Table 1. Odds ratios (95% confidence interval) from multilevel models predicting low numeracy.
| Model 1A: Incorrect or no answer (vs. correct answer)a | Model 1B: 10% is bigger than 1 in 10 (vs. correct answer) | Model 1B: 10% is smaller than 1 in 10 (vs. correct answer) | Model 1B: No answer (vs. correct answer) | ||
|---|---|---|---|---|---|
| World Bank country income categories | |||||
| Low-income (vs. high-income) | 5.47*** | 7.69*** | 3.22*** | 5.61*** | |
| (3.98, 7.50) | (5.22, 11.32) | (2.36, 4.40) | (3.83, 8.21) | ||
| p<0.001 | p<0.001 | p<0.001 | p<0.001 | ||
| Lower middle income (vs. high-income) | 3.54*** | 3.98*** | 2.44*** | 3.78*** | |
| (2.57, 4.88) | (2.89, 5.47) | (1.87, 3.17) | (2.54, 5.63) | ||
| p<0.001 | p<0.001 | p<0.001 | p<0.001 | ||
| Upper middle income (vs. high-income) | 2.34*** | 2.51*** | 1.77*** | 2.46*** | |
| (1.70, 3.22) | (1.79, 3.53) | (1.39, 2.24) | (1.65, 3.67) | ||
| p<0.001 | p<0.001 | p<0.001 | p<0.001 | ||
| Demographic variables | |||||
| Up to elementary school (vs college) | 4.30*** | 2.27*** | 2.09*** | 6.47*** | |
| (3.58, 5.17) | (1.80, 2.85) | (1.54, 2.83) | (5.07, 8.26) | ||
| p<0.001 | p<0.001 | p<0.001 | p<0.001 | ||
| High school (vs. college) | 2.11*** | 1.61*** | 1.27* | 2.78*** | |
| (1.73, 2.56) | (1.28, 2.03) | (1.01, 1.60) | (2.24, 3.44) | ||
| p<0.001 | p<0.001 | p = 0.04 | p<0.001 | ||
| Female (vs. male) | 1.31*** | 1.03 | 1.20 | 1.44*** | |
| (1.16, 1.48) | (.88, 1.19) | (.99, 1.47) | (1.27, 1.63) | ||
| p<0.001 | p = 0.75 | p = 0.07 | p<0.001 | ||
| Age (divided by 10) | 1.16*** | 1.10*** | 1.09*** | 1.19** | |
| (1.07, 1.26) | (1.05, 1.16) | (1.05, 1.13) | (1.07. 1.32) | ||
| p<0.001 | p<0.001 | p<0.001 | p = 0.002 | ||
| N | 151,712 | 151,712 | |||
| Fixed effects ANOVA | F(7, 151704) = 122.47*** | F(21, 151688) = 118.34*** | |||
| AIC | 5,350,468,634 | 15,998,386,296 | |||
| BIC | 5,350,468,644 | 15,998,386,326 | |||
aLow numeracy was defined as failing to provide a correct answer to the basic numeracy question, and giving one of the incorrect answers or no answer instead. P-values significant at ***p<0.001, **p<0.001, and *p<0.05. Model 1A represents multilevel logistic regression, Model 1B represents multilevel multinomial regression. Model fit was better for Model 1A than for Model 1B, seen in lower values of AIC = Akaike Information Criterion, corrected and BIC = Bayesian Information Criterion. According to the World Bank’s classification, low-income countries have a per capita gross national income of less than $1,026, lower middle income countries of $1,026-$3,995, upper middle income countries of $3,996-$12,375, and high-income countries of more than $12,375 [23].