Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Informing Microfinance Policy-Makers through an Agent-Based Modelling Lifecycle

A88fffb92966b9cd12b7d1010e0cd768?s=47 pl
December 11, 2014

Informing Microfinance Policy-Makers through an Agent-Based Modelling Lifecycle



December 11, 2014


  1. Informing  Microfinance  Policy-­‐Makers     through  an  Agent-­‐Based  Modelling  Lifecycle

      Pablo  Lucas   University  College  Dublin   Complexity  and  Social  Simula;on  
  2. overview:  ra<onale   Complexity  and  Social  Simula;on   •  Behavioural

     economics  highlights  the  role  of  social  preferences  in  decisions.       Popula;ons  in  real  (not  experimental)  public  goods  tend  to  be  very  heterogeneous.     Group  composi;on  may  impact  the  sustainability  of  voluntary  public  good  contribu;ons.           •  Conduct  agent-­‐based  simula<ons  of  contribu<ons  to  an  actual  public  good   (microfinance  groups),  varying  group  composi<on  and  social  preferences.     –  Why?    To  systema;cally  examine  the  effect  of  these  factors  on  contribu;ons,        as  an  actual  experimental  seIng  would  be  innapropriate.  
  3. agent-­‐based  modelling  (ABM)   Complexity  and  Social  Simula;on   Agents

       -­‐  computa;onal  model(s)  of  individual  behaviour(s)  with  mechanisms  for:      -­‐  sensing  the  environment  and  other  agents      -­‐  keeping  an  internal  state  about  what  they  know      -­‐  selec;ng  ac;ons  (i.e.  how  and  when  will  an  act  occur)         Interac<on    -­‐  rules  that  either  determine,  or  result  in  emergent,  organisa;on  paLerns    -­‐  asynchronous  /  dynamic  interac;on  may  prevent  a  full  analy;cal  approach    -­‐  scale  (number  and/or  density  of  agents)  maLers         Environment    (physical  /  symbolical)  representa;on  of  where  /  how      agents  (and  stakeholders!)  interact        
  4. ABM:  issues  and  purpose   Complexity  and  Social  Simula;on  

    Issues    -­‐  individual  /  social  behaviour  oRen  is  subject  to  influences  that  one  can    only  understand  with  detailed  contextual  (oRen  only  qualita;ve)  insights.        data  about  social  phenomena  is  oRen:        -­‐  non-­‐existent,  thus  requiring  fund/;me  to  collect  and  process  it        -­‐  unavailable  due  to  privacy  agreements,  such  as  non-­‐disclosure  agreements        -­‐  if  at  hand,  informa;on  is  typically  limited  (incomplete  or  outdated)       Purpose   -­‐  control  environments  and  repeat  experiments  to  test  hypotheses  according  to  a   implemented  theory  containing  assump;ons  about  individual  /  social  behaviour     -­‐  to  gain  insights  about  social  phenomena    (both  about  known  and  unknown  scenarios)     -­‐  discuss  findings  with  expert  stakeholders  /  policy-­‐makers        
  5. ABM:  lifecycle   Complexity  and  Social  Simula;on   Model design

    Implementation New knowledge? Social phenomena Evidence Simulation results Represent behaviour and processes Test hypotheses via what-if scenarios Validation and replication Observations and assumptions Data collection and analysis If needed, update representations How plausible is the model? 1 2 3 4 5 N.B.:  observa;ons  and  assump;ons  are  discussed  with  expert  stakeholders  to   differen;ate  what  should  be  modelled  from  what  one  should  simply  put  in  context.  
  6. public  goods:  value  and  importance   Complexity  and  Social  Simula;on

      •  The  provision  of  public  goods  provision  can  be  directly  linked  to  welfare  and  poverty  :   –  important  examples  of  public  goods  for  the  ci;zenry  include:      health  services,  schools,  clean  water,  urban  sanita;on,  roads,  microfinance,  etc.           •  Tradi<onal  economic  analysis  suggests  under-­‐provision  of  public  goods   –  stemming  from  individuals  considering  only  their  own  costs  and  benefits  associated  with  contribu;ng,  disregarding  benefits  accrued  from  others     –  behavioural  /  experimental  economics  suggest  that,  while  individuals      do  selfish  choices,  many  also  sacrifice  their  own  well-­‐being  for  others    (i.e.  there  is  significant  heterogeneity  in  social  preferences)    
  7. lThis  is  business  with  a  social  objec;ve,  which  is  to

