^{1}

^{1}

^{1}

Comparison shopping is good financial practice, but situations involving numbers and computations are challenging for consumers with math anxiety. We asked North Americans (N = 256) to select the better deal between two products differing in volume and price. As predicted, math anxiety was negatively related to performance on this Price Comparison Task. We then explored the mechanism underlying this relation by testing math competency, price calculation ability, need for cognition, and cognitive reflection as potential mediators. The results from a competing mediator analysis indicated that all factors, apart from need for cognition, served as significant independent mediators between math anxiety and performance on our Price Comparison Task. This study has important implications for how–and why–math anxiety relates to a person’s ability to accurately compare product prices. These data suggest that consumers higher in math anxiety may represent a financially vulnerable population, particularly in the context of financial tasks that are inherently mathematical.

Comparing products while shopping is good financial practice and is measured on several indices of ideal financial behaviour (

Math anxiety is a global phenomenon (

Indeed, math anxiety has been linked to difficulty interpreting medical risks, such as deciding between treatments when given outcomes (

Another factor related to people’s abilities to make sound financial decisions is math competency (see

Research indicates that people who are higher in math anxiety tend to be lower in their need for cognition (

Evidence suggests there is a negative relation between math anxiety and cognitive reflection both in general and in situations requiring math (

The first objective of the current research is to test the hypothesis that adults higher in math anxiety will perform worse on a price comparison task (i.e., a novel “Price Comparison Task”) than those lower in math anxiety (Hypothesis 1). Working under the assumption that a negative relation between math anxiety and task performance does indeed exist, our second objective is to test a series of theoretically driven hypotheses designed to elucidate the mechanism by which this relation occurs. Specifically, we test whether the relation between math anxiety and performance on the Price Comparison Task is mediated by math competency, price calculation ability (Hypothesis 2), NFC (Hypothesis 3), and cognitive reflection (Hypothesis 4).

We recruited North American participants using two modes: a student research pool at a university in Canada (

We removed five participants who failed two or more (out of three) directed attention checks in the data (

We designed a novel task to assess product comparison behaviour when consumers are presented with identical products that differ only by size (in terms of volume) and price. We simultaneously presented participants with two products with the price and size presented below the image (in dollars and metric units, respectively).^{1}

We recognize that many Americans are not accustomed to working with the metric system. However, average performance on the Price Comparison Task for participants residing in Canada (

^{2}

All materials for the Price Comparison and Calculation Tasks are available in the

The task included 40 trials, split into two parts. In the first 32 trials, we asked participants to identify which product was the better deal (i.e., cost less per metric unit) between the two products presented. We presented trials one at a time. Scores range from zero to 32, with higher scores indicating better performance on the task. We will herein refer to this first portion of the task as the “Price Comparison Task.”

We then showed participants four of the same product pairs with modified instructions. We chose only four of the original 32 product dyads in the second portion of the task to reduce survey fatigue while still capturing an accurate representation of their calculation abilities. We asked participants to calculate the exact price per metric unit for each product (i.e., eight products total) and report the price to the fourth decimal place. However, we coded all responses that demonstrated the participant understood how to accurately compute the product price (i.e., understood the procedure) as correct. For example, the correct response for Option 1 in

The purpose of this exercise was to determine if participants could correctly calculate the product’s price per unit when prompted. Previous work has linked higher math anxiety with general math competency (

We measured math anxiety with the nine-item Abbreviated Math Anxiety Scale (AMAS) (

We measured trait anxiety using the 20-item State-Trait Anxiety Inventory, Trait subscale (STAI-T) (

We measured math competency using the Brief Mathematics Assessment 3 (BMA-3) (

Due to a programming error, responses to Item Two of the BMA-3 did not record for participants recruited through the university student pool (^{3}

A series of

We measured participants’ motivation to engage in effortful cognitive activities using the 18-item Need for Cognition scale (NFC) (

We measured participants’ tendency to reflect or deliberate to verify intuitive insights using the four-item Cognitive Reflection Test 2 (

We asked participants demographic questions including age, gender, occupation, and education. We did not ask participants to disclose their ethnic background.

