Bank of Korea study finds won-spike veterans cut spending less when the currency falls

The panel covers the 27 survey years from 1998 through 2024, so its 4,460 observations come to about 165 a year ^[1]. The core result rests on that pool. The higher the current exchange rate, the smaller the relative spending cut among households with more past exposure to currency surges ^[3]. Every finding after that is a slice of the same data, cut by birth cohort, asset bracket, debt-service quintile and spending line ^[5][6][7][8].

A weaker won pushes up import prices and erodes real purchasing power ^[12]. Choi's explanation for the smaller cuts is learning: households that have been through several spikes understand how a weaker won eats into purchasing power and how to respond ^[9]. "The more experience households have with exchange rate surges, the more limited the negative impact of a rising exchange rate on their consumption," Choi said ^[10]. He measures against 1997, when household consumption contracted sharply after the currency spike. Spending in later episodes was far less sensitive ^[11].

Learning is one way to read those coefficients. Depletion is another. The effect is strongest in the lower asset brackets and among the top 20% of households by debt service ratio, the group with the largest principal and interest payments relative to income ^[6][7]. For that indebted group the average marginal effect of past exposure on consumption was generally positive, while the upper asset brackets showed no statistically clear effect ^[7][6]. A household that cut its discretionary budget in 2008 and again in 2022 ^[4] has less left to cut in the next spike. A regression would record the smaller cut as adaptation.

A third reading is life stage. Gen X members were 17 to 32 years old when the 1997 crisis hit ^[2]. MZ members were at most 16 then, and the youngest were 14 in 2024, the panel's final year ^[3]. Some of the gap between the cohort that cut less and the cohort that cut more ^[5] may be age as much as memory (or rather, memory may partly stand in for having run a household through a shock at all).

I think the learning effect is real, and narrower than the core result suggests. The category detail supports it. Adaptation is clearer in discretionary items than in necessities, clearest in telecommunications costs, and widens for vehicle maintenance as the rate rises ^[8]. Those are lines where a household decides how much to spend and when. The counter-case is that discretionary lines are also the first place an indebted household runs out of room, so the same pattern fits depletion.

For anyone forecasting Korean consumption off a weak won, the study settles the direction: households with more spike experience cut less ^[3]. As reported, the Bank of Korea study does not estimate how much smaller the aggregate hit is, and it does not test the result against the bank's own forecasting models ^[1]. To separate learning from depletion you would need spending levels from before each spike. If the smallest cuts come from households whose discretionary budgets were already the lowest going in, depletion explains more of the result than learning does.