) 5248 5825 Pest'(82'towns) 183052 188594 3'counties 291336 306695 3'counties 459669 481562 Control(group 12'counties 1613425 1704533   Note:  Based  on  statistics  of  the  Hungarian  Satistical  Office  (KSH)   Results   Before   presenting   the   DID   regression   estimates   of   the   reform   in   Table   7,   we   present   evidence   on   that   turnover   is   a   valid   proxy   for   investment   in   Table   6.   To   check   whether  there  is  a  positive  relation  between  turnover  and  investment  we  look  at  the   firm   level   corporate   income   tax   (CIT)   administrative   database   (available   for   the                                                                                                               4  The  parallel  trend  cannot  be  checked  based  on  the  municipality  level  micro  tax  data   as  it  is  not  available  for  many  pre-­‐reform  years,  neither  based  on  the  CIT  firm  level   admin  data  as  it  has  no  subsidiary  level  information.     universe  of  firms  for  the  period  of  2007-­‐2013).5  This  database  contains  no  subsidiary   level  information,  only  firm  level  information.  The  regressions  results  are  presented   in   Table   6,   control   variables   are   progressively   added   to   the   regressions,   the   last   column   containing   the   square   of   log   investment,   year   dummies,   number   of   employees,  profit  and  balance  sheet  variables.  The  positive  significant  coefficient  of   log  investment  confirms  the  positive  correlation  between  turnover  and  investment.     Table  6:  Positive  correlation  between  turnover  and  investment   (1) (2) (3) (4) Variables log(turnover) log(investment) .54*** .59*** .57*** .52*** (.00207) (.00215) (.00304) (.00785) log(investment)2 .0035*** (.000683) profit 1.3e?08*** 1.2e?08*** (4.17e?09) (4.10e?09) nbr@employee .00079*** .00075*** (.000264) (.000258) balance@sheet 1.0e?10 2.4e?11 (4.25e?10) (3.92e?10) d_2008 ?8.3*** ?8.6*** ?8.6*** (.177) (.182) (.181) d_2009 1.4*** 1.3*** 1.4*** (.0112) (.0125) (.0138) d_2010 1.4*** 1.3*** 1.3*** (.0114) (.0129) (.0142) d_2011 1.4*** 1.3*** 1.3*** (.0117) (.0132) (.0143) d_2012 1.4*** 1.3*** 1.3*** (.012) (.0133) (.0143) d_2013 1.3*** 1.3*** 1.3*** (.012) (.0131) (.0139) Constant 6.5*** 5.1*** 5.3*** 5.4*** (.0168) (.0212) (.026) (.0263) Observations 400,800 400,800 387,354 387,354 R?squared .367 .403 .409 0.409 Robust@standard@errors@in@parentheses ***@p<0.01,@**@p<0.05,@*@p<0.1   Note:  CIT  micro  database.    In  the  regressions  the  dependent  variable  is  log  turnover,  the  independent   variable   is   log   investment,   controls   includes   square   of   log   investment,   year   dummies,   number   of   employees,  profit,  balance  sheet.  Standard  errors  are  clustered  at  firm  level.                                                                                                                 5  Firms  do  not  report  investment  in  the  corporate  tax  form.  We  calculate  investment   based   on   balance   sheet   information   available   in   the   CIT   database   the   following   way:   investment   =   Δ   capital   +   amortization,   where   amortization   is   the   available   accounting   amortization   value   and   we   assume   investment   takes   place   at   the   beginning  of  the  year.   Table  7  reports  the  difference-­‐in-­‐difference  regression  results.  As  a  common  practice   in   the   literature,   the   dependent   variable   is   top   coded   for   the   top   1%   to   avoid   that   outliers  might  drive  the  result.  The  change  in  turnover  is  winsorized  at  the  bottom   1%  and  top  99%,  and  the  final  sample  in  the  regressions  contains  firms  that  were  not   dropped   during   the   winsorization   process.   Regressions   include   firms   with   less   than   50  subsidiaries  (220  firms  were  excluded  from  the  sample).  Variables  are  in  current   value  including  inflation.  The  panel  dataset  is  balanced.       In  the  first  four  columns  the  treatment  group  includes  firms  located  in  the  parts  of   the  country  where  the  aid  intensity  changed  (this  change  varies  between  -­‐30  and  5   percentage  point)  and  the  control  group  includes  firms  in  the  regions  where  the  aid   intensity   did   not   change.   Controls   are   added   gradually   to   the   regressions.   In   the   first   column   only   the   time   dummy,   treatment   intensity   and   interaction   terms   are   included,  and  then  county  level  controls  are  added  in  the  second  column.  The  third   column   also   includes   the   number   of   subsidiaries   and   county   dummies,   while   the   fourth  column  controls  for  industry  structure  also.  The  regressions  results  measure   together   the   extensive   (eg.   opening/closing   a   subsidiary)   and   intensive   responses   (changes   in   turnover).   The   coefficients   of   interest   are   significant   and   positive   in   all   specifications,   but   small   in   magnitude,   suggesting   a   positive   relation   between   aid   intensity  change  and  turnover.  The  0.01  coefficient  suggests  that  if  the  aid  intensity   is   increased   by   1   percentage   point,   then   on   average   the   turnover   is   expected   to   increase  by  1  percent.       