We programed our study using the Experiment Builder function on Gorilla.sc. We allowed participants to compete the study using a computer or tablet, but not a mobile device, as the small screen distorted our product images and textbox placements. We allowed participants to use a calculator during the study.

After providing informed consent, participants completed demographic questions followed by the measures of anxiety and cognition. We randomized presentation of the AMAS, STAI-T, NFC, and CRT-2. Participants then completed the Price Comparison Task (i.e., select the product that represents the “better deal”) followed by the Price Calculation Task (i.e., calculate the price per unit for each product). We did not impose a time restriction for these measures. Lastly, participants had 10 minutes to complete the BMA-3 before the survey timed out and directed participants to the debriefing form. We obtained ethical approval for this study from the Office of Research Ethics and Integrity at the University of Ottawa (file number H-11-19-4999).

We performed analyses using SPSS version 28. All analyses control for recruitment method (i.e., student pool or Prolific), gender, and general anxiety. We included gender as a covariate within our analyses as it is well-documented that women tend to report higher levels of math anxiety compared to men (

Variable | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
---|---|---|---|---|---|---|---|---|

1. Gender^{a} |
.16* | .07 | -.08 | -.02 | -.06 | -.02 | .09 | |

2. Math Anxiety^{b} |
– | .32*** | -.31*** | -.21** | -.23*** | -.12* | -.20** | |

3. General Anxiety^{b} |
-.22*** | .01 | .02 | .15* | .09 | |||

4. Need for Cognition^{b} |
.05 | .10 | .08 | .11 | ||||

5. Cognitive Reflection^{b} |
.37*** | .38*** | .18** | |||||

6. Math Competency^{b} |
– | .35*** | .29*** | |||||

7. Price Comparison Task^{b} |
.33*** | |||||||

8. Price Calculation Task^{b} |
||||||||

256 | 256 | 256 | 256 | 233 | 256 | 256 | ||

21.77 | 47.32 | 59.65 | 2.38 | 7.78 | 25.80 | 5.71 | ||

8.06 | 11.10 | 12.54 | 1.16 | 2.16 | 4.71 | 3.12 | ||

Minimum Score | 9 | 20 | 21 | 0 | 2 | 13 | 0 | |

Maximum Score | 44 | 76 | 85 | 4 | 11 | 32 | 8 | |

Possible Range | 9 – 45 | 20 – 80 | 18 – 90 | 0 – 4 | 0 – 11 | 0 – 32 | 0 – 8 |

^{a}Dichotomous variable; Gender 1 = male, 2 = female. ^{b}Mean scores.

*

We inferred missing data was missing at random as Little's MCAR test was statistically significant, χ^{2}(1683) = 1851.41,

In line with our first objective (i.e., to test the hypothesis that people higher in math anxiety will make more errors on the Price Comparison Task compared to people lower in math anxiety), we conducted a multiple regression analysis. Our dependent variable is performance on the Price Comparison Task. Consistent with our prediction, those who were higher in math anxiety were less likely to select the product that represented the better deal (β = -.20,

Variable | Average Score on Price Comparison Task |
|||||
---|---|---|---|---|---|---|

B | _{B} |
β | 95% CI for B |
|||

Constant | 0.78 | 0.09 | < .001 | 0.60 | 0.95 | |

Recruitment Method | -0.01 | 0.03 | -0.02 | 0.739 | -0.08 | 0.05 |

Gender | -0.00 | 0.02 | -0.01 | 0.851 | -0.04 | 0.03 |

General Anxiety | 0.23 | 0.07 | 0.21 | 0.001 | 0.09 | 0.36 |

Math Anxiety | -0.16 | 0.06 | -0.20 | 0.004 | -0.27 | -0.05 |

^{2} |
0.06 | |||||

△^{2} |
0.04 | |||||

3.73 | 0.006 |

_{B}^{2}^{2} = adjusted ^{2}

In line with our second objective (i.e., to understand the mechanism by which the relation between math anxiety and performance on the Price Comparison Task occurs), we conducted a competing mediator analysis. With this analysis we assessed whether various cognitive factors (i.e., math competency, price calculation ability, NFC, or cognitive reflection) uniquely explain the relation between math anxiety and performance on the Price Comparison Task. We used the PROCESS macro version 4.1, model number four in SPSS (

