Effects of Technology on Economic Growth and Poverty Traps

Research Challenges in
Humanitarian Engineering
Prof. Kevin M Passino
https://hecourse.engineering.osu.edu
Outline:
1. Introduction
2. Financial advisors
3. Poverty traps
4. Cooperative management of community
technology
5. Analysis of sustainable community
development
6. Brief: 5 other challenges
2
Source (free download):
K.M. Passino, “Humanitarian
Engineering: Advancing Technology for
Sustainable Development,” 3rd Edition,
Bede Pub., 2016.
Caveat: I am a control systems researcher
(stability, optimization, game theory)
3
Problem #1: Lack of financial services for
the poor (savings, loans, investments,
insurance)
• Sometimes, “micro-finance institutions”
• But, little help in how to manage money
4
Implementation?
Can we use a cell phone to
provide advice?
• User input: Daily income,
desired wealth
• Recommendation to user:
Spend $X today (and tell
device what actually spent)
Centralized computer for
community in a microfinance
institution?
7.3 Billion mobile
subscriptions in world
in 2015
5
Progress: Feedback control for financial
management
6
Proportional-Integral-Derivative
(PID) Financial Advisor
Matlab/Simulink model
7
Standard engineering approach!
Tuning the PID strategy:
Error, e = desired wealth - current wealth
• Proportional gain: Increase ⇒ Faster,
can cause overshoot
• Integral gain: Increase ⇒ (error → 0) but
can cause overshoot
• Derivative gain: Increase ⇒
predict/reduce overshoot
8
PID spending strategy performance (>0.6 spend)
Sample Results
Have extra money
for emergencies
9
Other Progress:
Gonzalez/Passino, “Feedback
Controllers as Financial Advisors for
Low-Income Individuals,” IEEE Trans.
on Control Systems Technology, 2017.
1. Loans and investments added to
model
2. Dynamic programming
3. Model predictive control
4. Monte Carlo analysis
Poverty
trap
10
Research Challenges:
1. Additional model features (e.g., insurance
against “shocks”)
2. Stability analysis of the control loop
(nonlinear and stochastic)
3. Robustness analysis (to ensure reliability)
4. Other control methods (e.g., adaptive
control)
5. Implementation and scale-up
11
Problem #2: Poverty traps
• Understanding mechanisms (e.g., effects
of technology)
• Breaking traps: Allocating funds to help
pull people out of poverty traps (e.g., how
should the UN/World Bank make
decisions?)
12
Country-level model
Progress: Effects of Technology on
Economic Growth and Poverty Traps
• “production function”, increasing in c(t) (e.g.,
pf(c(t))=pc(t), linear). Why?
• c(t)≥0, is “capital-labor ratio” (money and
equipment/tools per person)
• p is “total factor productivity” (proportional to
quality of technology, p≥0)
• s=national savings rate, g=population growth
rate, d=capital depreciation rate (>0, normal)
13
Economic growth ordinary differential
equation (ODE):
First term positive (increase rate), second
negative (decrease rate). Consider zeros for
parameter values.
⇒
If f(0)=0 then c=0 is an “equilibrium” (stagnant growth)
14
Production functions:
Common view of economic growth:
where a>0. What happens for c(t)=0? Note:
What happens as c(t) increases? Is large?
Things always get better!
15
Low Capital Poverty Trap:
Low amounts of capital get you nothing!
If c=0, f(0)=0, and derivative is zero
Equilibria: c=0, two solutions to (cT and cE):
Quadratic equation
16
Rich get richer, poor get poorer
Examples: s=0.1, g=1, d=0.1, a=0.75,
p=30 (low
tech) and
p=40 (high
tech), c(0)
values?
Poverty Trap
Tech helps!
Stability
Analysis…
17
← Tech →
quality
Technology impact? How to break a poverty trap?
18
Optimization for Economic Models
(example):
Consider how to change p and g (off
nominal, current values)
to improve (reduce) cT
19
0.8
1.5
0.6
1
0.4
0.5
0
1.2
45
40
1
0.8
0.4
30
0.2
25
Technology parameter, p
0
Contours of c T(p,g), white is c T(p,g)=c Td
1.1
991
0.59
1
0.9
0. 5
1989
87
0.439
0.9
0.8
5
0.3598
0.7
0.6
0. 3
0.4
5
0.3598
0.8
0. 3
0.27983
0.6
0.7
0.3
0.19981
0.2
0.1
0.19981
30
32
0.5
0.4
0.3.27983
0
0.5
0.3
28
0.2
35
0.6
Growth parameter, g
Growth parameter, g
Poverty threshold, c T(p,g)
1
34
36
38
Technology parameter, p
40
42
44
0
20
Cost to increase p and/or decrease g
How should development dollars be
invested. In p? In g? In both? What
proportions?
21
Optimization problem: Minimize poverty
trap threshold constrained by a fixed
total amount of spending
22
Graphical solution
1
0.9
Contours of c T(p,g) and cost constraint (white)
1
79 2119 446 74 01
3
9
7
5
0. 0.5 0.50 0.48 .471 4542 3756 084 411
0
0. 0.4 0.42 .40
0
Growth parameter, g
0.8
0.7
0.6
0.5
0.4
0.3
30
9
66
873 .370 35394
3
.
