Saving Lives versus Saving Livelihoods: Can Big Data Technology - - PowerPoint PPT Presentation

▶
saving lives versus saving livelihoods can big data
SMART_READER_LITE
LIVE PREVIEW

Saving Lives versus Saving Livelihoods: Can Big Data Technology - - PowerPoint PPT Presentation

Sep 30, 2023 240 likes •614 views

Saving Lives versus Saving Livelihoods: Can Big Data Technology Solve the Pandemic Dilemma? Kairong Xiao Columbia Business School COVID-19 and Economics: China, Asia and Beyond May 1, 2020 Xiao Can Big Data Technology Solve the Pandemic

slide-1
SLIDE 1

Saving Lives versus Saving Livelihoods: Can Big Data Technology Solve the Pandemic Dilemma?

Kairong Xiao

Columbia Business School

COVID-19 and Economics: China, Asia and Beyond May 1, 2020

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 1

slide-2
SLIDE 2

Motivation

COVID-19: an impossible choice between saving lives and saving livelihoods Population movement restrictions are deemed necessary to contain pandemics But such restrictions inflict steep economic costs U.S. GDP is foretasted to decline at a 37% annual rate from April to June (WSJ, April 29)

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 2

slide-3
SLIDE 3

How do we solve the pandemic dilemma?

Table 1: Top 10 Most Downloaded Contact-Tracing Apps

Country App Name Downloads India Aarogya Setu 50M Czech Republic Mapy.cz 1M Colombia CoronApp 1M South Korea Corona 100m 1M Israel The Shield 1M Singapore TraceTogether 0.5M India (Punjab) COVA Punjab 0.5M Spain (Catalonia) STOP COVID19 CAT 0.5M Norway Infection Stop 0.1M

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 3

slide-4
SLIDE 4

Big Data Technology

Advocates

◮ Detect potential carriers and allow the mass population to resume work ◮ Successful experience in China and South Korea

Critics

◮ Inconclusive evidence: unsuccessful experience in Singapore ◮ Privacy infringement and government surveillance Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 4

slide-5
SLIDE 5

This paper

Exploit the staggered adoption of contact-tracing apps in 322 Chinese cities Use high-frequency measures of economic activities

◮ Within-city population movements ◮ Emission of greenhouse gas Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 5

slide-6
SLIDE 6

Findings

Cities that adopt contact-tracing apps experience a significant increase in economic activities without suffering from higher infection rates Contact-tracing apps create an economic value of 0.5%-0.75% of GDP during the COVID-19 outbreak The economic benefits seem to outweigh the cost of privacy

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 6

slide-7
SLIDE 7

Institutional Background

slide-8
SLIDE 8

Contact-Tracing Apps: Health Code

Green: no restriction Yellow: isolation for 7 days (then it turns green) Red: isolation for 14 days (then it turns green)

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 7

slide-9
SLIDE 9

Contact-Tracing Apps: Health Code

First implemented in Hangzhou on Feb 11, 2020 Implemented by other cities in a staggered manner The implementation is uncoordinated by the central government Cities often have different versions of health code

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 8

slide-10
SLIDE 10

Data

slide-11
SLIDE 11

Data

Implementation dates of health code for 322 cities in China Within-city population movements Greenhouse gas level of each city Daily COVID-19 infection counts for 322 cities in China

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 9

slide-12
SLIDE 12

Adoption of Health Code in Chinese Cities

Wuhan Lockdown Hangzhou Health Code 100 200 300 Cummulative # of cities adopting health code 10 20 30 # of cities adopting health code 01jan2020 01feb2020 01mar2020 01apr2020 Date

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 10

slide-13
SLIDE 13

Adoption of Health Code in Chinese Cities

Date Feb 15 Feb 29 Mar 15 Mar 31 Data not available

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 11

slide-14
SLIDE 14

Data

Implementation dates of health code for 322 cities in China Within-city population movements Greenhouse gas level of each city Daily COVID-19 infection counts for 322 cities in China

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 12

slide-15
SLIDE 15

Economic Activities of Chinese Cities

Wuhan Lockdown Hangzhou Health Code 20 40 60 80 100 120 Economic activities 01jan2020 01feb2020 01mar2020 01apr2020 Date

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 13

slide-16
SLIDE 16

Data

Implementation dates of health code for 322 cities in China Within-city population movements Greenhouse gas level of each city Daily COVID-19 infection counts for 322 cities in China

