Perceived Inflation in Germany: Determinants and the Role of the Basket of Goods

2.1 Data

The data on perceived inflation used in this paper are based on the representative OpinionTRAIN survey for Germany, which has been conducted in five waves since 2020. Appendices A.1 and A.2 provide a detailed description of the survey as well as of the resulting data set. In essence, the survey data contain responses from nearly 2,000 individuals who were asked about various aspects of their professional and personal lives. In the two most recent survey waves conducted in April 2022 and April 2023, explicit questions about perceived price changes as well as about consumption and saving behavior were included. In the 2023 survey, participants were additionally asked to specify their monthly expenses across different consumption categories and to assess the price changes within each of the categories over the past 12 months. The descriptive analyses in Sections 3 and 4.1 include responses from 1,888 participants who provided complete and plausible answers regarding subjective inflation and expenditure structure. Due to missing covariates, the number of observations used for the regression analysis in Section 4.2 decreases to 1,723 (see Appendix A.2 on the issue of selection and non-response bias).

2.2 Descriptive Statistics

Table B.1 in the Appendix summarizes the unweighted characteristics of the respondents surveyed in April 2023. The age range spans from 18 to 87 years with an average age of 49.4 years. Men and women are represented almost equally in the data. About four-fifths of the respondents live in Western Germany. Less than 10 % receive social assistance benefits. Regarding monthly household net income, the largest group earns between 2,000 and 3,000 Euros with the distribution being right-skewed, which is in line with the overall population. More than half of the respondents hold a college or high school degree as their highest level of education, while nearly 17 % exhibit only basic schooling with or without an apprenticeship. 6 % of respondents are currently in training and 58 % are employed full or part-time. 28 % are retired and approximately 9 % are neither in training, employed, nor retired. With regard to city size, all types of municipalities in Germany are represented approximately equally.

4.1 Heterogeneity in Inflation Perceptions Between Income Groups

A central question in the design of fiscal and distributional policies during times of high inflation is whether certain population groups are more than proportionately affected by rising prices compared to others. With regard to income, the literature unanimously shows that lower-income households are on average more strongly hit by inflationary pressures than households with higher incomes (see, e.g., European Central Bank (2022) for Europe, and Penn Wharton (2021) for the US). According to both studies, this can partly be assigned to the fact that low-income households spend a larger share on goods that are subject to particularly high price increases and at the same time difficult if not impossible to substitute (like food, energy, and shelter). While we cannot directly measure whether lower-income households are de facto more severely affected by rising inflation rates, based on the survey data we are able to examine whether these households feel more affected by rising prices.^[1] To do so, we have segmented the responses to the question of how much respondents perceive prices to have changed over the past 12 months by household net income as well as by perceived changes in their expenditures on food. The dark blue bars in Figure 2 show the share of respondents in 2023 who have reported a value above the official inflation rate of 7 %, light blue bars the opposite. Red dashed lines show the respective value for the previous year.

Figure 2:

Subjective inflation by population groups (April 2023). The figure shows the perceived inflation rate by population group based on the answers to the question ‘By how many percent do you think consumer prices have gone up/down over the past 12 months?’. Dark blue bars in the histogram indicate the share of respondents who reported a perceived inflation rate above the official rate of 7 %, while light blue bars represent the opposite. Red dashed lines provide the values from April 2022. Source: exeo Strategic Consulting AG / Rogator AG; own calculation.

The following insights can be derived from Figure 2. With values between 84 and 93 %, the share of persons reporting a subjective inflation rate above the official rate of about 7 % is high across all groups. The group-specific averages range between 18 and 25 %. Hence, the high inflation rates in 2022 and 2023 seem to subjectively affect all income groups. Notably, the share of persons who report a perceived rise in prices above the official inflation rate has increased throughout. At the same time, despite similarities in average values and proportions, moderate differences exist between population groups. With a value of 25 %, persons living in low-income households report the highest average perceived inflation rate. In addition, individuals who indicate that their expenditure on food had strongly increased report the second-highest rate at 23 %.^[2] In the next section, we examine whether this first indication that persons living in lower-income households feel more than proportionally affected by rising prices continues to hold when other individual characteristics are being controlled for.

