Asymmetric price adjustment over the business cycle

Avichai Snir; Daniel Levy; Elliot Charette; Haipeng (Allan) Chen; Mark Bergen; Sourav Ray; Weihong Zhao; Xiao Ling

REVIEW 4 major objections 5 minor 13 references

Asymmetric price adjustment over the business cycle

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Using 98 million weekly grocery prices from 1989–1997, this paper argues that small price increases outnumber small decreases far more in booms than in recessions, and traces the gap to consumer attention.

desk verdict A clean, honest test of a specific prediction that lines up with the data, but the single-cycle design leaves the mechanism unproven. read the letter →

arxiv 2506.10640 v2 pith:KESHBBLG submitted 2025-06-12 econ.GN q-fin.EC

Daniel Levy , Haipeng (Allan) Chen , Sourav Ray , Elliot Charette , Xiao Ling , Weihong Zhao , Mark Bergen , Avichai Snir This is my paper · ORCID

classification econ.GNq-fin.EC

keywords asymmetricpriceadjustmentsmallchangesconsumerinattentionrigiditybusinesscyclesunemploymentretailscannerdatastickyprices

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Small price changes behave differently across the business cycle. The paper combines two established findings—consumers often ignore small price changes, and shoppers pay more attention to prices when unemployment is high—into one testable prediction: the well-known tilt toward small price increases should weaken in recessions and strengthen in expansions. Using 98 million weekly store-level price observations from a large US grocery chain over 1989–1997, the authors compare the asymmetry in small price changes across an NBER recession, two peak-unemployment windows, and a low-unemployment window. They find the asymmetry is substantially larger in the low-unemployment period, with an average threshold of 10.30 cents versus 0.62 cents in the recession, and conclude that firms' price-setting behavior varies over the business cycle.

What carries the argument

The central object is the asymmetry threshold $\bar{A}$, defined as the last absolute price change size, in cents, at which the frequency of price increases exceeds the frequency of price decreases at the $z \geq 1.96$ significance level. A larger threshold means a wider range of small price changes in which increases dominate, which is the signature of consumer inattention. The argument runs through two links: inattentive shoppers do not react to small price cuts but do not punish small price increases, so retailers have an incentive to make small increases more often than decreases; and shoppers' attention rises with unemployment, so the incentive weakens in downturns. The empirical engine is the Dominick's scanner dataset, 400 weekly observations for 18,037 products in 29 categories, with the same four 8-month windows compared for each of 27 product categories.

What would settle it

Compute the asymmetry threshold for the same 27 product categories month by month over the full 1989–1997 sample and regress it on the monthly Chicago unemployment rate. The attention story predicts a systematic negative relationship within the sample; if the threshold only differs between the two selected windows and does not track unemployment in between, the central claim is not supported.

Watch

Extended reading notes

Core claim

The paper's claim is that the asymmetry in small price changes is cyclical: the range of small price changes in which increases outnumber decreases is larger when unemployment is low. In the lowest-unemployment window (September 1996–April 1997) the average asymmetry threshold across 27 product categories is 10.30 cents, compared with 0.62 cents during the NBER recession (August 1990–March 1991), 4.15 cents in the highest-unemployment Chicago window, and 3.59 cents in the highest-unemployment US window. The comparison favors the hypothesis in 62 of 75 category-window comparisons, and robustness checks that exclude temporary sales and clearance sales leave the pattern intact. The authors interpret this as evidence that consumer inattention—and therefore retailer pricing incentives—varies with the business cycle.

Load-bearing premise

The load-bearing premise is that the larger asymmetry in the September 1996–April 1997 window is due to lower consumer attention during a boom, not to some other change in the chain, its competitors, or the economy between 1989 and 1997.

