Baruch Fischhoff (email@example.com) is the Howard Heinz University Professor in the Department of Social and Decision Sciences and the Department of Engineering and Public Policy of Carnegie Mellon University
The views expressed are those of the authors and do not necessarily reflect those of the Federal Reserve Bank of New York. We thank Michael Bryan, Jeff Dominitz, Eric Johnson, Arthur Kennickell, Chuck Manski, Athanasios Orphanides, Simon Potter, Robert Rich and Ken Wolpin for their advice on this project, as well as Sandy Chien, Tim Colvin, Daniel Forman, Peter Fielding, Daniel Greenwald, Tania Gutsche, Mandy Holbrook and Bas Weerman for their help with conducting the research.
When financial decisions have consequences beyond the immediate future, individuals' economic success may depend on their ability to forecast the rate of inflation. Higher inflation expectations have been reported by individuals who are female, poorer, single and less educated. Our results suggest that these demographic differences in inflation expectations may be partially explained by variations in expectation formation and financial literacy. Specifically, higher inflation expectations were reported by individuals who focused more on how to cover their future expenses and on prices they pay (rather than on the US inflation rate) and by individuals with lower financial literacy.
In the course of everyday life, people make a variety of financial decisions about saving, investing and borrowing, among other things. When their effects extend into the future, financial decisions require accurate assessments of inflation rates. Inflation expectations have been studied by economists, psychologists, marketing scientists and others concerned about individuals' financial wellbeing and the impacts of their choices on the economy. Indeed, individuals' perceptions and expectations of inflation may affect actual realized inflation and other economy-wide outcomes (Katona 1975). A better understanding of these inflation expectations can help economists and central bankers to improve their forecasts of future macroeconomic trends and formulate monetary policy.
One explanation proposed for demographic differences in reported inflation expectations is that individuals from population groups who report higher inflation expectations also experience a relatively higher rate of inflation in their actual consumption. For example, the elderly may experience a higher rate of inflation due to their health care expenditures (Hobijn and Lagakos 2003; McGranahan and Paulson 2006). However, even though the rate of inflation varies widely across product categories,1 actual inflation experiences of individual households do not seem to vary much. For example, Hobijn et al. (2009) reported that between 1995 and 2005, annual inflation rates experienced varied by only .2% to .4% across different demographic groups (see also Kokosi 2000).
Here, we considered three possible explanations for demographic differences in inflation expectations. First, we examined whether individuals who report higher inflation expectations reveal systematic differences in how they form their inflation expectations. Bruine de Bruin et al. (2008) found considerable variability in the factors people considered when forming their inflation expectations. In addition to thinking about the US inflation rate, they also reported thinking about their personal experiences with prices they pay. Psychological theories suggest that larger price changes are usually more salient than smaller ones, and that increasing prices are usually more salient than decreasing or stable ones (Bruine de Bruin et al. 2008; Brachinger 2008; Fluch and Stix 2005; Jungermann et al. 2007; Kahneman and Tversky 1979). As a result, individuals who think relatively more about their personal experience with prices, compared with the US inflation rate, may give higher inflation expectations.
Second, we examined whether individuals who report higher inflation expectations have shorter financial planning horizons. Low-income populations tend to be more myopic when making financial decisions (e.g., Zikmund-Fisher and Parker 1999). Such a near-term focus could make them more sensitive to transient price shocks as well as less informed about the longer-term price trends captured in the inflation rate. As a result, they may experience more uncertainty about what levels of inflation to expect, leading to more volatile inflation expectations. Indeed, density forecasts have suggested less uncertainty about future inflation among men (vs. women), among individuals with (vs. without) a college education, for married (vs. single) individuals and for those with higher (vs. lower) levels of income—and that those with less uncertainty gave less volatile point estimates of inflation expectations (in terms of making smaller absolute revisions of their forecast) over time (VanderKlaauw et al. 2008).
Perhaps because inflation is more typical than deflation, inflation expectations tend to have an implicit lower bound of 0%, with only 3% of the 200,000 responses to the Michigan Survey of Consumers given between 1978 and 2004 being below 0% (Blanchflower and Coille 2009; Curtin 1996, 2006; Lombardelli and Saleheen 2003). The resulting floor effect would bias inflation expectations upward, among those whose translation process is noisier.
This study examined these three hypotheses, predicting that higher inflation expectations would be reported by individuals who (1) focused on their expenses and the prices they pay when forming their inflation expectations, (2) had shorter financial planning horizons and (3) had lower financial literacy. We further examined whether these relationships explained demographic differences in inflation expectations.