4.2 Individual Determinants

Intuitively, it seems unlikely that inflation perceptions are uniformly distributed across the population. Consistently, the theoretical literature has argued that they are likely to differ along various dimensions.^[3] D’Acunto et al. (2023) among others explain that historical experiences and in particular exposure to earlier periods of inflation may be relevant for the formation of subjective views on current price changes. Similarly, an understanding of the institutional setup of the financial system as well as the capacity to process information about monetary policies (termed ‘financial literacy’ by Bruine de Bruin et al. (2010)) may influence inflation perceptions. In addition, persons with higher incomes may feel less affected by inflation as they face less difficulties in financially compensating higher prices (D’Acunto et al. 2016). One somewhat special dimension in this context is the one of gender as there is ex ante no theoretical rationale why men and women should perceive changes in prices differently. Despite this, the empirical literature has consistently found that women tend to report higher subjective inflation rates than men. We discuss potential reasons for this finding in the context of presenting our own results below. Finally, regional differences may prevail between different regions due to local variations in the underlying basket of goods or the magnitude of price changes for certain goods and services (e.g., housing and mobility).

Since the seminal contribution by Jonung (1981), the empirical literature has confirmed the relevance of each of these factors. Studies for the United States by D’Acunto et al. (2023), Weber et al. (2022), and Angelico and Di Giacomo (2024) among others provide evidence that perceived inflation decreases with age, education level, and household income, and is generally higher among women than men. For Germany, Treu and Hartwig (2022) report similar findings. Taking the theoretical considerations as well as these empirical findings as a point of departure, we examine the relationship between subjective inflation and individual characteristics for Germany for the year 2023 by means of the following regression model:

where the dependent variable I _i is the perceived inflation rate. X denotes the matrix of individual characteristics, which includes information on age, employment status, net household income, educational attainment, gender, and city size. β represents the vector of estimated coefficients, α is the intercept, and ϵ _i the error term.

The first column in Table 2 provides the results from the regression. The significance of the coefficients shows that consistent with the existing literature individual age, gender, and education have a significant influence on perceived changes in prices. When interpreting the size of the coefficients, it needs to be kept in mind that age is measured in full years, while all other variables are of categorical nature. The reference categories for each of the binary variables are described at the bottom of the table. The results indicate that each additional year of age is associated with a 0.11 % point decrease in perceived inflation.^[4] Hence, an increase in age by one decade lowers inflation perceptions on average by 1.1 % points. This finding provides support for the theoretical argument by D’Acunto et al. (2023) that lifetime experience and exposure to different economic phases support the formation of a more realistic perception of the current economic situation. Furthermore, in line with the existing literature, the results indicate that perceived inflation is on average 3.6 % points higher for women than for men. A number of contributions have aimed to rationalize this finding. One of the key arguments relates the result to the ‘Availability Hypothesis’ formulated by Kahneman and Tversky (1973), which states that price increases for frequently purchased goods are assigned a greater weight (we test this notion empirically in Section 4.3). Along these lines, D’Acunto et al. (2022) show that the gender gap in inflation perceptions exists only within households where women do the groceries while it disappears if both spouses equally share this task (see D’Acunto et al. (2023) for an overview of the related literature). Regarding the individual level of education, the results in Table 2 emphasize the profound correlation of this variable with perceived inflation. Subjective inflation declines monotonically with rising formal education and is about 11 % points lower for college graduates than for individuals with only a basic school diploma. As shown by Hayo and Neuenkirch (2018), a higher level of education correlates with a deeper knowledge about economic relations including monetary policies and the role of Central Banks. Hence, the education variables included in the regression can be regarded as proxies for this (at least in our survey data) unobservable economic understanding (see Hayo (2023) for empirical evidence on this relation). In light of the discussion on differential perception of inflation by income in Section 4.1, it is noteworthy that net household income does not exert an independent influence on perceived inflation when the individual level of education is being controlled for. This suggests that differences in the perceived inflation rate are primarily related to varying levels of financial literacy (see Bruine de Bruin et al. (2010)), which are correlated with household income. This relationship is somewhat different with regard to receiving social assistance, which has an independent effect on inflation perception amounting to about 4 % points even when controlling for education level.