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 14

slide-17
SLIDE 17

Nitrogen Dioxide Level of Chinese Cities

Wuhan Lockdown Hangzhou Health Code 20 40 60 80 100 NO2 01jan2020 01feb2020 01mar2020 01apr2020 Date

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 15

slide-18
SLIDE 18

Nitrogen Dioxide Level of Chinese Cities from NASA

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 16

slide-19
SLIDE 19

Data

Implementation dates of health code for 322 cities in China Within-city population movements Greenhouse gas level of each city Daily COVID-19 infection counts for 322 cities in China

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 17

slide-20
SLIDE 20

Nitrogen Dioxide Level of Chinese Cities

Wuhan Lockdown Hangzhou Health Code 20000 40000 60000 80000 # of cases 01jan2020 01feb2020 01mar2020 01apr2020 Date Confirmed Cured Deaths Active

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 18

slide-21
SLIDE 21

Summary Statistics

N mean sd p5 p25 p50 p75 p95 Within-city movements 28658 78 26 32 56 84 99 109 NO2 24742 63 30 23 40 58 81 119 PM2.5 24742 78 51 20 43 68 102 171 Infection rate 28658 2 5 20 Confirmed cases 28658 144 1986 8 31 213 Cured cases 28658 83 1252 3 18 140 Deaths 28658 5 91 3 Emergency level 28658 2 1 2 3 3

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 19

slide-22
SLIDE 22

Results

slide-23
SLIDE 23

Effect of Health Code on Within-city Movement

Regression model EconomicActivityi,t = βHealthCodei,t + γXi,t + ǫi,t

(1) (2) (3) (4) Movement Movement Movement Movement Health code 2.859∗∗∗ 2.687∗∗∗ 2.552∗∗∗ 3.118∗∗∗ [0.410] [0.345] [0.437] [0.399] Control Yes Yes Yes Yes City F.E. Yes Yes Yes Yes Time F.E. Yes Yes Yes Yes Emergency F.E. Yes Yes Yes Yes Sample Full sample

  • Excl. Hubei

Match by cases Match by act. Observations 28,658 27,145 28,658 28,658

  • Adj. R-squared

0.852 0.862 0.867 0.852

The introduction of health code leads to around 2-3% increase in within-city movement.

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 20

slide-24
SLIDE 24

Dynamic Effect of Health Code on Economic Activities

  • 2

2 4 6

Treatment effect

  • 20-19-18-17-16-15-14-13-12-11-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Time since treatment

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 21

slide-25
SLIDE 25

Effect of Health Code on Greenhouse Gas

Regression model EconomicActivityi,t = βHealthCodei,t + γXi,t + ǫi,t

(1) (2) (3) (4) NO2 NO2 NO2 NO2 Health code 1.792∗ 1.973∗ 1.417 1.980∗∗ [1.006] [1.057] [1.106] [0.990] Control Yes Yes Yes Yes City F.E. Yes Yes Yes Yes Time F.E. Yes Yes Yes Yes Emergency F.E. Yes Yes Yes Yes Sample Full sample

  • Excl. Hubei

Match by cases Match by act. Observations 24,742 23,674 24,742 24,742

  • Adj. R-squared

0.541 0.534 0.555 0.541

The introduction of health code leads to around 2% increase in NO2 level.

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 22

slide-26
SLIDE 26

Effect of Health Code on Greenhouse Gas

Regression model EconomicActivityi,t = βHealthCodei,t + γXi,t + ǫi,t

(1) (2) (3) (4) PM2.5 PM2.5 PM2.5 PM2.5 Health code 4.989∗∗∗ 4.830∗∗∗ 4.514∗ 5.152∗∗∗ [1.671] [1.767] [2.287] [1.626] Control Yes Yes Yes Yes City F.E. Yes Yes Yes Yes Time F.E. Yes Yes Yes Yes Emergency F.E. Yes Yes Yes Yes Sample Full sample

  • Excl. Hubei

Match by cases Match by act. Observations 24,742 23,674 24,742 24,742

  • Adj. R-squared

0.358 0.352 0.378 0.360

The introduction of health code leads to around 4% increase in PM2.5 level.