After controlling for recruitment method, gender, and general anxiety, the total indirect effect was significant as the 95% confidence interval did not include zero (effect = -.15., 95% CI [-.22, -.08]). The total indirect effect accounted for 91% of the total effect (%C = .91). The indirect effect of NFC was not significant (effect = -.01, 95% CI [-.04, .01]), suggesting that NFC does not mediate the relation between math anxiety and performance on the Price Comparison Task. Math competency (indirect effect = -.04, 95% CI [-.08, -.01]), price calculation ability (indirect effect = -.04, 95% CI [-.08, -.02]), and cognitive reflection (indirect effect = -.06, 95% CI [-.10, -.02]) served as significant independent mediators. The indirect effect of math competency accounted for 23% of the total effect (%C = .23), the indirect effect of price calculation ability accounted for 27% (%C = .27), and the indirect effect of cognitive reflection accounted for 34% (%C = .34). The direct effect between math anxiety and performance on the Price Comparison Task was no longer significant after including the mediators (effect = -.01, 95% CI [-.12, .09]). See

*Indicates a significant 95% confidence interval. **

Model | B | ^{†} |
β | %C | 95% CI for B |
||
---|---|---|---|---|---|---|---|

Model without mediators | |||||||

AMAS → Price Comp (c) | -.16 | .06 | -.20 | .004 | -.27 | -.05 | |

^{2} AMAS → Price Comp |
.06 | .006 | |||||

Model with BMA-3, Price Calc, NFC, CRT-2 as mediators | |||||||

AMAS → BMA-3 (a) | -.29 | .07 | -.26 | < .001 | -.43 | -.14 | |

AMAS → Price Calc (a) | -.54 | .14 | -.25 | < .001 | -.83 | -.26 | |

AMAS → NFC (a) | -.20 | .05 | -.26 | < .001 | -.30 | -.10 | |

AMAS → CRT-2 (a) | -.41 | .11 | -.25 | < .001 | -.62 | -.19 | |

BMA-3 → Price Comp (b) | .13 | .05 | .18 | .005 | .04 | .22 | |

Price Calc → Price Comp (b) | .08 | .02 | .22 | < .001 | .04 | .13 | |

NFC → Price Comp (b) | .06 | .06 | .05 | .355 | -.06 | .18 | |

CRT2 → Price Comp (b) | .14 | .03 | .27 | < .001 | .08 | .12 | |

AMAS → Price Comp (c’) | -.01 | .05 | -.02 | .792 | -.12 | .09 | |

Indirect effects (a*b) | |||||||

AMAS → BMA-3 → Price Comp | -.04 | .02^{†} |
-.05 | .23 | -.08 | -.01 | |

AMAS → Price Calc → Price Comp | -.04 | .02^{†} |
-.05 | .27 | -.08 | -.02 | |

AMAS → NFC → Price Comp | -.01 | .01^{†} |
-.01 | .07 | -.04 | .01 | |

AMAS → CRT-2 → Price Comp | -.06 | .02^{†} |
-.07 | .34 | -.10 | -.02 | |

AMAS → Total → Price Comp | -.15 | .04^{†} |
-.18 | .91 | -.22 | -.08 | |

^{2} |
.26 | < .001 |

Math anxiety is a widespread phenomenon and, recently, researchers have begun to explore relations between math anxiety and performance on math-related tasks that goes beyond what happens in a math classroom or on a standardized test. The results from the current study add to the growing body of literature indicating that math anxiety is relevant in a host of everyday situations involving math. In support of our first hypothesis, we demonstrated a novel finding that consumers who experience higher levels of math anxiety made worse decisions on our Price Comparison Task compared to people lower in math anxiety. That is, when asked to determine which product represented the better deal (i.e., cost less per metric unit) between two products, consumers higher in math anxiety were less likely to correctly select the product that provided superior savings. Importantly, we went beyond simply demonstrating that math anxiety relates to performance on this task. We also tested a series of theoretically driven hypotheses geared at understanding the mechanism underlying this negative relation. We found that math competency, price calculation ability, and cognitive reflection–but not NFC–each significantly mediated the relation between math anxiety and performance on the Price Comparison Task. After considering all mediators, there was no longer a significant direct effect between math anxiety and task performance.