Editorial extensions

If this is right

The asymmetry threshold can serve as a measurable proxy for consumer attention in retail price data, giving future work a way to track attention over time.
Price-setting models that include consumer inattention should predict that the small-price-change distribution shifts with unemployment, since the incentive to post small increases versus decreases changes over the cycle.
Inflation measurement should treat the small-price-change margin as cyclically sensitive: the mix of small increases and decreases is not constant, so the micro-content of measured inflation varies with the business cycle.
Retailers' pricing strategy is state-dependent: they can rely more on unnoticed small increases in booms, while in downturns the same small increases are more likely to be noticed and resisted.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

An extension the paper leaves implicit: the asymmetry threshold should track monthly unemployment within the same 1989–1997 sample, not just differ between selected boom and bust windows; that time-series test would separate attention from a one-off regime change.
If the mechanism generalizes, the same threshold gap should appear in other chains and countries that have scanner data spanning at least one full recession; the prediction is that every boom shows a larger threshold than every nearby bust.
A sharper test would compare chains serving different clienteles in the same city: a chain whose shoppers have a high opportunity cost of time should show a larger asymmetry threshold in every period, because its customers are the least attentive.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: this is a clean, honest test of a specific prediction, and the data line up. But the identification is thin enough that I read it as a documented regularity, not a demonstrated mechanism. Worth reading and citing with a caveat.

The new thing here is the test itself: Chen et al. (2008) documented more small price increases than decreases and attributed it to consumer inattention. Several papers show shoppers pay more attention to prices when unemployment is high. This paper puts those together and predicts the asymmetry should shrink in recessions. Nobody else has tested that directly. Using the Dominick's scanner data over 1989–1997, they compare the asymmetry threshold (the last cent amount where increases significantly outnumber decreases) across an NBER recession, two highest-unemployment windows, and the eight-month lowest-unemployment window at the end of the sample. In 27 categories, the low-u threshold is larger in about 83% of comparisons, and the averages are dramatically different (10.3¢ vs 0.6–4.2¢). They also run sensible robustness checks: removing V-shaped sales, removing clearance sales, checking sample-size artifacts, and checking inflation mechanics.

The soft spot is the research design, not the execution. The low-unemployment window is September 1996–April 1997, the last eight months of the sample. All the high-unemployment windows are in 1990–1992. So calendar time and the cycle are nearly collinear. Any monotonic drift in Dominick's pricing strategy, promotional calendar, price-endings policy, or competitive environment would push the measured threshold up at the end of the sample and produce exactly the pattern they see. The authors acknowledge the single-chain, single-cycle limitation in Section 7, but they don't attempt to model time trends or chain-strategy drift. That's a real gap, though for a short Economics Letters it may be acceptable if framed as tentative evidence.

There's also a measurement subtlety: the asymmetry threshold is defined by a significance test (z ≥ 1.96). Because the z-statistic grows with sample size, the measured threshold is mechanically sample-size dependent. The paper tries to address this by looking at categories where the low-u sample is smaller, but that doesn't solve the issue for the whole sample, and they don't report confidence intervals or an alternative continuous measure. So the magnitude of the asymmetry difference is likely overstated, though the direction seems robust.

On the citation pattern: the heavy reliance on Chen et al. (2008) is appropriate since this is a direct extension using the same data. No circularity.

Bottom line: The paper is a legitimate, if preliminary, contribution. It deserves a serious referee and, if the editors accept the limited scope, it should be published. I'd cite it as evidence that the asymmetry may be cyclical, with a caution about identification. Bring it to the reading group if you want a good discussion of why single-cycle designs can mislead.

Referee Report

4 major / 5 minor

Summary. The paper tests whether the asymmetry in small price changes (more frequent small increases than decreases) varies over the business cycle. Using Dominick's scanner data covering 1989–1997, it compares the asymmetry threshold (Chen et al., 2008) across the NBER recession (August 1990–March 1991), the highest-unemployment 8-month periods (late 1991–1992), and the lowest-unemployment period (September 1996–April 1997). The paper finds a larger threshold in the low-unemployment period and interprets this as evidence that consumer inattention, which weakens in downturns, shapes price-setting behavior.

