Did media outlets overreport on rising inflation during the pandemic? In my corpus of Washington Post, New York Times, and USA Today articles, the answer is a qualified yes. Two of the three outlets devoted more coverage than you would expect and that coverage remained very elevated as inflation receded. Coverage from The Washington Post was not elevated.
Separately, I consider whether these outlets' tone on inflation matched inflation dynamics. I find that tone during the pandemic was no different from previous periods of inflation. In certain regression specifications, there was a slight negative impact of the pandemic on tone, but the effect size is small.
An important caveat for all these findings is that because I only have data from 2000 forward, I don't have a period of inflation comparable to the pandemic. So the pandemic is a bit out-of-sample for prediction, but I think there's enough variation to provide useful information.
The first figure below compares all coverage in my corpus to year-over-year PCE inflation. I've standardized all variables for all of the below graphs to make comparison easy. You can see that total coverage of inflation is slightly elevated in the beginning of the pandemic as inflation is falling, but is especially elevated starting in 2023, as inflation starts to fall rapidly but coverage falls only modestly.
The figure below shows the same thing, but with my three outlets broken out. You can see how The Washington Post has a markedly different trajectory than the other outlets, with a volume of coverage that doesn't rise as far and then tapers off much quicker.
The pandemic coverage bump is large. If we regress volume of coverage on PCE inflation and a pandemic dummy variable, the pandemic dummy is significant and indicates that pandemic coverage is 1.3 standard deviations higher than expected. If we instead use one dummy variable for the pandemic's era of increasing inflation and one for its era of decreasing inflation, both are significant, but the magnitude during the initial era of increasing inflation is less than one standard deviation, whereas the magnitude during the era of decreasing inflation is nearly two standard deviations. Within outlet regressions mirror the visual findings. The Washington Post's volume of coverage is well-predicted by PCE alone, and pandemic dummy variables are insignificant.
Another interesting observation from these graphs is that in every period except for the pandemic, coverage actually isn't very responsive to levels of inflation at all. Regressing coverage volume on inflation level prior to 2020 shows that the level of inflation is not a significant predictor of the amount of coverage. This might be because we don't observe a month with inflation higher than 4.1% outside the pandemic (where it rises to 7.2% at the peak).
One more thing about coverage. What happens if we divide articles into those contain the word "election" and those that don't? I thought this might be an interesting look into how inflation played into political narratives. The graph below shows raw counts of how many times inflation was mentioned in an article that also contains "election" and articles that don't contain "election."
You can see how significant the spikes are in 2022 and 2024. Prior to the pandemic, inflation is mentioned alongside elections only rarely. So inflation was a big horserace topic, but we shouldn't take too much from this graph; without a comparable inflation spike in the rest of the series, it's hard to say if this is unusual. But election coverage was a key reason that the volume of inflation coverage remained elevated in 2024, despite decreasing levels of inflation.
The story is quite different for sentiment, where we see no strong effect from the pandemic. Because sentiment is quite volatile, I've used a 5-month rolling average to make the graphs easier to interpret. I've also flipped the polarity of sentiment such that positive values indicate negative sentiment. Again, this just makes the graphs much easier to read, as it's quite difficult to parse a negative correlation with this much volatility.
Regression results mirror the visual: pandemic dummies for sentiment are insignificant except in a couple of cases, where their magnitude is small.
Outside of the pandemic, inflation is a significant predictor of sentiment, although the relationship is fairly weak. There is relatively less variation by outlet in sentiment than in volume of coverage. The Washington Post and USA Today are both a bit more negative in tone than you would expect when inflation is rising during the pandemic.
Volume of coverage is simply a count of the number of times inflation is mentioned in the context of the U.S. economy. I filter out articles that don't mention the Federal Reserve, CPI, PCE, the Bureau of Labor Statistics, or the Bureau of Economic Analysis to remove articles about inflation in other countries.
To obtain media tone on inflation, I'm adopting the approach of Luca Barbaglia, Sergio Consoli, and Sebastiano Manzan in their paper on economic forecasting with economic news (more here). This is a relatively recent approach that is economic domain specific, has been validated against human coders, and has shown promise for economic forecasting.
My pandemic dummy run from May of 2020 to September of 2024, when inflation returned to basically pre-pandemic levels. I also look at a period of increasing inflation, from the beginning of the pandemic to June of 2022, and the subsequent decreasing period of inflation.