While novel, the current findings are nonetheless in line with previous research on math anxiety and its relation to math competency, NFC, cognitive reflection, and consumer behaviour. We replicated the well-established finding that math anxiety is negatively related to math competency (

Consistent with this theory, and in support of our second hypothesis, both price calculation ability (as measured by the Price Calculation Task) and math competency (as measured by the Brief Math Assessment 3) were significant independent mediators of the relation between math anxiety and performance on the Price Comparison Task. Indeed, math competency and price calculation ability were not as highly correlated as we expected (

Further, when looking at the distribution of scores on the Price Calculation Task, 49 participants (19.1% of the sample) obtained an accuracy score of 15% or lower whereas 179 participants (69.9% of the sample) obtained an accuracy score of 75% or higher. We interpret this to mean that the people in the low-scoring group either do not know which procedure to apply or have extremely poor calculation abilities. People in the high-scoring group, on the other hand, know the procedure but occasionally made calculation errors. It is likely that calculation ability is detecting variability between people who know the correct procedure versus those who do not, whereas math competency is detecting variability between participants who know which procedure to apply but differ with respect to their other mathematical abilities (e.g., conceptual understanding). An important next step would be to understand the procedure participants apply when deciding which product to select.

While we did not find support for our third hypothesis (i.e., that NFC would mediate the relation between math anxiety and performance on the Price Comparison Task), we did replicate the finding that math anxiety is negatively related to NFC as reported by Maloney and Retanal (

We further replicated the finding that people higher in math anxiety are less reflective in their thinking (

The current research has important implications for our understanding of the financial vulnerability of consumers with math anxiety. We demonstrate a novel finding that, when prompted, consumers higher in math anxiety are less likely to correctly select a product that costs less per unit compared to those lower in math anxiety. While the current results are based on a Price Comparison Task completed in the lab, this task was designed to parallel decisions about price comparisons that consumers encounter on a regular basis. Thus, assuming that consumers higher in math anxiety are also making errors when engaging in price comparisons in the aisles of their local grocery stores, then this research suggests that math anxiety may be associated with financial consequences above and beyond those already demonstrated in the literature (e.g.,

The present study had limitations. First, we are unable to draw causal links from our mediation analysis given the correlational design of our research. For example, the current data does not allow us to determine whether participants’ lower math competency and their inability to accurately compute the price was caused by a math anxiety induced transient reduction in working memory resources (

In the present study we established a negative relation between math anxiety and performance on a novel Price Comparison Task and investigated possible mediators of this relation. The negative relation between math anxiety and task performance can be partially explained by the evidence suggesting that people higher in math anxiety are less capable of performing the necessary computations to discern the most cost-effective deal, are lower in math competency in general, and are less likely to engage in reflective thinking (i.e., verify that their calculation is correct). Considering the prevalence of math anxiety, future research should continue to investigate how math anxiety impacts consumers’ daily decisions. Potential means of mitigating these negative impacts include bolstering calculation and math competencies as well as a willingness to override intuitive responses.

Due to a programming error, responses to Item Two of the BMA-3 did not record for participants recruited through the university student pool (

Work on this project was funded by an NSERC DG and a SSHRC IDG to EAM and an OGS and SSHRC Joseph-Armand Bombardier CGSD to AS.

For this article, a data set is freely available (

The Supplementary Materials contain the following items (for access see

all materials for the Price Comparison and Calculation Tasks

the research data for this study

The authors have declared that no competing interests exist.

The authors have no additional (i.e., non-financial) support to report.