0
0
0.
2
372
0.3
9
204
0.3
7
0 37
3
.
0
704
0.28
032
0.27
3 59
0.25
687
0.23
014
0.22
342
0.20
69
0 . 1 86
97
0.169
9
6
873
4
3
706
.
3
0
.
0
539
3
.
0
2
372
3
.
0
9
204
3
.
0
7
037
3
.
0
704
0.28
032
0.27
359
0.25
0.6
6 87
0.23
014
0.22
0.3
0.5
0.4
342
0.20
0.186
0.2
69
97
0.169
0.1
24
0.153
2
0.1365
35
40
Technology parameter, p
45
0
23
Research Challenges:
1. Better economic models
2. Individual-level models
3. Large-scale models
4. Sensitivity analysis
5. Other optimization problems
24
Problem #3: Cooperative management of
community technology
1. “Management”=operation+maintenance
2. Too often projects fail due to poor
operation and maintenance
25
Community technology
examples:
• Water pump
• Energy sources (e.g.,
biodigester)
• Sanitation service (e.g.,
toilets)
• Cell phone charging station
(multiple outlets)
Any technology that is shared
26
2
1
3
Resource 1
(cell phone
charging station)
Example:
• 5 people
• 3 technology
“resources”
Resource 2
(water pump)
4
Resource 3
(sanitation,
toilets)
5
27
Automated/Semi-Automated
Management of Community Technology:
• Electromechanical $ collection devices at
each j
• Display price, gather money
• Data tracking: Payments, use
levels/patterns, monitoring from a
distance (to study effectiveness)
• A network?
• Automation costs, up-front/on-going
28
Study of automation could be useful (pricing)
Semi-Automated: People+Technology
• People attend, collect money, get paid
• Prices set by the computer (to cope with
complex pricing issues)
• If person not there (e.g. at night), go to
full-automated
• Could cope with inequality, via pricing
strategies (e.g., with community input)
29
Feedback Control for Community
Technology Management: Example
• M=1 technology resource, N=3 people
• Notation: ri(k), pi(k)
• ri(k) are constants of 5, 7, and 8 with
uniformly distributed noise added on from
[-2,2]
• cm(k), maintenance costs at k
• m(k), stored maintenance money
30
Past amount, minus expenditures, plus total
gathered (modified)
Amount paid to attendant
31
Desired stored maintenance money, md(k)
Set as ramp/constant
Let cm(200)=50
Pricing signal
Nonlinear proportional feedback control
Tries to adjust prices to make m(k) the same
as md(k) (copes with random failures)
32
Allocate
Pricing
Choice of gi
• Equality, gi=1/N
• Inequality: Community picks the gi
33
Resource demands
Prices, price signal
0.3
0.25
p i, p (dashed)
8
6
4
2
0
0.15
0.1
0.05
0
100
200
Money for each i, total gathered (black)
2
1.5
1
0.5
0
100
200
Time step, k
0
300
2.5
0
0.2
300
m d (k) (red), m(k) (blue), cm (o)
r i, i=1,2,3
Equality
gi=1/3,
i=1,2,3
0
100
200
300
Regulation of maintenace funds
120
100
80
60
40
20
0
0
100
200
300
Time step, k
34
Research Challenges:
1. Stability/convergence of the control loops
(nonlinear stochastic problem)
2. Fully distributed/networked case, stability
3. Network optimization approach?
4. Deployment
35
Problem #4: Analysis of sustainable
community development
1. Can we show that humanitarian
technologies will work in a community
before deployment?
2. Can we predict their effects (e.g., on the
environment)?
36
Wealth dynamics (N random communities)
Why a product?
Economics
Health
Mincer earning function
Extend ecology
37
Health, education, resource dynamics
Utilization proportional to spending
Recall earlier ecological models
A resource common to the community
(e.g., farm, forest, or fishery)
38
Representing technologies (each
highlighted above):
• To make money: p
• To improve health: bh
• To improve education: be
• To make resource use efficient: bu
Each technology is an “amplifier” or an
“attenuator” (i.e., it is a gain)
39
How do we measure how well a community
is doing? Sustainable community
development index (SCDI)
SCDIw: Total wealth
(discounts per inequalities, like UN
Inequality-Adjusted Human Development
Index)
40
Effects of technology, p
Consider
effects of
technology
failures and
quality of
technologies
41
Research Challenges:
1. Better models of community, validated
with data
2. Stability/optimality analysis
3. Our current work: Embed “savings club”
vs donation strategies in community and
show that it pulls people out of a poverty
trap
42
Other problems are in the book:
“Humanitarian Engineering: Advancing
Technology for Sustainable
Development”
1. Analysis of (a) sustainability, (b) wealth
distribution, (c) democracy
2. Technology diffusion and poverty traps
3. Participation (e.g., Giraldo/Passino,
Dynamics of Cooperation in a Task
Completion Social Dilemma,” PLOS
ONE, 2017.)
43
Summary:
1. Here: Financial advisors, poverty traps,
CMCT, sustainable community
development
2. Research directed toward (i) engineering
applications or (ii) systems analysis
3. Challenging/interesting mathematical and
computational research problems
4. Ones whose solution really matters!
44