We conducted a Web-based survey with RAND's American life panel (ALP), whose members were recruited from respondents participating in the Michigan Survey of Consumers in 2007. These survey respondents were originally reached through random-digit dialing. Those who indicated willingness to participate in Web-based surveys and gave consent to have their information transferred were contacted by RAND and provided with Web TV if they did not have Internet access.
A random sample of 740 ALP panel members were invited to participate in the ALP's 16th monthly survey. Of those, 613 completed the survey (82.8% response rate). Our sample included 299 respondents randomly assigned to receive the questions analyzed here. The survey was fielded between December 22, 2007, and May 20, 2008; 41.8% completed it by December 31, 2007 and 84.6% by January 31, 2008. Ages ranged from 19 to 82 (M = 47.4, SD = 14.3), with a median of 48. In total, 54.5% were female, 70.6% were married or living with a partner, 59.9% had at least a bachelors degree and 86.6% were white. The median reported income range was $60,000 to $75,000, with 45.5% reporting incomes more than $75,000.
Over the entire period that the survey was online, the monthly samples of the Michigan Survey of Consumers included 53.4%–59.4% females, 58.0%–64.7% married or living with a partner, 64.4%–69.9% with at least a bachelors degree, 36.1%–43.6%% reporting income more than $75,000, 61.3%–64.4% aged 48 or older and 79.6%–83.7% white. Compared with the ranges observed for the Michigan sample, our sample was slightly more likely to be married or living with a partner, slightly less likely to have a college education, somewhat younger, and slightly more likely to be white.
Respondents received $20 for answering the entire Internet survey, which included the measures described below, and took about 35 minutes to complete. Although respondents were allowed to skip questions, those who tried to do so received a prompt encouraging them to provide an answer. Respondents reported their race, marital status, highest level of education completed, age, gender and total combined income2 across all family members over the past 12 months.
Respondents received an adaptation of the Michigan Survey of Consumers question (Curtin 1996): “During the next 12 months, do you think that prices in general will go up, or go down or stay where they are now?” with the response options of “Go up,”“Stay the same” and “Go down.” Those who responded “Stay the same” were asked whether they meant that prices would go up at the same rate, or that prices would not go up. Those who chose “go up at the same rate” were categorized as having indicated that prices would increase. Those who indicated expectations for prices to go up or down were asked by what percent, and to give “your best guess or your best guess for a range.” Those who provided only the lower or upper bound of a range were prompted to complete the other. Those who provided a range then were asked for a best guess. Here, we report only on the point estimates given as a “best guess.”
Following the Michigan Survey of Consumers procedure, respondents who gave a best guess greater than 5% were given the opportunity to revise their answer, with the prompt, “Let me make sure I have that correct. You said that you expect prices to go up during the next 12 months by [repeat response] percent. Is that correct?” Finally, those who did not give a best guess or a range were asked, “How many cents on the dollar do you expect prices to go [up/down] on the average, during the next 12 months?”
Forming Inflation Expectations
Respondents were asked what they thought the inflation expectations question was “asking for the most.” Response options (shown in Table 2) reflect topics mentioned in cognitive interviews asking participants to think aloud while generating inflation expectations (Bruine de Bruin et al. 2008). Respondents then rated how much they had thought about each topic, which they “may or may not have thought of” when generating their inflation expectations, on a scale from 1 (not at all) to 7 (very much).
Table 2. Topics Respondents Thought About When Forming Inflation Expectations
aRatings of how much respondents thought of these topics.
bFor each topic, a one-sample t-tests examined whether the mean rating was significantly different from the scale midpoint of 3.50.
cAn additional “other” option was used by 2.9% of respondents.
dExcept for ratings of annual raises, Levene's test for equality of variances showed significant group differences in variances for each of these ratings (p < .01). Hence, we used the nonparametric Mann–Whitney test (Siegel and Castellan 1988) to examine group differences in mean ratings.
Respondents were asked “In planning your [family's] spending, which of the following time periods is most important to you?” Response options ranged from the next day (= 1) to longer than 10 years (=10). A parallel question asked about decisions concerning how much income to save.
3. True/False: If the interest rate on your savings account is 1% per year and inflation is 2% per year, after one year, you will be able to buy more with the money in this account than you are able to buy today.
7. In the BIG BUCKS LOTTERY, the chances of winning a $10.00 prize are 1%. What is your best guess about how many people would win a $10.00 prize if 1,000 people each buy a single ticket from BIG BUCKS?