In  the  last  column  only  subsidiaries  situated  in  the  municipalities  of  Pest  county  are   included.   Within   Pest   county   three   different   changes   in   aid   intensity   were   introduced.   It   might   be   that   towns   within   Pest   are   more   similar,   and   hence   the   change   in   turnover   is   more   determined   by   the   change   in   aid   intensity.   In   the   regression   the   baseline   contains   firms   located   in   municipalities   with   30   percentage   point   reduction   in   aid   intensity,   group   D1   contains   firm   located   in   towns   with   10   percentage  point  reduction  and  D2  with  5  percentage  point  increase  (see  Table  2).   The   coefficient   of   t*D1   is   not   significant,   possibly   due   to   that   only   few   firms   were   located  in  the  4  towns  affected  by  the  10  percentage  point  reduction.  The  positive   and  significant  coefficient  of  t*D2  suggests  that  firms  in  municipalities  where  the  aid   intensity   was   increased   also   increased   their   turnover   compared   to   the   baseline   group.     Table  7:  Difference-­‐in-­‐difference  estimates  for  changes  in  firm  revenue  between  pre-­‐reform  (2013)   and  after  reform  years  (2015-­‐2016)   (1) (2) (3) (4) (5) Region All1geographical1area Only1Pest1county Variables log1revenue1(in1million1HUF) time1dummy .05*** @.17*** @.49*** @.51*** @.95*** (.0103) (.0141) (.0155) (.0155) (.0309) treatment1intensity .05*** .01*** .01*** .00** change (.000985) (.00103) (.00114) (.00112) time%X%treatment%intensity .02*** .02*** .01*** .01*** change (.000991) (.00102) (.00102) (.00102) log1investment1(county) @.48*** .33*** .32*** (.0141) (.0333) (.0333) log1avg1wage1(county) @1.8*** 13*** 13*** (.0906) (.235) (.235) log1unempl1rate1(county) @.59*** .75*** .72*** (.0232) (.0311) (.0311) nbr.1of1subsidiaries X X county1dummy X X industry1dummy X D11(towns1with1@101pp1int.1 @.084 change1in1Pest) (.115) D21(towns1with151pp1int.1 .089** change1in1Pest) (.0429) D1%X%time .17 (.116) D2%X%time .61*** (.0448) Constant 7.2*** 36*** @169*** @170*** 6.9*** (.0103) (.995) (3.25) (3.25) (.0359) Observations 3,330,134 3,330,134 3,330,134 3,330,066 412,647 R@squared .00445 .0138 .0182 .0419 .00519 Robust1standard1errors1in1parentheses ***1p<0.01,1**1p<0.05,1*1p<0.1   Note:   For   the   first   four   columns   the   treatment   group   includes   firms   located   in   part   of   the   country   where  the  aid  intensity  changed  and  the  control  group  includes  firms  in  where  the  aid  intensity  did   not   change   (see   Table   2).   Column   5   contains   only   firms   located   in   Pest   county   where   the   baseline   group  includes  firms  located  in  municipalities  with  -­‐30  pp.  intensity  change,  group  D1  in  municipalities   with  -­‐10  pp.  change  and  D2  municipalities  with  5  pp.  increase.  Standard  errors  are  clustered  at  firm   level.     The   coefficient   of   interest   is   significant   positive   in   the   above   specifications,   suggesting   firms   reacted   to   the   aid   intensity   change   and   turnover   decreased   more   on  average  where  the  aid  intensity  was  reduced  more.         Conclusions     This   study   examined   whether   firms   react   to   changes   in   investment   incentives.   In   mid-­‐2014,   in   accordance   with   new   European   Commission   regulations,   the   aid   intensity   ceiling   aligning   the   amount   of   regional   development   aid   (state   and   European   Union)   that   firms   might   receive   was   significantly   modified   in   Hungary.   The   study   is   based   on   the   municipality   business   tax   registry,   which   contains   administrative  microdata  on  the  universe  of  firm  subsidiaries  for  the  period  of  2013– 2016.   As   the   municipality   dataset   does   not   contain   information   on   investment,   we   used  the  available  revenue  as  a  proxy.  We  estimated  how  the  revenue  changed  for   firms   located   in   regions   differently   affected   by   the   aid   intensity   change   by   using   a   difference  in  difference  (DID)  setup.  The  estimated  coefficient  is  0.01  suggesting  that   if  the  aid  intensity  is  increased  by  1  percentage  point,  then  on  average  the  turnover   is  expected  to  increase  by  1  percent.       References   Bellak   C.,   Leibrecht   M.   (2009):   Do   low   corporate   income   tax   rates   attract   FDI?– Evidence  from  Central-­‐and  East  European  countries,  Applied  Economics  41  (21)   Beyer,  J.  (2002):  Please  invest  in  our  country.  How  successful  were  the  tax  incentives   for   foreign   investment   in   transition   countries?   Communist   and   post-­‐communist   studies,  35,  191-­‐211   Bronzini   R.,   Blasio   de   G.,   Pellegrini   G.,   Scognamiglio   A   (2008):   The   effect   of   investment  tax  credit:  Evidence  from  an  atypical  programme  in  Italy,  Banca  D’Italia   WP  N661   Devereux   M.,   Griffith   R.,   Simpson   H.   (2006):   Agglomeration,   regional   grants   and   firm   location,  IFS  WP  04/06   van   Dijk,   J.,   Pellenbarg,   P.   (2000):   Firm   relocation   decisions   in   The   Netherlands:   An   ordered  logit  approach,  Papers  Regional  Science  79/191