Significance. If the result holds, it extends the inattention-based explanation of small price changes to a business-cycle dimension and links micro price-setting to aggregate conditions. The paper uses a large, publicly available dataset and provides several thoughtful robustness checks, including V-shaped sales filters, clearance-sale exclusions, and sample-size subsample analyses. The main weakness is identification: the low-unemployment window is the terminal segment of the sample, so calendar time is nearly collinear with the cycle indicator, and the single-chain/single-cycle design limits causal attribution to consumer attention.

major comments (4)

[Section 4] The low-unemployment window (September 1996–April 1997) is the last 8 months of the sample, while all three high-unemployment windows fall in 1990–1992. This makes calendar time almost collinear with the business-cycle indicator. Any monotonic drift in Dominick's pricing strategy over 1989–1997—such as changes in price-ending policies, private-label assortment, promotional calendars, or competitive pressure—would produce exactly the reported pattern. The paper acknowledges the single-chain/single-cycle limitation, but it does not provide a test of the time-trend alternative. This is load-bearing because the concluding attribution to consumer attention requires ruling out such trends. I would like to see an explicit model in which the asymmetry threshold (or a sample-size-invariant measure of asymmetry) is regressed on unemployment using all monthly windows with a time-trend control, or a placebo comparison using another low-unemployment window within the sample (e.g., 1994–1995).
[Section 3] The asymmetry threshold is defined as 'the last point at which the frequency of price increases exceeds the frequency of price decreases of the same absolute magnitude with z≥1.96.' Because the threshold depends on statistical significance, it is not invariant to sample size: larger samples in the lowest-u period can mechanically produce a larger threshold even when the raw difference in frequencies is unchanged. The robustness checks in Section 6 partially address this, but they do not re-estimate the thresholds after equalizing sample sizes (e.g., by random subsampling) or using a measure that does not depend on statistical significance, such as the raw difference in the share of increases minus decreases at each change size, or the integrated difference over a fixed range. This matters because the quantitative claim of '2.5–16.6 times stronger asymmetry' is based on this threshold measure.
[Section 5] The cross-category tests treat the 27 product categories as independent, but all categories are priced by the same chain and likely share common shocks, including chain-wide promotional calendars, common cost shocks, and the same regional labor market. The reported t-statistics and the binomial sign test do not account for this clustering, so the statistical significance is probably overstated. Reporting cluster-robust standard errors at the chain-week or category-group level, or at least acknowledging the correlation, would be appropriate. The sign pattern is strong, but the p-values should not be taken at face value.
[Section 6] The inflation robustness check in Section 6 compares only the average inflation rates over the 8-month windows (Table A1). This does not control for inflation dynamics within the windows, and because Chen et al. (2008) document asymmetry in small price changes across inflation regimes, the concern that inflation drives the cyclical pattern is not fully addressed. A category-level inflation control or a regression framework would be more convincing.

minor comments (5)

[Section 5] The sentence 'We find 2.5–16.6 times stronger asymmetry when unemployment is low' uses ratios to a near-zero baseline (e.g., 10.30/0.62); this ratio is unstable when the denominator is close to zero. Reporting absolute differences or the distribution of thresholds would be more informative.
[Table 1] The text refers to 'the LHS panel of Table 1,' but Table 1 contains four threshold columns and four sample-size columns, not a dedicated left-hand side panel. Please rephrase to 'the threshold columns of Table 1.'
[Section 5] The notation for the threshold variable, A, appears in the text without the overbar that is used in the tables and equations. Please use a consistent notation throughout, and define the symbol at first use.
[Section 6] The degrees of freedom are reported inconsistently (e.g., t20 = 5.18 versus t26 = 3.43). Please use the standard form t(df) = value, and verify that the correct degrees of freedom are reported for each paired test, especially when some categories have missing observations.
[References] The sign test in Section 5 cites Chakraborty et al. (2015), but a standard reference for the binomial test would be more appropriate; please add one or clarify why that citation is relevant.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the business-cycle comparison of asymmetry thresholds is a new empirical result, not a restatement of the measure or theory it builds on.