The mean reported expectation for prices in general over the next 12 months was 6.73% (SD = 9.02), with a median of 5.0%. Among the respondents who completed the survey by January 31, 2008, the median was 4.0%, slightly higher than the raw median of 3% and the imputed median of 3.4% observed with the Michigan Survey of Consumers for both December 2007 and January 2008. However, we cannot make a confident comparison, because we do not know the exact interpolation and sample weighting method used to compute Michigan's median from the individual, usually integer, responses or the imputation of missing responses (VanderKlaauw et al. 2008).3 The overall distribution (see Figure 1) showed strong positive skewness (5.33), indicating the mean was higher than the median, and strong positive kurtosis (43.04), suggesting a relatively flat and long-tailed distribution. We dealt with these extreme values in two ways, using both conceptualizations in all of the reported analyses, and finding similar patterns of results in most cases. First, we examined reported inflation expectations, after removing twelve extreme outliers, defined as values that exceeded the 75th percentile by more than three times the interquartile range (Frigge, Hoaglin, and Iglewicz 1989), here equal to 23%. Doing so reduced skewness (1.59) and kurtosis (2.78), as well as the mean (5.37) and standard deviation (4.47), with the median remaining at 5%. Second, we created a binary measure reflecting whether or not respondents gave inflation expectations greater than 5%, retaining all responses. We chose 5% as a threshold for unusually high expectations, because (1) the Michigan Survey of Consumers treats inflation expectations over 5% as suspect, offering respondents who report such expectations an opportunity to revise their answer (Curtin 1996), (2) the CPI has not been above 5% since 1990 (Bryan and Venkatu 2001b) and (3) median inflation expectations have not been above 5% since the mid-1980s (Bryan and Venkatu 2001b). Overall, 30.4% of our sample gave values greater than 5%.
Forming Inflation Expectations
Table 2 shows respondents' ratings of how much they thought about each of ten topics when forming their inflation expectations. The ratings are presented in decreasing order. Six were above the scale midpoint of 3.50, including ratings of how much respondents thought about prices they pay (t(298) = 23.51, p < .001), prices Americans pay (t(298) = 17.60, p < .001), changes in the cost of living (t(298) = 12.64, p < .001), the US inflation rate (t(297) = 9.78, p < .001), specific prices (t(298) = 2.51, p < .05) and how their life will change (t(297) = 1.84, p = .07). Although these ratings suggested that respondents thought about various topics when forming their inflation expectations, the majority seemed to interpret the inflation expectations question in ways consistent with economic definitions. When asked what the inflation expectations question was asking about the most, respondents selected prices Americans pay (39.5%), prices they pay (21.4%), changes in the cost of living (17.1%) and the US inflation rate (15.1%). Overall, respondents who selected a topic as the main focus of the inflation expectations question also gave higher ratings of how much they thought about it when forming their inflation expectation, suggesting consistency across measures. Subsequent analyses focused on the ratings of how much respondents thought about each topic when forming their expectations, rather than their choice of main question topic, because (1) how respondents formed their inflation expectations was the main focus of one of our hypotheses, (2) the ratings provided more information and (3) the ratings allowed us to conduct the factor analysis described below.
The ten ratings were highly correlated, with Pearson correlations ranging from .02 to .74 and a median correlation of .20. To reduce the large number of correlated ratings, we conducted a principal axis factor analysis with a varimax rotation to identify orthogonal factors. Table 3 shows the three resulting factors, which resembled those found by Bruine de Bruin et al. (2008). The first factor involved topics regarding respondents' personal financial situation, with the highest loading (.88) for ratings of how much respondents thought about how to cover their expenses next year. Other high loadings reflected how much they thought of how to pay for loans and debts as well as how their life will change over the next year. The second factor seemed to reflect general indicators of inflation, with the highest loading for ratings of how much respondents thought of the US inflation rate (.63), and the second highest loading for changes in the cost of living (.58). The highest loading on the third factor (.94) reflected ratings of how much respondents thought about the prices of things they usually spend money on. In the reported analyses, we represented each factor with the item that had the highest loading. Replacing it with the average across high-loading items for each factor did not affect the overall pattern of results discussed below.
Table 3. Factor Analysis on Ratings of How Much Respondents Thought About Topics When Forming Inflation Expectations
Factor 1: Personal Finances
Factor 2: General Indicators
Factor 3: Prices You Pay
Note: This table presents the structure matrix for the principal factor analysis with varimax rotation. For each factor, the highest loading is underlined.
