advertisement

Giving is Caring: Understanding Donation Behavior through Email

56 %
44 %
advertisement
Information about Giving is Caring: Understanding Donation Behavior through Email
Technology

Published on February 19, 2014

Author: YelenaMejova

Source: slideshare.net

Description

Every day, thousands of people make donations to humanitarian, political, environmental, and other causes, a large amount of which occur on the Internet. In our paper presented at CSCW'14, we describe a comprehensive large-scale data-driven study of donation behavior. We analyze a two-month anonymized email log from several perspectives motivated by past studies on charitable giving: demographics, user interest, external time-related factors and social network influence. We show that email captures the demographic peculiarities of different interest groups, for instance, predicting demographic distributions found in US 2012 Presidential Election exit polls. Furthermore, we find that people respond to major national events, as well as to solicitations with special promotions, and that social connections are the most important factor in predicting donation behavior.
advertisement

Giving  is  Caring   Understanding  Dona1on  Behavior  through  Email   @   circa  Fall  2012   Yelena  Mejova,  Qatar  Compu1ng  Research  Ins1tute   Ingmar  Weber,  Qatar  Compu1ng  Research  Ins1tute   Venkata  Rama  Kiran  Garimella,  Aalto  University   Michael  C.  Dougal,  University  of  California,  Berkeley   @  Computer  Supported  Coopera1ve  Work  and  Social  Compu1ng  (CSCW14)   February  19,  2014  

Mo1va1on   Impact  real  life!   hTp://www.opensecrets.org/pres12/  

Mo1va1on   1.  Can  we  detect  dona1ons  in  email?   2.  Can  we  verify  sociological  theories  on  charitable  giving?  

What  could  affect  dona1on  behavior?   •  Demographics   –  “individual  capacity”:  educa1on  &  income1   –  belonging  to  a  social  group2   •  Interest   –  seeing  as  more  needy  and  deserving3   •  Solicita1ons   –  influence  of  the  email  deluge4   •  Social  network   –  influence  /  homophily,  poli1cal  affilia1on5   •  External  influence   –  events  triggering  a  new  interest  or  awareness6  

1.  Shier,  M.,  and  Handy,  F.  Understanding  online  donor  behavior:   the  role  of  donor  characteris1cs,  percep1ons  of  the  internet,   website  and  program,  and  influence  from  social  networks.   Interna'onal  Journal  of  Nonprofit  and  Voluntary  Sector  Marke'ng   (2012).     2.  Tajfel,  H.,  and  Turner,  J.  An  integra1ve  theory  of  intergroup   conflict.  The  social  psychology  of  intergroup  rela'ons  33  (1979),   47.     3.  Bekkers,  R.  Who  gives  what  and  when?  a  scenario  study  of   inten1ons  to  give  1me  and  money.  Social  Science  Research  39,  3   (2010),  369–381.     4.  Chan,  M.  The  impact  of  email  on  collec1ve  ac1on:  a  field   applica1on  of  the  side  model.  New  Media  &  Society  12,  8  (2010),   1313–1330.     5.  Bekkers,  R.,  and  Wiepking,  P.  A  literature  review  of  empirical   studies  of  philanthropy.  Nonprofit  and  Voluntary  Sector  Quarterly   40,  5  (2011),  924–973.     6.  Olson,  M.  The  logic  of  collec've  ac'on:  Public  goods  and  the   theory  of  groups,  vol.  124.  Harvard  Univ  Pr,  1965.    

Data   •  Collec1ng  “chari1es”  (total:  480)   –  Scraping  Forbes  and  US  News  top  chari1es  lists   –  Top  100  US  poli1cal  campaign  organiza1ons   –  Chari1es  relevant  to  the  prominent  news  stories  in  the  1me   period  (Wikipedia  Current  Events)   hTp://www.forbes.com/lists/2011/14/200-­‐largest-­‐us-­‐chari1es-­‐11_rank.html   hTp://www.usnews.com/usnews/biztech/chari1es/lists/intl_deve-­‐lopment.htm   hTp://www.fec.gov/data/CommiTeeSummary.do?format=html&elec1on_yr=2012   hTp://en.wikipedia.org/wiki/Portal:Current_events   •  Anonymized  Yahoo!  Mail   –  user  agrees  to  research   –  email  addresses  replaced  with  hashes   –  fields:  from,  to,  1tle   •  July  19  –  September  19,  2012  

