Assessing Impacts as Changes in Economic Output

1
Assessing Impacts as Changes in Economic Output

• Anand Patwardhan
• Upasna Sharma

2
Stock Vs. Flow

• Conventionally impacts of cyclones (or other
  climate hazards) measured as changes in stocks of
  human, social and economic capital.
• Alternatively, they may be measured as changes in
  flow of goods and services, (typically economic
  output).

3
Motivation for assessing impacts in terms of
changes in flow variables

• Provides information about length of recovery
  period
• Relief vs. recovery debate in disaster
  mitigation.
• Could be used to establish the validity or
  explanatory power of measures of adaptive
  capacity (generic or specific adaptation)
• Distinguish between loss of capital assets vs.
  loss of income for affected communities

4
Natural Hazard Tropical Cyclones

• Why tropical cyclones is a good starting point
  for exploring the concept
• Bounded in space and time (unlike droughts).
• Impact large enough to disrupt economic activity
  over an area, and for a duration long enough that
  it may be resolved / detected

5
What do we expect the data to reveal?

• We expect a drop in the output of the affected
  economic sector as a consequence of the event in
  relation to the baseline
• Event year output should differ from non-event
  year output
• We expect to observe gradual recovery in the
  period following the event
• Confounding factors
• Secular change
• Other, non-event related disturbances
• Variability (signal to noise ratio problem)

6
Output variables being studied
7
Linkage between spatial extent and administrative
units for reporting of data

• Considerations for selecting spatial unit/scale
  of analysis
• The scale of hazard and its impact
• The scale at which the socio-economic data is
  reported

Hierarchy of Administrative Units in India
8
Descriptive statistics for paddy output
9
Plots of Time Series of Paddy Output for Some
Districts
Balasore
Tanjavur
East Godavari
Nellore
10
Secular Trend in the Data
All values are significant at 1 level of
significance
11
Assessing Impact on Agricultural Output

• Grouped districts into 2 categories based on the
  number of cyclonic events that occurred during
  the study period

Districts with few events.
Districts with many events
12
Districts with many events

• Existence of secular trend in the agricultural
  output data required de-trending the data
• Classified the residuals into those for event
  years and those for non-event years
• Examined the difference in these two populations
  using the non-parametric Kolmogorov-Smirnov (K-S)
  test

13
Results Districts With Many Events
denotes ? 0.01 (one-tailed)
denotes ? 0.05 (one-tailed) denotes ?
0.10 (one-tailed) ? Level of significance
14
Districts with few events
Compared residual in event year, with residuals
of baseline years (after accounting for secular
trend) by using the Students t-test. The
hypothesis can be specified as Ho Re
Rne Ha Re lt Rne Where Re denotes the
residuals for the event years and Rne denotes
mean of residuals for the non event years
15
Results Districts With Few Events
Figures in bold are statistically significant at
5 level of significance.
16
Alternative approach for districts with many
events

• To account for secular trends, earlier approach
  focused on residuals after detrending the data
• Alternatively, we can look at year to year
  changes in output. Four kinds of changes in
  output are possible
• Non-event to event change in output,
• Event to non-event change in output,
• Non-event to non-event change in output and
• Event to event change in output.
• We can postulate certain expectations regarding
  these changes in output
• Non-event to event change in output to be
  negative,
• Event to non-event change in output to be
  positive,
• Non-event to non-event and event to event change
  in output could either be positive or negative
• If there is a statistically significant increase
  / recovery in output after the event year, then
  it provides for a more robust basis for
  attributing the decrease in output in the event
  year to a cyclone.

17
Year to Year Changes in Agricultural Output for
Nellore District
18
Statistical Technique Used

• To test whether there is a statistically
  significant difference between the means of the
  three types of changes in output, we use Analysis
  of Variance (ANOVA) technique.
• Analysis of Variance (ANOVA)
• The hypothesis can be specified as
• Ho µne-e µe-ne µne-ne
• Ha µne-e, µe-ne, µne-ne are not all equal.
• Where µne-e denotes the mean of non-event to
  event changes in output
• µe-ne, denotes the mean of event to non-event
  changes in output
• µne-ne denotes the mean of non-event to non-event
  changes in output

19
Results of ANOVA for effect of cyclones
on agricultural output
denotes ? 0.01 denotes ?
0.05 denotes ? 0.10
20
Observations

• Impacts of a cyclone can be measured in terms of
  flow of goods and services of the affected
  socio-economic sectors, if appropriate spatial
  and temporal resolution is chosen.
• This approach can provide new impact metric for
  linking generic adaptive capacity to observable
  impacts at the ground level.
• This not only a relatively low cost methodology
  but also uses in a different manner the data for
  which well established reporting mechanisms
  already exist.
• Assessment of impacts through a different route
  which could act as a check on biases and errors
  of the conventional impact assessment methods.
• This methodology can be replicated for natural
  disasters other than cyclones (for instance,
  floods, earthquakes etc.).

21
Work in Progress

• Extend the work - Fisheries output (fish catch),
  Electricity Consumption (much higher temporal
  resolution then the agricultural data)

Impact of Cyclone on Fisheries Output
Chengalpattu District
Cyclonic event Nov-85
Recovery
Impact