Prices you pay
Prices Americans pay
Changes in cost of living
US inflation rate
How your life will change
How to cover expenses
Seasonal changes in prices
How to pay for loans and debts
Percent variance explained (%)
We computed Spearman rank correlations between reported inflation expectations and respondents' ratings of how much they thought about the topics representing the three factors. Ratings of how much respondents thought about how to cover expenses (Factor 1) were positively correlated with reported inflation expectations (rs = .19, p < .001) and with the binary measure of whether expectations were greater than 5% (rs = .15, p < .01). Ratings of how much respondents thought about prices they pay (Factor 3) also were positively correlated with reported inflation expectations (rs = .11, p < .05) and the binary measure (rs = .17, p < .05). In contrast, ratings of how much respondents thought about the US inflation rate (Factor 2) were not significantly correlated with reported expectations (rs = .07, p = .22) or the binary measure (rs = .00, p = .99).
Financial Planning Horizon
The two questions asking about respondents' planning horizons for spending and saving decisions had good internal consistency, with a Cronbach's (1951) alpha of .74. Responses to the two questions were averaged (M = 5.46, SD = 1.86). Individuals with lower composite scores, indicating shorter financial planning horizons, did not report significantly higher inflation expectations (rs = −.08, p = .20) but were more likely to report expectations greater than 5% (rs = −.13, p < .05).
Financial Literacy and Confidence
Table 1 reports descriptive statistics for responses to the financial literacy items. The percent of respondents answering each item correctly varied from 29.8% (Item 8) to 94.6% (Items 3 and 13). The items reliably measured a common underlying construct; Cronbach's (1951) alpha was .71. Removing one item (#4) with a very low item-total correlation increased Cronbach's alpha to .73. Removing additional items did not improve it further. A factor analysis with oblimin rotation, allowing correlated factors, suggested a three-factor solution (Tabachnick and Fidell 1989). The three factors had eigenvalues of 3.38, 1.62 and 1.27, explaining 22.5%, 10.8% and 8.5% of the variance, respectively. However, the factors had no ready interpretation, considering their topic, source (Table 1) or response mode. As a result, our analyses used one overall score reflecting the percent of correct answers to the items shown in Table 1 (excluding Item #4).
Financial confidence was calculated across the four of the five true/false financial literacy questions (excluding Item #4).4 Cronbach's (1951) alpha was .78, indicating good internal consistency. Mean confidence was 85.3% (SD = 12.2), only slightly higher than the 83.7% (SD = 20.2) correct responses across these same items, indicating appropriate confidence. Financial confidence was significantly correlated with total financial literacy scores (rs = .33, p < .001), indicating that respondents who knew more also had greater confidence in their knowledge.
As expected, respondents with lower financial literacy scores reported higher inflation expectations (rs = −.21, p < .001) and were more likely to report expectations greater than 5% (rs = −.26, p < .001). However, those reporting lower financial confidence did not report higher inflation expectations (rs = −.07, p = .26) nor were they more likely to report inflation expectations greater than 5% (rs = −.05, p = .38).
Table 4 presents the demographic categories previously found to be correlated with inflation expectations. For education, income, age and financial literacy, we created two groups, reflecting individuals reporting values above and below the median of that variable. Levene's test for inequality of variances showed that there was significantly more disagreement between the inflation expectations of respondents with lower (vs. higher) levels of education (F(1,285) = 15.98, p < .001), lower (vs. higher) income (F(1,284) = 5.42, p < .05) and lower (vs. higher) financial literacy (F(1,285) = 21.98, p < .001) (Table 4). As a result, we used nonparametric Mann–Whitney (M–W) tests to compare groups' inflation expectations (Siegel and Castellan 1988). Chi-square tests examined group differences in reporting expectations greater than 5%. Spearman rank correlations examined relationships of both measures of inflation expectations with the continuous variables for age and financial literacy (Siegel and Castellan 1988).
Table 4. Descriptive Statistics by Demographic Groups and Financial Literacy
M (SD) of Ratings of Formation of Inflation Expectations
How to Cover Expenses
Prices You Pay
US Inflation Rate
Financial Planning Horizon
Note: Nonparametric Mann–Whitney tests (Siegel and Castellan 1988) examined group differences in continuous variables, and chi-square tests examined group differences in dichotomous variables. Significance levels for age and financial literacy were based on Spearman rank correlations with the full-range variables.
