When data lie: correcting a conflict variable reverses a significant finding in Chad's agricultural time series (1961–2024)

In the vast, sun-scorched landscapes of the Sahel, agriculture is a high-stakes gamble played against the sky. For millions of people in countries like Chad, the difference between a full belly and an empty one depends on the rhythm of the rains and the stability of the land. Scientists have long tried to understand exactly how much these two forces—climate and conflict—push and pull on crop harvests. They look at data stretching back decades, searching for patterns that can help predict famine or guide policy. But this search relies entirely on the accuracy of the records kept by researchers. If the numbers in a spreadsheet are wrong, even slightly, the entire story they tell can be a lie. This is not a matter of malicious deception, but rather the quiet, accidental accumulation of errors that happen when files are passed from hand to hand without careful checking.

A recent study focusing on Chad, a nation where nearly seventy percent of the workforce depends on farming, exposes just how fragile these conclusions can be. The country has faced a complex history of violence, from a long civil war in the mid-twentieth century to various rebellions and insurgencies in the decades that followed. Researchers wanted to know how this fighting, combined with changing rainfall patterns, affected the harvests of three staple crops: millet, sorghum, and maize. To do this, they relied on a dataset that tracked whether the country was at war or at peace for every year from 1961 to 2024. The initial analysis suggested a startling result: that years of armed conflict actually led to higher yields of millet. This finding seemed robust and statistically significant, leading to a conclusion that would have reshaped how experts view the relationship between war and food security in the region.

However, the author of this study, Abderamane Mahamat Abdel-Aziz, decided to go back to the source. Instead of trusting the spreadsheet as it was, he systematically compared the data against the original, primary databases where the information was first recorded. He checked four different sources: records for crop yields, climate data, fertilizer use, and the history of armed conflict. This process of verification revealed two major problems. First, the crop yield numbers had been multiplied by a thousand in a cyclical error, requiring a complete rebuild of that section of the data. Second, and more critically, the record of armed conflict was fundamentally broken. The original file only counted the years of the major civil war from 1965 to 1987 as times of conflict. It treated the remaining decades as peaceful, completely ignoring decades of active fighting that were well-documented in official international records.

When the researcher corrected this mistake, coding the conflict variable to reflect the true history of violence over the entire sixty-four-year period, the story changed completely. The previously significant finding that war boosted millet harvests vanished. The statistical link between conflict and higher yields disappeared, leaving a result that showed no real effect at all. The idea that fighting helped farmers grow more food was not a hidden truth waiting to be discovered; it was a statistical illusion created by missing data. The corrected analysis showed that for nearly half the years studied, the country was actually at war, but the old file had labeled them as peace. This error had created a false correlation that made it look like the crops were thriving during the wrong years.

With the conflict variable fixed, a different, more reliable pattern emerged regarding the weather. The study confirmed that rainfall patterns, measured by a standard index of drought and wetness, had a clear and positive impact on the harvests of millet and sorghum. When the rains were good, these crops grew better. This effect was strong and consistent, matching what agronomists expect for crops that rely entirely on the rain. Maize, which is often grown in wetter southern areas and might have different needs, showed no such clear link to the rainfall index in this specific analysis. The study also found that the use of nitrogen fertilizer helped sorghum yields, but did not significantly change the harvests of the other two crops.

The most important lesson from this work is not about the specific crops of Chad, but about the necessity of checking the facts before drawing conclusions. The study demonstrates that a single error in how a variable is coded can reverse a finding that appears to be highly significant. It shows that in the complex field of studying African agriculture, where data is often assembled from many different sources, the act of verification is not just a luxury for experts but a basic requirement for credibility. By returning to the primary records and correcting the record of conflict, the researcher did not just fix a number; he corrected the narrative, ensuring that future policies and scientific understandings are built on the reality of the land, not on the errors of a spreadsheet.