Cost Realism in Deploying Technologies for Development

1
Cost Realism in Deploying Technologies for
Development

• Aishwarya Lakshmi Ratan (Microsoft Research
  India)
• Mahesh Gogineni (London School of Economics)
• May 30, 2008

Confronting the Challenge of Technology for
Development Experiences from the
BRICS Department of International Development,
University of Oxford
TEM, Microsoft Research India
2
Outline

• Problem area
• Analytical framework
• Hypotheses
• Case studies
• Implications
• Takeaways
• Limitations

TEM, Microsoft Research India
3
Problem area

• At 24-36 per annum, microcredit is cheaper than
  moneylender-loans, but not cheap enough.
• High transaction costs drive up the price (even
  the best MFIs have operating cost/asset ratios
  gt10)
• Can technology deliver cost savings through
  efficiency gains?
• If yes, will this drive down the price of credit
  and expand outreach?
• Examine client-facing information collection and
  processing transaction tasks in microfinance
  workflows

TEM, Microsoft Research India
4
Analytical framework
Cost for a given transaction t is described by a
cost function C (Vl, Vk, O, L, F, N) Where Vl w
(wage or labour cost per unit time) A (inverse
productivity indicator or no. of time units per
transaction t) Vk Variable capital cost per
transaction t O Operating costs per unit labour
for transaction task t L Total labour hired for
transaction task t F Fixed costs for
transaction task t N Number of transactions of
task t Measure efficiency through input cost
minimisation Examine the relative cost accrued
for task t under alternate arrangements, LT
(low-tech, baseline channel)and HT (high-tech)
TEM, Microsoft Research India
5
Analytical framework
Per-transaction gains from using a HT option G
(Vl Vk)LT - (Vl Vk)HT G w (ALT - AHT)
Vk,LT - Vk,HT Total gains across all t
transactions TG GN (w (ALT - AHT) Vk,LT -
Vk,HT ) N OG (OLTLLT) - (OHTLHT) OG L
(OLT - OHT) Profit through cost savings p
TGOG RoI p / FLT FHT
Where ? is the inflation rate and d is the
opportunity cost of capital
TEM, Microsoft Research India
6
Hypotheses
Hypothesis 1

• Cost savings from a HT channel maximised when TG
  is maximised, requiring
• High wage rate (w)
• High productivity differential (ALT AHT )
• High variable cost differential (Vk,LT - Vk,HT )
• Large number of transactions t per unit time (N)

Hypothesis 2

• Cost savings from a HT channel are maximised when
  OG is maximised. Since OHT will typically be
  higher than OLT , this requires
• Low operating cost differential (OLT - OHT)

Hypothesis 3
For a given level of net profit through cost
savings, the larger the fixed cost differential
(FLT FHT ), the lower the likely financial
sustainability of the HT channel.
TEM, Microsoft Research India
7
Cases

• Examine 3 microfinance institutions (MFIs) in
  southern India
• Chosen for
• the spread they offered as small, medium and
  large organizations,
• their choice of varying microfinance operating
  models (Joint Liability Groups vs. Self-Help
  Groups), and
• their experimentation with ICTs in client-facing
  information collection and processing tasks.
• Follow Kumar (2004) in an activity-based costing
  approach
• Two kinds of microfinance data examined Loan
  Customer Acquisition and Loan Installment
  Processing
• LT channel PC-based back-end paper-based
  front-end
• HT channel PC-based back-end electronic
  device based front-end
• In the cases examined, cash transport costs
  equivalent across LT and HT