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 23

slide-27
SLIDE 27

Effect of Health Code on Between-city Migration

Regression model Inflowi,j,t = βDestinationHealthCodej,t + γXi,j,t + ǫi,t,

(1) (2) (3) (4) Inflow Inflow Inflow Inflow Health Code (destn) 12.694∗∗∗ 13.185∗∗∗ 11.183∗∗∗ 12.557∗∗∗ [1.700] [1.696] [1.709] [1.687] Control Yes Yes Yes Yes City pair F.E. Yes Yes Yes Yes Source-time F.E. Yes Yes Yes Yes Emergency level F.E. Yes Yes Yes Yes Sample Full sample

  • Excl. Hubei

Match by cases Match by act. Observations 1,888,652 1,798,439 1,888,652 1,888,652

  • Adj. R-squared

0.857 0.859 0.863 0.858

The introduction of health code increases inflows to a city.

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 24

slide-28
SLIDE 28

Effect of Health Code on Between-city Migration: Inflows

Ningbo Hangzhou (with health code) Shanghai (w/o health code)

13%↑

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 25

slide-29
SLIDE 29

Effect of Health Code on Between-city Migration

Regression model Outflowi,j,t = βSourceHealthCodei,t + γXi,j,t + ǫi,t

(1) (2) (3) (4) Outflow Outflow Outflow Outflow Health Code (source)

  • 14.840∗∗∗
  • 14.647∗∗∗
  • 13.681∗∗∗
  • 14.549∗∗∗

[1.778] [1.744] [1.612] [1.749] Control Yes Yes Yes Yes City pair F.E. Yes Yes Yes Yes Destination-time F.E. Yes Yes Yes Yes Emergency level F.E. Yes Yes Yes Yes Sample Full sample

  • Excl. Hubei

Match by cases Match by act. Observations 1,887,544 1,834,631 1,887,544 1,887,544

  • Adj. R-squared

0.860 0.862 0.861 0.860

The introduction of health code in the source city decrease flows from a city.

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 26

slide-30
SLIDE 30

Effect of Health Code on Between-city Migration: Outflows

Ningbo Hangzhou (with health code) Shanghai (w/o health code)

15%↓

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 27

slide-31
SLIDE 31

Effect of Health Code on Infection Rates

Regression model InfectionRatei,t+7 = βHealthCodei,t + γXi,t + ǫi,t

(1) (2) (3) (4) Infection rate Infection rate Infection rate Infection rate Health code 0.024 0.028

  • 0.000

0.027 [0.078] [0.081] [0.092] [0.082] Control Yes Yes Yes Yes City F.E. Yes Yes Yes Yes Time F.E. Yes Yes Yes Yes Emergency F.E. Yes Yes Yes Yes Sample Full sample

  • Excl. Hubei

Match by cases Match by act. Observations 26,404 25,010 26,404 26,404

  • Adj. R-squared

0.411 0.388 0.412 0.408

The introduction of health code does not lead to a significant increase in infection rates.

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 28

slide-32
SLIDE 32

Dynamic Effect of Health Code on Infection Rates

  • 2
  • 1

1 2

Treatment effect

  • 20-19-18-17-16-15-14-13-12-11-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Time since treatment

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 29

slide-33
SLIDE 33

Saving Lives vs. Saving livelihood

Regression model InfectionRatei,t+7 = β1HealthCodei,t × EconomicActivityi,t + γXi,t + ǫi,t

1 1.05 1.1 1.15 Infection rate 10 20 30 40 50 60 70 80 90 100 Economic activities With Health Code Without Health Code

Without health code, economic activities have to decrease by 19% to reduce the daily infection rate by 1% from its average level. With health code, economic activities only need to decrease by 17% to achieve the same amount of reduction in infection rate.

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 30

slide-34
SLIDE 34

Do Benefits Justify Costs?

2-3% increase in economic activities The outbreak lasted for a quarter in China Health code creates an economic value of 0.5%-0.75% GDP Chinese GDP per capita: $10,000 Economic value per person: $50 − $75 Value of privacy: $33 (Huang, 2019) Caveat 1: culture differences in value of privacy (Athey et al, 2017) Caveat 2: data anonymization

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 31

slide-35
SLIDE 35

Why Is Big Data Effective in Fighting Pandemics?

The key amplification mechanism: incomplete information Because of the hidden virus, people are afraid of going out, which brings economy actitivities to a standstill Governments have to impose quarantines on the whole population just to stop a few hidden carriers Big data technology can effectively alleviate information frictions

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 32

slide-36
SLIDE 36

Conclusion

Cities adopt contact-tracing apps experience a significant increase in economic activities without suffering from higher infection rates Contact-tracing apps create an economic value of 0.5%-0.75% of GDP during the COVID-19 outbreak The economic benefits seem to outweigh the cost of privacy

Xiao Can Big Data Technology Solve the Pandemic Dilemma? May 1, 2020 33