Data   •  Emails  from  charity:  matching  from  field   •  Emails  thanking  for  dona1on:  manually  tuned  regex  (86%  assessed   precision)   –  from:  info@barackobama.com    subject:  Thank  you  for  your  dona'on!   •  Manually  categorized  chari1es  which  have  at  least  100  emails  in   dataset:   –  Medical,  Humanitarian,  Poli1cs,  Environmental,  Chris1an/Religious,   Military,  Children,  Public  Broadcas1ng,  Animals,  Internet  

Data   •  Donors  (≈100,000)  –  charity  thanked  them  for  a  dona1on   •  Interested  (≈100,000)  –  got  email  from  a  charity  but  did  not   donate   •  General  (≈10,000)  –  a  random  sample  of  the  rest                                    ≈  1  billion  emails  total   •  Is  the  email  treated  as  bulk  or  spam?    <10%  labeled  as  spam:  cancer.org,  lls.org,  redcross.org    >50%  labeled  as  bulk:  stjude,  dscc.org    >50%  labeled  as  spam:  wikimedia.org   •  In  the  analysis,  we  pay  aTen1on  only  to  emails  which  reach   the  inbox  

Demographics  

Demographics   •  Self-­‐reported  from  user  profiles   –  age,  gender,  zip  code   •  US  Census  to  es1mate     –  %  of  Bachelor  degrees,  median  household  income   age   gender   (male=1)   %  bachelors   median  HHI  

Demographics   age   gender   (male=1)   In  agreement  with  the   US  Presiden1al  Elec1on   exit  polls:   –  Younger   –  Female   –  Less  affluent   …  voters  favor  Obama   hTp://elec1ons.msnbc.msn.com/ns/poli1cs/ 2012/all/president/#exitPoll   %  bachelors   median  HHI  

Interest  

Interest   •  Classify  email  1tles  into  topical  categories  using  manually-­‐compiled   keyword  regexes  (avg  precision:  86.2%):   for  par1cular  topic   PBS   WGBH   incoming   for  par1cular  topic   outgoing  

Solicita1on  

Solicita1on   •  Does  increase  in  solicita1on  prompt  more  dona1ons?   1.  2.  3.  Compute  number  of  dona1ons  per  day  for  each  charity   Divide  into  three  terciles:  low,  medium,  high   Compute  Cohen’s  kappa  with  incoming  mail  from  charity   43%  of  organiza1ons   have  Cohen’s  kappa  >  0.3     That  is,  there  is  a   moderate  to  high   rela1onship  between   solicita1ons  and  dona1on  

Solicita1on   •  Can  we  detect  this  on  a  personal  scale?   –  Compute  probability  that,  given  the  user  will  donate  to  an   organiza1on,  that  he  or  she  donates  within  some  number  of  days  of   receiving  a  solicita1on   –  Compare  to  a  uniform  baseline   On  average,  cumula1ve   probability  of  a  dona1on  a{er   a  solicita1on  is  higher  by  8%   na1onalmssociety.org   wycliffe.org   lls.org   opera1onsmile.org   intervarsity.org   feedthechildren.org   dscc.org   worldvision.org   tbn.org   wikimedia.org   marchofdimes.com   mdausa.org   ronpaul2012.com   greenpeace.org   irusa.org   Medical   Chris1an   Medical   Medical   Chris1an   Humanitarian   Poli1cs   Humanitarian   Chris1an   Internet   Medical   Medical   Poli1cal   Environmental   Humanitarian   48.99   37.98   33.82   29.41   18.78   18.26   15.39   13.88   13.82   12.99   11.82   11.45   8.92   7.89   7.5   increase  in  dona1on  probability  

Social  network  

Social  network   •  Is  there  a  rela1onship  between  your  dona1ons  and  those  of  your  friends?   1.  2.  3.  4.  For  each  donor,  find  other  Yahoo  users  they  exchanged  emails  with   Normalize  number  of  friends  by  sampling  with  replacement  to  get  100   friends   Within  those  friends  we  can  find  other  donors  and  interested  users   Measure  strength  of  rela1onship  by  minimum  emails  sent  or  received   Donors  to  poli1cal  causes  have   almost  no  interac1on  with  others   who  donated  to  the  opposite  side   Spike  in  7th  bucket  (closest  friends)  is   greater  than  buckets  1-­‐6  at  p  <  0.01   each  bucket  contains  the   same  number  of  users  

