The world is not just getting warmer, our climate is increasingly more variable too. How does this affect agriculture, industry, health, schools, and poverty levels? We dive into the data...
As countries race to deliver new pledges to tackle global warming following COP26, one issue related to temperature receives relatively less attention – its variability. Temperature variability affects whole economies, households, firms, and - as with most impacts of climate change - will disproportionately be felt by the poor. Climate models predict that temperature variability is increasing for poor countries, making it a threat to global poverty reduction efforts. But in order to design adequate responses, we need to better understand its relationship with poverty.
Enter big data. As part of DEEP, we are exploring the use of modern data science techniques to get a more precise and localised picture of poverty and its relationship with temperature variability. But first, how bad exactly is temperature variability for economic outcomes?
A macro view
Recent research shows that an extra degree of day-to-day temperature variability results in a 5-percentage point reduction in regional growth rates. Another study finds that negative effects are larger for day-to-day than seasonal variation. On an annual scale, the impact is positive at low and negative at high temperatures. Since poor countries tend to be hotter, the unequal impact of climate change strikes again. Indeed, an analysis of 136 countries shows temperature fluctuations reduce economic growth in poor countries but have little effect in rich countries.
Can temperature variability be deadly? Some recent evidence suggests it can. In the US, increased variation of monthly temperatures was found to cause increased mortality. Several other health indicators are also affected. In Colombia, temperature shocks negatively influenced birth outcomes, such as weight and length at birth, making it 0.4 percentage point less likely that a newborn will be classified as healthy.
Poverty traps revisited
The concept of poverty traps is not new to development economics theory, but it increasingly helps us to understand the role of climate in driving poverty. The basic idea is that individuals can be “stuck” in a low-income equilibrium unless they generate sufficient “threshold” capital to break out of it. Temperature variability, by introducing uncertainty and negative shocks to income and assets, can bring households below that threshold level, trapping them in poverty.
Recent research from Vietnam shows that temperature variability leads to lower consumption, an important poverty indicator. Similar evidence emerged for Mexico and Malawi. In Ghana and Tanzania, lower farm revenues due to temperature variability meant households spent less on food and non-food items. A variable climate can also keep children out of school, as shown by a study from Ethiopia.
The agriculture channel
Most of the world’s poor work in agriculture. The best way to understand how temperature variability might be related to farm income theoretically is to transport oneself back to the 18th century and ask what happens to the value of land in the face of climatic shocks. As famously observed by David Ricardo in that period, the value of land depends on how much output you can produce with a given number of inputs. Temperature variability, by lowering crop yields, reduces not just landowners’ income, but also the incomes of wage workers employed at the farm.
Ricardian insights may be old, but they do pretty well at describing how farmers are affected by climate variability around the world today. In Togo, higher standard deviations of temperature lower cereal crop income for local farmers. Similarly, cereal yields in Nigeria remain highly vulnerable to temperature variability. In El Salvador, temperature variability negatively affected not just corn yield, but also the number of workers hired in corn production and their wages.
Survival of the most productive
Temperature variability can affect industrialised economies too. Early theories of heterogeneous firms, where only the most productive firms survive, are now used to think about climate change impacts. Recent adaptations of this model show that temperature shocks can raise firms’ production costs, lower their productivity, and force some of them to exit.
In the US, seasonal temperature extremes lead to high work absenteeism. Similarly, extreme heat reduces weekly production from automotive plants in the US by 8%. In Cote d’Ivoire, more days with high temperatures decreased firms’ revenues and profits by 14.83% and 21.71%, respectively, raising the exit rate by 0.04%.
How can big data help?
The above shows that evidence already exists on the negative impact of temperature variability on poverty. However, crucial methodological issues related to the data that is used in most studies in low-and middle-income countries persist:
- There are long gaps between rounds of household surveys used to measure poverty
- Missing values or lack of data on poverty for all regions
- Long periods between collection of temperature data from meteorological stations
- Aggregated temperature data obscuring regional variation.
While some studies have already tackled points 3 and 4 by relying on high-frequency geospatial data, challenges remain on the poverty side. Infrequent data points for poverty at a high level of aggregation make it difficult to identify exactly which households are most vulnerable to impacts of temperature variability. New approaches to small area estimation (SAE) combine geo-spatial data with household surveys in a machine learning algorithm to predict poverty rates at a much more localised level. The punchline: a high-frequency poverty map that accurately identifies poor households at a very low spatial level that we previously knew nothing about due to lack of survey data.
It is clear that policymakers should recognise the threats of not just a hotter, but also a more variable climate. Temperature variability has a negative impact on economic growth, mortality, health, consumption, school enrolment, farm income, crop yields, and productivity. While this is a lengthy list of indicators, we still lack a precise understanding of who is affected - mostly due to the use of highly aggregated and infrequent data on poverty. Governments need to think precisely about who the poor are – and how to design better targeted social programs to protect them. SAE offers a solution via the prediction of poverty at a very disaggregated spatial level. But its success will depend on the willingness of policymakers to invest in big data approaches to monitoring poverty and temperature variability.