387 American Economic Review: Papers & Proceedings 2013, 103(3): 387–392 http://dx.doi.org/10.1257/aer.103.3.387 Deposit Collecting: Unbundling the Role of Frequency, Salience, and Habit Formation in Generating Savings † By Suresh de Mel, Craig McIntosh, and Christopher Woodruff* * De Mel: Department of Economics, University of Peradeniya, Peradeniya 20400, Sri Lanka (e-mail: demel. suresh@gmail.com); McIntosh: Graduate School of International Relations and Pacifc Studies, University of California-San Diego, La Jolla, CA 92093 (e-mail: ctm- cintosh@ucsd.edu); Woodruff: Department of Economics, University of Warwick, Coventry CV4 7AL, UK (e-mail: c.woodruff@warwick.ac.uk). This project would not be possible without the collaboration of the National Savings Bank, particularly the Bandarawela and Mahiyangana branches, or without funding provided by the Consortium for Financial Services and Poverty (CFSP) and the Center for Competitive Advantage in the Global Economy. We thank Mike Callen for assistance with the project,and Shawn Cole for comments. Kandy Consulting Group carried out all of the surveys and treatment logistics. † To view additional materials, and author disclosure statement(s),visit the article page at http://dx.doi.org/10.1257/aer.103.3.387. Deposit collectors are common in many countries (Ashraf, Karlan, and Yin 2006). Poor households prove willing to pay for services such as Chit collectors in India or Susu collec- tors in Western Africa, indicating the value of illiquid savings to those whose fnancial lives are transacted largely in cash (Rutherford 2000). While small-scale deposit collection can provide a valuable service to households, it is unlikely to create the same multiplier effects as savings that are on-lent through modern fnancial insti- tutions. For this reason, a service that could link deposit collection to the formal banking sector would hold the promise of beneftting both sav- ers and the broader economy through improved intermediation. To examine this, we conducted an experiment in which employees of the National Savings Bank (NSB) worked as deposit collectors for 12 months in semirural areas of Sri Lanka. NSB employed a technological innovation to deliver remote banking services in a highly credible way: a mobile POS device links the agent to the bank’s computer system via the mobile tele- com network, allowing the depositor to receive a receipt providing full account information, including the amount of the latest deposit. Withdrawals can be made only at the bank, pro- viding a form of savings commitment via asym- metric costs of deposit and withdrawal. Weekly home visits by the POS agents resulted in sig- nifcant increases in bank savings and overall savings, as reported in Callen et al. (2012). While frequent visits by POS-enabled bank agents are effective at increasing savings, they are unlikely to be cost effective given the sala- ries and capital costs involved. In this paper we report on a subsequent experiment which altered the collection method to reduce collection costs. The experiment allows us to “unbundle” various attributes of the weekly home POS visit in order to understand which components of the weekly home visit are central to its impact. Theory points us in several directions. First, there is the question of the frequency at which people are prompted to transact (Field and Pande 2008, McIntosh 2008). Next, there is the question of salience; the extent to which an individual is actively prompted to save (Karlan et al. 2010). Finally, frequent face-to-face collection may be effective simply because it inculcates a habit of saving. The unbundling experiment altered the fre- quency of savings by switching some house- holds from weekly to biweekly (fortnightly) collection. We altered the salience of savings by having the bank install a neighborhood sav- ings lockbox rather than sending the deposit collector door-to-door. Only the bank agent has the key to this box, and POS receipts are left behind once the box has been collected on its regular schedule. We switched both control and treatment areas to the box treatment, leaving us with the six treatment combinations illustrated in Figure 1. These are Weekly Home (the origi- nal treatment, which we refer to as “treatment”), Treatment to Biweekly Home (the experiment in decreasing frequency), Treatment to Weekly Box (the experiment in decreasing salience),