External  Influence  

Dona(on  Volume   1.2   1   0.8   0.6   0.4   0.2   0   Paul  Ryan  announced  as  VP  candidate   9/19/1 9/17/1 9/15/1 9/13/1 9/11/1 9/19/1 9/17/1 9/15/1 9/13/1 9/11/1 9/9/12   9/7/12   9/5/12   9/3/12   9/1/12   8/30/1 8/28/1 8/26/1 8/24/1 8/22/1 8/20/1 8/18/1 8/16/1 8/14/1 8/12/1 8/10/1 8/8/12   8/6/12   8/4/12   8/2/12   Solicita1ons   1   0.8   0.8   0.6   0.6   0.4   0.4   0.2   0.2   Solicita1ons   0   1.2   1   0.8   0.6   0.4   0.2   0   Solicita(on  Volume   1.2   Solicita(on  Volume   T-­‐shirt  promo  “I  built  this”   9/9/12   9/7/12   9/5/12   9/3/12   9/1/12   8/30/1 8/28/1 8/26/1 8/24/1 8/22/1 “Photo  going  around  on  Facebook”   8/20/1 8/18/1 8/16/1 8/14/1 8/12/1 8/10/1 8/8/12   8/6/12   8/4/12   8/2/12   Dona1ons   7/31/1 7/29/1 7/27/1 7/25/1 Dona1ons   7/31/1 7/29/1 7/27/1 7/25/1 1   7/23/1 7/21/1 1.2   7/23/1 7/21/1 7/19/1 0   7/19/1 Dona(on  Volume   External  Influence   barackobama.com   Democra1c  Na1onal  Conven1on   Republican  Na1onal  Conven1on   miTromney.com  

Pu|ng  it  all  together   •  Sample  data  to  mi1gate  influence  of  chari1es  with  more   representa1on     •  Logis1c  regression  predic1ng  whether  user  donates  to  a  charity   For  each  friend  who  donates  to  a  charity,  the  likelihood  for   the  user  to  also  donate  to  that  charity  goes  up  by  24%   all  coefficients  are  significant  at   p  <  0.01  except  I4  and  S4  

Conclusions   •  Want  to  run  a  campaign?   –  Email  campaigns  can  be  effec1ve  –  solicita1on  works!   –  Tailor  to  your  campaign  to  your  audience   –  Leverage  social  network   –  Don’t  get  stuck  in  spam/bulk  folders   •  What’s  next?   –  Social  influence  or  homophily?   –  How  to  tailor  solicita1ons?   The  Language  that  Gets  People  to  Give:   Phrases  that  Predict  Success  on  Kickstarter   Tanushree  Mitra  |  Eric  Gilbert   [images  by  Wikimedia]  

Add a comment

Related presentations

Presentación que realice en el Evento Nacional de Gobierno Abierto, realizado los ...

In this presentation we will describe our experience developing with a highly dyna...

Presentation to the LITA Forum 7th November 2014 Albuquerque, NM

Un recorrido por los cambios que nos generará el wearabletech en el futuro

Um paralelo entre as novidades & mercado em Wearable Computing e Tecnologias Assis...

Microsoft finally joins the smartwatch and fitness tracker game by introducing the...

Related pages

“Giving is caring: understanding donation behavior through ...

Giving is Caring: Understanding Donation Behavior through Email Yelena Mejova Yahoo Labs Av. Diagonal 177 E-08018 Barcelona, Spain ymejova@yahoo-inc.com
Read more

Giving is caring: understanding donation behavior through ...

Figure 2. Demographics of donors to two political campaigns - "Giving is caring: understanding donation behavior through email"
Read more

Giving is Caring: Understanding donation behavior through ...

Enter the email address you signed up with and we'll email you a reset link.
Read more

Giving is Caring: Understanding donation behavior through ...

Giving is Caring: Understanding donation behavior through email on ResearchGate, the professional network for scientists.
Read more

Online fundraising donations and ideas - JustGiving

Fundraise or donate with JustGiving, ... "Through JustGiving, ... Contains Royal Mail data © Royal Mail copyright and database right ...
Read more

Donate Now & Help Change the World | CARE

Empower Her Through ... Relief Aid with Your Donation Today. Every day, CARE helps people in the poorest ... Donate by Mail; Legacy & Planned Giving;
Read more

Giving Statistics : Charity Navigator

How To Stop Solicitations By Mail; ... and how much funding flows through them each year. ... Where are donations coming from? Giving increased in every ...
Read more

Nonprofit Donation and Church Giving App - Givelify

Givelify is the free app for church giving and ... Givelify is the free app for church giving and nonprofit donations ... the Givelify email ...
Read more

Personal Health Journals for Any Condition | CaringBridge

... giving you time to focus on what matters. ... CaringBridge Stories. Through CaringBridge, ... Email. Join Our Email List ...
Read more