Consistent with previous research, inflation expectations were higher among respondents who were nonwhite (z = 1.81, p < .10), single (z = 1.65, p < .10), less educated (z = 1.96, p < .10) and poorer (z = 2.06, p < .05). Expectations greater than 5% were more likely for respondents who were less educated (χ2(1) = 7.22, p < .01) and had lower income (χ2(1) = 9.08, p < .01), but not for other demographic groups. We did not find significantly higher inflation expectations for women (p < .10), although results were in the expected direction.
Older adults reported higher inflation expectations (rs = .12, p < .10), with a seemingly monotonic trend over the four age quartiles, with means of 4.89 (SD = 4.79) for respondents younger than 36, 5.24 (SD = 4.11) for ages 36–46, 5.54 (SD = 4.49) for ages 47–57 and 5.76 (SD = 4.55) for those over 57. There was a similar pattern for the percent of respondents reporting expectations greater than 5% (23.5%, 29.3%, 36.4% and 31.7%, respectively). However, age was not significantly correlated with the binary measure (rs = .09, p > .10).
There also were some demographic differences in how respondents formed their inflation expectations, with singles (vs. those who were married or living with a partner) giving lower ratings for how much they thought about the US inflation rate (M–W z = −2.37, p < .001), and respondents with lower levels of education (M–W z = 3.05, p < .01) and income (M–W z = −1.80, p = .07) thinking more about how to cover their expenses. Single (M–W z = −3.52, p < .001), less educated (M–W z = −2.54, p < .05) and poorer (M–W z = −3.95, p < .001) respondents reported shorter financial planning horizons.
Most demographic groups reporting higher inflation expectations also had lower financial literacy, including respondents who were nonwhite (z = −1.77, p = .08), single (z = −1.77, p = .08), less educated (z = −5.79, p < .001), lower income (z = −4.26, p < .001) and female (z = −5.38, p < .001). Financial literacy was unrelated to age (p > .10). Similar group differences were observed for financial confidence (Table 4), which was lower for respondents who were single (z = −2.25, p < .05), had lower incomes (z = −3.42, p < .001), were younger (rs = .22, p < .001) and were female (z = −5.68, p < .001).
The last two rows of Table 4 correlate financial literacy with other measures. As reported above, respondents with lower financial literacy reported higher expectations (rs = −.21, p < .001) and were more likely to report expectations greater than 5% (rs = −.26, p < .001). Respondents with lower financial literacy thought more about covering expenses (rs = −.19, p < .01) and marginally less about the US inflation rate when forming inflation expectations (rs = .11, p = .06). They had significantly shorter financial planning horizons (rs = .21, p < .001) and lower financial confidence (rs = .38, p < .001).
Linear Regression Predicting Inflation Expectations
We conducted linear regression analyses to examine the relative contributions of the different variables. The left panel of Table 5 shows a linear regression predicting reported inflation expectations, adding demographic variables in Model 1, ratings for what respondents thought about when forming their inflation expectations and their financial planning horizons in Model 2, and financial literacy and confidence scores in Model 3.
In the first model (R2 = .05), education was the only demographic variable that remained significantly related to higher inflation expectations after controlling for the other demographic variables, with a marginal relationship for older age. The second model had greater predictive power (R2 = .10), with ratings of how much respondents thought about how to cover expenses and prices they pay being significantly related to higher inflation expectations, and education no longer being significant. The third model explained yet more variance (R2 = .15), finding significantly higher inflation expectations for respondents with lower financial literacy scores. Adding financial literacy further reduced the coefficients of some demographic variables, most notably education, income and gender. Thus, individuals' financial literacy, and how they form inflation expectations, may help to explain the relationship between demographic variables and inflation expectations.
The right panel of Table 5 shows the results of a logistic regression predicting the binary measurement of whether respondents reported inflation expectations greater than 5%. The results paralleled those of the linear regression, with education and income the significant demographic predictors in Model 1, and respondents' ratings of how much they thought about prices they pay (added in Model 2) and financial literacy scores (added in Model 3) explaining the relationship between demographic variables and giving inflation expectations greater than 5% (Table 5).5
We found support for three hypotheses as to why members of these demographic groups reported higher inflation expectations. As expected, inflation expectations were higher among individuals who thought relatively more about how to cover expenses and about specific prices when forming their inflation expectations and among individuals with lower levels of financial literacy. Similar relationships were observed whether the dependent variable reflected actual reported inflation expectations or a binary measure of whether inflation expectations were seemingly high, defined as greater than 5%. The expected relationship between shorter financial planning horizons and higher inflation expectations was found only for the binary measure of inflation expectations. Regression analyses controlling for multiple demographic predictors (reported in Table 5) showed independent contributions of each hypothesized relationship to reported inflation expectations, explaining previously reported demographic differences in inflation expectations.