TEM, Microsoft Research India
8
Cases
1
2
3
Installment Processing Rural NBFC Streamline
installment data collection Handheld device used
by field officer Data uploaded through USB Cut
variable cost by 73 Positive NPV over 6 years
Customer Acquisition Urban NBFC Improve
efficiency of customer acquisition process Smart
phone used by field officer Data sent via SMS or
GPRS Cut variable cost by 50 Positive
RoI Negative NPV over 6 years
Installment Processing Rural SHG Fed Streamline
book-keeping and installment data
collection Smart phone used by field officer Data
sent via SMS Little reduction in variable
costs High fixed costs for HT channel Negative
NPV over 6 years
TEM, Microsoft Research India
9
Cost savings comparison
Case 1 Case 2 Case 3
Labour component 1(front-end) Labour component 1(front-end) Labour component 1(front-end) Labour component 1(front-end)
w1 (per month, Rs.) 4500 800 4500
A1,LT - A1,HT (Productivity gain in minutes) 19-19 0 2-2 0 3.46 - 0.81 2.65
Productivity gain 1 (share of baseline) 0 0 0.76
Labour component 2 (back-end) Labour component 2 (back-end) Labour component 2 (back-end) Labour component 2 (back-end)
w2 (per month, Rs.) 4000 800 6000
A2,LT - A2,HT (Productivity gain in minutes) 14-0 14 2-1.75 0.25 0.6 - 0.033 0.57
Productivity gain 2 (share of baseline) 1 0.125 0.95
Vk,LT - Vk,HT (Rs.) 5.65-0.05 5.6 0 0.78 - 0.33 0.45
Variable capital cost reduction (share of baseline) 0.99 0 0.58
G (Rs.) 10.97 0.019 1.92
N (number of task t transactions per yr per branch) 2400 22,992 64,800
TG (per yr per branch, Rs.) 26,304 437 124,416
OLT OHT (per L, Rs.) -2030 -1680 -970
L (per branch) 8 8 12
OG (per yr per branch, Rs.) -16,240 -13,440 -11,640
p (cost savings, Rs.) 10,064 -13,003 112,776
FLT - FHT (per branch, Rs.) -98,462 -68,000 -138,000
NPV of p over 6 years lt0 lt0 gt0
(PV of p over 6 years) / FLT - FHT 0.5 lt0 3
TEM, Microsoft Research India
10
Implications

• The higher the labour productivity gains from the
  HT channel, the greater the transaction cost
  savings.
• The higher the local wage for the task, the
  higher the productivity-linked transaction cost
  savings.
• The higher the variable capital cost reduction,
  the greater the transaction cost savings.
• A larger number of transactions per unit of
  labour/per device greatly multiplies the power of
  productivity gains per transaction from the use
  of the HT channel.
• The larger the operating costs required to run
  the HT channel (e.g. connectivity costs), the
  lower the gains from overall cost reduction.
• The higher the fixed capital investments called
  for in the HT channel, the more substantial the
  requirements for high transactional cost savings
  and low operating cost differentials to ensure
  the HT channels financial sustainability.

TEM, Microsoft Research India
11
Is tech always cost-efficient?

• Classic banking, US

• Microcredit, India

22,000/yr
200
700
240/yr
12.5 productivity improvement 30
12.5 productivity improvement 2750
Parallels in PDA-based data collection for
healthcare, telecentres for accessing lean data,
individual computers for education, etc.
TEM, Microsoft Research India
12
Takeaways

• Can technology deliver cost savings through
  efficiency gains?
• Yes, but conditional on the cost context
• the local wage rate for adequately skilled
  labour
• labour productivity,
• variable capital costs,
• fixed and operating costs per channel,
• scale of transactions per device.
• Among our cases, the successful deployment of a
  HT channel involved
• - operating in an environment of high wages,
• - achieving high improvements in labour
  productivity,
• - reducing variable capital costs significantly,
  
• - having low operating costs for the HT channel,
• - processing a large number of transactions per
  device.

TEM, Microsoft Research India
13
Limitations and extensions

• ICTs dont always trump paper!
• The view that though capital-intensive,
  electronic ICTs, with their speed and accuracy,
  will automatically trump labour-intensive
  paper-based ways of fulfilling the same
  functionalities, is mistaken.
• Extending this work will involve
• Adopting a dynamic, not static lens
• Specifying a functional form to calculate
  allocative and productive efficiency, returns to
  scale
• Endogenising variables w, N, L
• Assessing costs with bundling of transaction
  tasks per channel (data and payments across
  products)
• Incorporating data quality gains/error
  correction costs across channels
• Testing against a larger sample of data points

TEM, Microsoft Research India
14
Acknowledgements Ujjivan, CCD/Ekgaon, BASIX,
Shabnam Aggarwal, Angelin Baskaran, Kentaro
Toyama, Rajesh Veeraraghavan, Rahul
DeExcel-based costing template available at
http//research.microsoft.com/aratan/costing.htm
? aratan_at_microsoft.com
TEM, Microsoft Research India