     help  people  get  out  of  poverty.z        (Muhammad  Yunus)         ( a  public  good:  microfinance   Complexity  and  Social  Simula;on   Key  concepts  of  microfinance  (credit,  savings,  insurance,  etc.)      -­‐  clientele:  typically  those  who  cannot  access  tradi;onal  banking      -­‐  social  collateral:    the  collec;ve  responsibility  over  credit  effec;vely            subs;tutes  tradi;onal  assets  to  backup  funding      -­‐  methodology:  local  ins;tu;ons  implement  the  general  framework  with        adapta;ons  to  local  needs  (i.e.  developed  /  developing  economies)  
  8. microfinance:  value  and  importance   Complexity  and  Social  Simula;on  

    Non-poor Vulnerable non-poor Moderate poor Extreme poor Safety net unexpected shocks (economic, illness, crop failure, etc.) with poverty mitigation strategies without poverty mitigation strategies Poverty line TIME microfinance  services  (insurance,  credit,  etc.)  can  be  used  effec;vely  to  improve   clients’  circumstances  and  resilience  to  absorb  unexpected  shocks       in  other  words:  the  debate  is  similar  to  those  about  development  aid.   it  is  not  whether  it  is  effec;ve  or  not,  but  how  to  make  it  more  effec;ve.  
  9. the  case  study:  loca<on   Complexity  and  Social  Simula;on  

  10. 3 to 7 members, following an adaptation of the (Grameen

    Bank) pioneer -­‐  based  in  Chiapas,  Mexico,    with  over  20.000  ac;ve  clients  in  groups  of  3  to  7,    following  an  adapta;on  of  the                                                                      (Grameen  Bank)        use  assorta;ve  matching  (endogenous  and  geographically  bound  peer  selec;on)         -­‐    instead  of  considering  tradi;onal  assets,  required  by  tradi;onal  banks,    social  collateral  of  applicants  are  assessed  along  with  the  local  poverty  line         -­‐  understanding  the  financial  and  social  status  quo  of  their  clients  is  key  to  adapt  services    with  less  risk  (e.g.  how  ins;tu;onal  norms  interplay  with  group  compliance  strategies)       the  case  study:  context   Complexity  and  Social  Simula;on  
  11. the  case  study:  cross-­‐sec<onal  fieldwork   Complexity  and  Social  Simula;on

      Key  :  social  collateral  force  sharing  the  burden  of  repaying           4  surveys:  600  clients,  261  groups,   35  credit  officers,    2  policy-­‐makers     and  analysis  of  all  their  loans           3  clients  per  credit  centre  (200)   were  chosen  within  a  350  km  radius.   6  Mayan-­‐descendant  languages  and  Spanish.  
  12. discussion:  fieldwork  findings   Complexity  and  Social  Simula;on   -­‐

     financially  stable  ins;tu;ons  are  interested,  from  a  sustainability  perspec;ve,    to  beLer  understand  –  both  qualita;vely  and  quan;ta;vely  –  their  groups         -­‐    microfinance  ins;tu;ons  (MFIs)  have  to  avoid  overpowering  external    dependence  and  harness  the  social  collateral  (the  group  dynamics)       -­‐    yet  collec;ng  and  analysing  such  data  (specially  longitudinally)      can  be  prohibi;vely  expensive  and  ;me-­‐consuming  for  MFIs         -­‐    despite  the  likelihood  of  improving  understanding  on  how  clients  self-­‐  organise  regarding  defaul;ng  members  and  debt,  there  is  no  a  priori  guarantee  that    such  effort  will  produce  findings  directly  useful  for  policy-­‐making.    
  13. discussion:  fieldwork  findings   Complexity  and  Social  Simula;on   -­‐ 