full rationale

The paper is an empirical test of a cross-prediction derived from two bodies of prior work: Chen et al. (2008), which documents asymmetric small price changes and attributes them to consumer inattention, and independent studies showing that consumers are more price-attentive during downturns. The paper does not fit any parameter to the target result. It measures the asymmetry threshold, defined in Chen et al. (2008) as 'the last point at which the frequency of price increases exceeds the frequency of price decreases of the same absolute magnitude (z≥1.96)', separately in low- and high-unemployment 8-month windows from the same Dominick's scanner data. The comparison across business-cycle states is new: the reported thresholds (A=10.30¢ for the lowest-unemployment period versus 0.62¢, 4.15¢, and 3.59¢ for the high-unemployment periods) are computed from the data and are not imposed by the definition of the threshold. Although the asymmetry measure and the inattention theory come from prior work by overlapping authors, that work is a published, externally grounded Journal of Monetary Economics article; citing it is not circular because the present paper's contribution—cyclical variation in the threshold—does not reduce to the cited result. The sales-filter robustness checks also rely on externally developed filters, and the acknowledged single-chain, single-cycle limitation is an identification threat, not a circularity.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper rests on the established inattention theory, the empirical regularity that attention rises with unemployment, and several measurement choices (threshold definition, period selection). No free parameters are fitted. The most fragile input is the single-cycle identification assumption, which the paper itself flags as a caveat.

assumptions (5)

domain assumption Consumers are inattentive to small price changes, leading to more frequent small price increases than decreases.
This is the core theory from Chen et al. (2008) that motivates the existence of asymmetry.
domain assumption Consumer attention to prices increases when unemployment is high and decreases when unemployment is low.
Taken from cited studies (Aguiar and Hurst 2007, Nevo and Wong 2015, Coibion and Gorodnichenko 2015); needed to generate the cyclical prediction.
domain assumption The asymmetry threshold, defined as the last price change size where increases significantly outnumber decreases at z >= 1.96, is a valid measure of the inattention range.
The significance-based definition makes the threshold sensitive to sample size, which is a core methodological concern for the comparison.
ad hoc to paper The chosen 8-month low- and high-unemployment periods differ only in business cycle conditions, with no confounding time trends or other shocks.
The sample has one recession and one boom, so this assumption cannot be tested with the data; it is the paper's main identification assumption.
domain assumption The Dominick's price data are recorded without systematic measurement errors that vary with the business cycle.
Measurement error in price changes could affect the asymmetry measure; the paper assumes it does not differ across the compared periods.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Asymmetric price adjustment over the business cycle." pith.science (2026). https://pith.science/paper/KESHBBLG

@misc{pith2026250610640,
  author       = {Pith},
  title        = {Pith review of: Asymmetric price adjustment over the business cycle},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KESHBBLG}},
  note         = {Machine review of arXiv:2506.10640}
}

read the original abstract

Studies of micro-level price datasets find more frequent small price increases than decreases, which can be explained by consumer inattention because time-constrained shoppers might ignore small price changes. Recent empirical studies of the link between shopping behavior and price attention over the business cycle find that consumers are more attentive to prices during economic downturns, and less attentive during economic booms. These two sets of findings have a testable implication. The asymmetry in small price changes should vary over the business cycle. It should diminish during recessions and strengthen during expansions. We test this prediction using a large US store-level dataset with more than 98 million weekly price observations for the years 1989-1997, which includes an 8-month recession period, as defined by the NBER. We compare price adjustments between periods of recession - high unemployment, and expansion - low unemployment. Focusing on small price changes, we find, consistent with our hypothesis, that there is a greater asymmetry in small price changes during periods of low unemployment compared to the periods of high unemployment, implying that firms price-setting behavior varies over the business cycle.

Figures

Figures reproduced from arXiv: 2506.10640 by the authors.

**Figure 2.** Monthly unemployment rate in the US and Chicago, 1989–1997 Notes: 1. The left-hand side shaded area marks the NBER recession period, August 1990–March 1991. 2. The middle-shaded area marks the highest unemployment rate periods, February 1992–September 1992 in the US, and December 1991–July 1992 in Chicago. 3. The right-hand side shaded area marks the lowest unemployment periods, September 1996 –April 1997, in both t… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

[1]

The left-hand side shaded area marks the NBER recession period, August 1990–March 1991

work page 1990
[2]

The middle-shaded area marks the highest unemployment rate periods, February 1992–September 1992 in the US, and December 1991–July 1992 in Chicago

work page 1992
[3]

