Measuring people's economic resilience to natural disasters


As temperatures rise and oceans warm, regions will face drastic changes and will be increasingly affected by climate hazards.

The analysis of financial transaction data

can provide critical insights into understanding the economic resilience of people affected by natural disasters. This is the story of a research project from United Nations Global Pulse, BBVA Bancomer and BBVA Data & Analytics, the centre of excellence on data science of the financial institution BBVA.

Economic activity
5,000 MX 7,500 MX 10,000 MX

In this project, we analyzed Sale payments (PoS) and ATM cash withdrawal data

from more than 100,000 BBVA Bancomer clients, the main financial institution in Mexico, totaling a number of 25,000 daily transactions.

In Mexico, 50% of the population has a bank account. The data analyzed in this study represents approximately 30% of all bank account holders in the country.


The first major, and most destructive, hurricane to hit Baja California Sur in 25 years.

Wind speed

Odile made landfall in September 2014 near Cabo San Lucas as a large category 3 hurricane, with a windspeed of approximately 110kt
(> 200km/h)


Resilience /rɪˈzɪlɪəns/
-The capacity of individuals, communities and systems to adapt and survive in the face of stress and shocks.-

The resilience of populations affected by Hurricane Odile was measured by analyzing economic activity before, during and after the disaster.

Data analytics were employed to derive proxy indicators of the economic impact and market resilience of people in the region.


A Normality Model

was built to estimate what the economic activity in Baja California Sur would be under normal conditions based on the activity of other Mexican regions that were not affected by the hurricane . Then, we compared this normal projection with what actually happened.

Hurricane's Landfall

Overall, the average recovery time was observed to be around 2 weeks. When broken down further by location, recovery times varied from 2 days (possibly due to businesses temporarily closing) to more than 1 month (for the towns located on the south coast where the hurricane struck with its highest intensity). The economic recovery time was also compared with maximum sustained windspeed measurements. Higher maximum sustained windspeeds corresponded with slower recovery times, a pattern possibly due to more damage occurring in these areas. See our publication for more details.

expected vs real


derived from this data could be used to target emergency response and reconstruction programmes and policies.

The following figures show some of the findings this study revealed.

Impact on spending behaviour

Total expenditure on items such as food and gasoline in the days immediately before the hurricane struck increased by 50%. The mean amount per transaction did not change, implying that the increase in the number of transactions overall was due to more people preparing for the event. In addition, higher income groups spent more in total – in proportion to their income level – in the days before the hurricane struck. Analysis of recovery time by income level revealed a tendency for people with lower income to recover faster.

Level of Income per year
Low [< 120,000 MXN]
Medium [120,000 MXN - 250,000 MXN]
High [> 250,000 MXN]

Click on the filters below.

Low Medium High     |    
Male Female

Preparation /ˌprɛpəˈreɪʃ(ə)n/
We define preparation as something done to get ready for an event or undertaking.

In an effort to measure community preparation the amount of transactions
conducted the day before the hurricane was analyzed.

The day before the hurricane’s arrival, the spending patterns shifted: transactions in categories such as food or gasoline increased (+20%), suggesting that people were stocking up on food and fuel, while the number of transactions on less essential categories dropped (-10%).

Recovery /rɪˈkʌvə/
We define recovery time as the time needed for a community to return to baseline activity levels.

In the aftermath of the hurricane, economic activity decreased across the region. It took two weeks for PoS and one week for ATM transactions to bounce back to normal.

In the 30 days after the impact, there were 30% fewer POS transactions and 12% fewer cash withdrawals than during a normal 30-day period.

The analysis of financial transaction data

can provide new, real-time insights into the economic resilience of people affected by natural disasters. We believe such insights it could be used to support the design of recovery programmes and policies.

For instance, the following visualization displays Baja California Sur’s broader economic context and differences in preparation and recovery patterns of population groups segmented according to their income

Preparation & Recovery

Our study also revealed that the increase in women's expenditure in preparation for the event was double that of men. Also, women took slightly longer to return to baseline levels than men.

How to
read it

Low income

Less than 120,000 MXN per year
50% of the overall population

Our analysis shows that the lower the income level of a community, the shorter the time needed for economic activity to return to normal levels.

Medium income

Between 120,000 and 250,000 MXN per year
30% of the overall population

Among this income group it has been measured the narrowest difference between men and women recovery time.

High income

More than 250,000 MXN per year
20% of the overall population

Women of high income is the slowest community to recover their spending habits.

For a quick summary of the project, check out this video animation.

United Nations Global Pulse

Jeremy Boy
Robert Kirkpatrick
Miguel Luengo-Oroz
Mila Romanoff
Anoush Tatevossian
Felicia Vacarelu
Makena Walker
Amanda Zerbe

BBVA Data & Analytics

Elena Alfaro

Project lead
Juan Murillo

María Hernandez
Roberto Maestre
Dario Patane

Data Visualization
Jordi Aranda
Iskra Velitchkova
Alejandro Vidal

Data provided by BBVA Bancomer Mexico

Data Privacy

This project was conducted in a way that fully protected the privacy of individuals, and complied with applicable data privacy and data protection principles, as well as high standards of ethical and moral conduct. Appropriate risk assessment was conducted prior to the project’s launch. No personal data was shared for the purposes of this research project.

Anonymised data was analysed by an authorized BBVA team of researchers under strict confidentiality, and employing appropriate data security measures. Each transaction was processed as a unique data point, and no longitudinal data associating multiple data points to a single person was used in the project. Only aggregated analysis – (insights drawn from the project research) – were published to remove any possibility of re-identification. To ensure accuracy of the results, relevant domain expertise was used to assist the research.

For more information on the privacy practices of Global Pulse please refer to Global Pulse Principles.

Política de cookies BBVA.