    98%  of  surveyed  groups  paid  on  ;me  (72%  rural,  44%  with  credit  equally  distributed)   -­‐  groups  indeed  varied  from  a  minimum  of  3  and  maximum  of  7  members   -­‐  data  suggests  that  group  loca;on  can  influence  behaviour  for  dealing  with  defaulters   (rural  clients  tend  do  apply  social  sanc;oning,  whilst  urban  ones  tend  to  sanc;on  economically)         -­‐  majority  (>80%)  of  groups  had  rela=ves  in  their  business  and  social  collateral  networks     (tes;ng  this  with  an  ABM  lead  the  MFI  to  change  their  policy,    encouraging  rather  than  prohibi;ng  groups  formed  by  rela;ves)  
  14. discussion:  ABM  lifecycle   Complexity  and  Social  Simula;on    

      what  it  did  help  with?      -­‐  guidance  on  how  to  systema;cally  collect,  analyse  and  model  data  /  behaviours    -­‐  cross-­‐validate  (qualita;vely  at  the  micro  level,  quan;ta;vely  at  the  macro  level)    -­‐  differen;ate  what  are  the  emergent  and  intrinsic  proper;es  of  the  ABM    -­‐  impact  the  actual  phenomena  (through  the  process  of  modelling)             what  it  didn’t  help  with?      -­‐  determine  the  level  of  support  or  guidance  for  decision-­‐making    -­‐  frequency  and  depth  of  researcher-­‐prac;;oner  collabora;on  
  15. discussion:  ABM  lifecycle   Complexity  and  Social  Simula;on   -­‐

       decision-­‐makers  par;cipated  in  every  stage  of  the  research  process  for  elici;ng    data  and  modelling  assump;ons  into  the  ABM           -­‐  fieldwork  findings  were  more  promptly  related  to,  both  in  terms  of    ;me  and  scale,  to  a  known  state  of  their  socio-­‐economic  phenomenon.      (i.e.  it  is  important  to  relate  results  to  events  that  can  s;ll  be  observed)           -­‐    the  ini;al  unfamiliarity  of  decision-­‐  and  policy-­‐makers  with  ABM      did  influence  some  of  their  skep;cism  when  assessing  /  evalua;ng  it.  
  16. discussion:  the  ABM  role   Complexity  and  Social  Simula;on  

    -­‐  computa<onal  models  are  formal,  synthesised  representa;ons  of  a  phenomenon   that  is  usually  not  analysed  as  such,  so  that  can  be  difficult  to  discuss  with  stakeholders         -­‐  the  increasing  demand  /  popularity  to  assess  poten<al  outcomes  of  socio-­‐economic   phenomena  is  turning  simula;ons  into  a  form  of  surrogate  reasoning             -­‐  computa;onal  methods  and  techniques  are  oRen  alien  to  the  stakeholders   (i.e.  informing  actual  policy  by  interpre;ng  ABM  results  is  not  straighjorward)  
  17. discussion:  the  ABM  role   Complexity  and  Social  Simula;on  

    -­‐    communica;ng  with  non-­‐experts  the  inner  workings  of  an  ABM  is  difficult    (the  concept  is  grasped  easily,  but  that  is  not  enough  to  inform  policy-­‐making)         -­‐    there  are  major  difficul<es  to  communicate  and  assess  the  actual  usefulness    of  an  ABM  (par;cularly  when  datasets  cannot  be  directly  compared)         -­‐  the  laborious  but  fruijul  endeavour  of  involving  stakeholders  in  the  research    process  can  provide  them  with  a  sense  of  ownership  about  what  is  being  developed   -­‐    tradi;onal  quan;ta;ve  and  qualita;ve  research  methodologies  can    lead  to  useful  descrip;ons  of  an  actual  socio-­‐economic  phenomenon   (so  that  an  ABM  can  complement  these  approaches)  
  18. discussion   Complexity  and  Social  Simula;on    (a)  engaging  stakeholders

     in  the  modelling  process  can  minimise  issues  of    communica;on  and  provide  a  sense  of  ownership  over  the  generated  results;          (b)  lack  of  (or  minimal  frac;onaliza;on)  in  social  preferences  can  impact    posi;vely  in  the  provision  of  public  good  (i.e.  credit  quotas  are  repaid  on  ;me);         -­‐  this  study  has  some  scalability  poten;al,  despite  insights  being  context-­‐dependant     -­‐  a  par;cipatory  approach  can  bridge  the  gap  between  the  stakeholders  and  researchers     -­‐  this  informed  an  actual  policy  change  regula;ng  the  funding  of  over  20.000  credit  clients    
  19. Thanks.   Pablo  Lucas   University  College  Dublin   Complexity

     and  Social  Simula;on