The right-hand side shaded area marks the lowest unemployment periods, September 1996 –April 1997, in both the US and Chicago. 14 Table 1 Variation in the asymmetry threshold in cents over the business cycle Asymmetry threshold ()A in cents Sample size Product Categories Lowest u NBER Recession Highest-u Chicago Highest-u US Lowest u NBER Recession Highes...

work page 1996
[4]

Lowest u denotes the lowest unemployment rate period for both the City of Chicago and the US

work page
[5]

NBER Recession denotes the NBER recession period. 15

work page
[8]

The empty cells are the cases of missing observations. 1 Online Supplementary Web Appendix (Not for Publication) Asymmetric Price Adjustment over the Business Cycle Daniel Levy* Department of Economics Bar-Ilan University, Emory University, ICEA, ISET at TSU, and RCEA Haipeng (Allan) Chen Tippie College of Business University of Iowa Sourav Ray Lang Schoo...

work page 2025
[13]

The blank cells are the cases of missing observations 5 Table A3 Variation in the asymmetry thresholds in cents over the business cycle, after excluding clearance sales events ()A Asymmetry threshold Sample size Product Categories Lowest u NBER Recession Highest-u Chicago Highest-u US Lowest u NBER Recession Highest-u Chicago Highest-u US Analgesics 16 0 ...

work page
[18]

The blank cells are the cases of missing observations 6 Table A4 Variation in the asymmetry thresholds in cents over the business cycle, after simultaneously excluding V-shaped sales and clearance sales events ()A Asymmetry threshold Sample size Product Categories Lowest u NBER Recession Highest-u Chicago Highest-u US Lowest u NBER Recession Highest-u Chi...

work page

Show all 13 references

[19]

Lowest u denotes the lowest unemployment rate period, for both the City of Chicago and the U.S
[20]

NBER Recession denotes the NBER recession period
[21]

Highest-u Chicago denotes the highest Chicago unemployment rate period
[22]

unemployment rate period

Highest-u US denotes the highest U.S. unemployment rate period
[23]

The blank cells are the cases of missing observations

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

The left-hand side shaded area marks the NBER recession period, August 1990–March 1991

work page 1990

[2] [2]

The middle-shaded area marks the highest unemployment rate periods, February 1992–September 1992 in the US, and December 1991–July 1992 in Chicago

work page 1992

[3] [3]

The right-hand side shaded area marks the lowest unemployment periods, September 1996 –April 1997, in both the US and Chicago. 14 Table 1 Variation in the asymmetry threshold in cents over the business cycle Asymmetry threshold ()A in cents Sample size Product Categories Lowest u NBER Recession Highest-u Chicago Highest-u US Lowest u NBER Recession Highes...

work page 1996

[4] [4]

Lowest u denotes the lowest unemployment rate period for both the City of Chicago and the US

work page

[5] [5]

NBER Recession denotes the NBER recession period. 15

work page

[6] [8]

The empty cells are the cases of missing observations. 1 Online Supplementary Web Appendix (Not for Publication) Asymmetric Price Adjustment over the Business Cycle Daniel Levy* Department of Economics Bar-Ilan University, Emory University, ICEA, ISET at TSU, and RCEA Haipeng (Allan) Chen Tippie College of Business University of Iowa Sourav Ray Lang Schoo...

work page 2025

[7] [13]

The blank cells are the cases of missing observations 5 Table A3 Variation in the asymmetry thresholds in cents over the business cycle, after excluding clearance sales events ()A Asymmetry threshold Sample size Product Categories Lowest u NBER Recession Highest-u Chicago Highest-u US Lowest u NBER Recession Highest-u Chicago Highest-u US Analgesics 16 0 ...

work page

[8] [18]

The blank cells are the cases of missing observations 6 Table A4 Variation in the asymmetry thresholds in cents over the business cycle, after simultaneously excluding V-shaped sales and clearance sales events ()A Asymmetry threshold Sample size Product Categories Lowest u NBER Recession Highest-u Chicago Highest-u US Lowest u NBER Recession Highest-u Chi...

work page