In short, individuals who expect higher inflation may be thinking about different issues when forming their inflation expectations. Especially for respondents with lower levels of income and education, questions about inflation may trigger relatively stronger concerns about their personal financial experiences, relative to the US inflation rate. Because large price changes tend to be more salient than smaller ones, and increasing prices tend to be more salient than decreasing ones (Brachinger 2008; Fluch and Stix 2005; Jungermann et al. 2007), focusing on these issues would be expected to bias inflation expectations upward.
The additional contribution of financial literacy to higher inflation expectations might reflect the increased difficulty individuals with lower literacy have in forming such expectations. These individuals may be more uncertain, resulting in more volatile estimates of inflation over time (VanderKlaauw et al. 2008). Because inflation expectations appear to be bounded at 0% (Curtin 2006), increased volatility should produce higher estimates. Although we did not have the data to examine volatility in individuals' expectations over time, we did find more variable inflation expectations between respondents with lower financial literacy, which, among other things, may be explained by higher individual-level uncertainty.
Our data were collected at a time of relatively low inflation, when demographic differences in inflation expectations tend to be less strong (VanderKlaauw et al. 2008). Nonetheless, we replicated all but one of the demographic differences observed in previous studies (Bryan and Venkatu 2001a, 2001b; Jonung 1981). Although the trend was in the same direction as in other studies, we did not find significantly higher inflation expectations for women than for men.
We suspect that the relationships we found were inflated by the question used to measure respondents' inflation expectations, which asked about expected “prices in general.” The present study followed the standard question wording used on the well-respected Michigan Survey of Consumers. Our recent research has suggested that asking for expectations for the “rate of inflation” may be less likely to evoke biased thoughts of increasing prices (Bruine de Bruin et al. 2008).
Our results do not indicate whether the seemingly unrealistic inflation expectations reported by people with low financial literacy will affect their financial decisions. However, they do suggest that people with low financial literacy also have less confidence in their financial knowledge and shorter financial planning horizons (Table 4). Although these variables showed no significant relationship to their inflation expectations per se, they may affect related financial decisions. That is, having less financial confidence may make individuals feel they do not have the ability to make complicated financial decisions, such as those that extend into the future. Combined with their already shorter financial planning horizons, the result may be the avoidance of long-term financial planning. If so, individuals with low financial literacy may benefit from validated programs targeting their financial literacy and understanding of inflation (Fox, Bartholomae, and Lee 2005).
UK data for December 2006 showed that the overall annual consumer price index (CPI) increase of 3.0% included rates of −4.1% for clothing and footwear, 4.6% for food and 14.0% for education (Office for National Statistics 2008). US data for the same period showed that the overall CPI increase of 3.2% included rates of −0.7% for communication, 2.4% for food and 6.2% for education (Bureau of Labor Statistics 2008).
Income was specified as “including money from jobs, net income from business, farm, or rent, pensions, dividends, interest, social security payments and any other money income received by members of your family who are 15 years of age or older.”
Our sample was slightly less likely to have a college education, which may have contributed to the somewhat higher median inflation expectations reported in the Michigan Survey of Consumers. However, other unobserved differences between samples, as well as variations in survey administration (such as using a self-assisted online computer survey at the RAND ALP vs. telephone interviews at the Michigan Survey of Consumers) also may have played a role (VanderKlaauw et al. 2008).
Financial confidence was significantly related to self-ratings of financial knowledge (r = .19, p < .01). Replacing our measure of financial confidence with these self-ratings does not affect the results reported in Table 5 (α = .10).
For both regressions (Table 5), replacing ratings of how much respondents thought about how to cover future expenses, the US inflation rate and prices they pay with the corresponding dichotomous measures of whether or not respondents selected these topics as the main focus of the inflation expectations' question as predictors showed no significant relationship between selecting each of these topics and inflation expectations (p > .10), perhaps due to their lower variability. The results were, however, in the expected direction.
In a later study, some respondents (n = 261) indicated for seven items (including housing, food and transportation) whether they were in the top three items on which they spent money, as well as the percent of their yearly budget spent for each item in the top three. Adding a fourth model to the regression including whether each of the seven items were in the spending top three, or including the percent spent on the most common items reported by 160 respondents, did not improve reported predictions of inflation expectations (left panel, Table 5) or whether respondents expected inflation to be greater than 5% (right panel, Table 5) (p > .10).