A puzzlingly bad September for Polish retail sales 

We expected that last September was not the best month for the Polish consumer - our retail sales forecast was 0% yoy, while the consensus optimistically forecast a much higher figure (2.5% yoy). Meanwhile, sales tanked (-3% yoy), one of the biggest negative sales surprises on record. Indeed, this data is just too bad to be true. Before we get into speculation, we'll sum up today's reading. 
First, sales in September alone fell (after excluding seasonal and calendar factors) by 6.7% mom, an unprecedented figure outside of the months in which stores were forced to close down due to COVID-19. This bump not only erased a year and a half of slow recovery, but pushed sales back to late 2019 levels. 

Retail sales (constant prices, s.a., 2015 = 100)

Source: Macrobond, Statistics Poland, Pekao Research

Second, virtually all categories of consumer goods performed poorly - their growth rates were often several percentage points lower than our assumptions. This makes analyzing the mechanics of the September reading futile. Paradoxically, the only category that minimally surprised on the positive side was the one we knew would perform badly, i.e. sales of textiles, clothing and footwear (due to high temperatures delaying the start of autumn sales). 

Retail sales – breakdown into categories (constant prices, % yoy)

Source: Statistics Poland, Pekao Research

Private consumption has generally disappointed us to the downside in recent months, but never on a scale similar to the September retail sales. This data makes no sense in the context of what we know about the Polish economy and other already known data from this period. For we know that consumer sentiment, however mediocre, improved just then, and real income growth remain high. Nor was there a sharp deterioration in the labor market. In general, no new economic factor affecting the propensity to consume and the resources available for consumption emerged in September. Because of that, our attention must turn to non-economic factors, and the analogy with store closures works surprisingly well. The problem is that we have no obvious and uncontroversial candidates. 

• The flood, in our view, was too small in scope and lasted too short a time to reduce consumption on such a scale. Besides, its impact is not unidirectional (precautionary shopping of supplies, flood relief, building materials, etc.). We can't rule out psychological effects, but we doubt they exist. The much more severe flooding of 2010 cannot be found in retail sales data.
• The unusually warm and sunny weather in September may have discouraged consumption of goods and encouraged switching to services, but September 2024 was a repeat of the previous year, and there was no bump in sales then.
• Statistics Poland has not reported any problems in data collection and processing, so we assume that the September data is complete. 

The conclusion will be disappointing for readers: we don't know why this happened. What we do know, however, is that such a result must negatively affect economic growth forecasts. After September's batch of data (all of which disappointed), we are revising our estimate of Q3 GDP down from 3.2 to 2.8% y/y, suggesting that the full-year average will also be just below 3%. We assume that most of the September sales decline will be made up in October, but the uncertainty about this is significant.