As a final step in this section, we compare the regression results for the year 2023 with the ones for 2022. While the nature of the survey as repeated cross-sections inhibits panel analyses with individual fixed effects, a comparison of the estimates obtained separately for the two waves allows to examine whether the results are qualitatively stable over time. Table B.2 in the Appendix contains the regression results for the year 2022. Overall, while the coefficients are estimated less precisely, the results point in the same direction as for 2023. Consistently, since perceived inflation is on average 10 % points lower in 2022 than in 2023, the coefficients are smaller for the earlier year. As in 2023, perceived inflation in 2022 tends to fall with age, household income and education and is higher for women than for men. It is surprising to see, however, that in contrast to 2023 the level of schooling does not offset the effects from household income. Despite this, it is reassuring that the results are overall qualitatively similar and also in line with the existing theoretical and empirical literature.

4.3 Subjective Inflation and Consumption Structure

A related discussion in the literature pertains to the question whether a relationship exists between individual consumption structure and subjective inflation. The idea goes back to the ‘Availability Hypothesis’ (Kahneman and Tversky 1973), which if applied to the context of inflation states that inflation perceptions are determined first and foremost by changes in the prices of those goods that are purchased most frequently.^[5] The assumed underlying mechanism is that a higher exposure to price signals for those items that are bought more frequently like, e.g., groceries, will lead to a higher subjective weight of these goods in the formation of inflation perceptions. Several empirical studies provide evidence in support of this argument. Among others, Brachinger (2008), Vogel et al. (2009), Del Missier et al. (2016), and D’Acunto et al. (2021) show that perceived inflation rises with the frequency of purchases and not with the share of expenditures within single categories of goods in individual budgets.^[6] Contributing to this debate, we examine the same question for Germany based on the survey data for the year 2023. As outlined in Section 2, this wave contains differentiated questions about perceived changes in prices and expenditure shares by categories of consumer goods. These categories encompass ‘Housing’, ‘Food’, ‘Mobility’, ‘Leisure’, ‘Lifestyle’ and ‘Other Expenditures’.^[7]

Columns (I) and (II) in Table 4 show average expenses as well as the perceived inflation rate per category. Costs for ‘Housing’ (43.5 %) and ‘Food’ (26.1 %) are mentioned as by far the most financially significant categories. Together, they account for almost 70 % of average individual expenses. Costs for ‘Mobility’ (10.6 %), ‘Lifestyle’ (9.4 %), ‘Leisure’ (7.5 %), and ‘Other’ (2.9 %) together contribute the remaining 30 %. At the same time, with the exception of ‘Lifestyle’ and ‘Other’, differences in perceived inflation across categories are relatively moderate. Notably, however, all of them range far above the official rate of 7 %. The highest perceived inflation rates are reported within the categories ‘Food’ (24.5 %) and ‘Housing’ (21.7 %), which is in line with sharply increased prices for groceries and heating during the period of observation (Möhrle and Wibault 2023). Column (III) in Table 4 provides the correlation between expenditure shares and perceived inflation at the individual level. With the exception of the category ‘Other’, this relationship is practically nonexistent. Together with the fact that ‘Food’ exhibits the highest perceived inflation rate, these results support earlier findings regarding the importance of purchase frequencies for the formation of individual inflation perceptions.

Complementing this analysis, we examine in a final step whether individual determinants of perceived inflation vary across expenditure categories. The remaining columns in Table 2 and Table 3 provide the results from estimating Equation (1) separately for each category of goods. The significance of the coefficients is mostly similar across the regressions. Deviations exist, however, in the following respects: the age-specific differences observed elsewhere are not evident for ‘Housing’ and ‘Lifestyle’ (e.g., clothing), i.e. price changes in these categories are perceived in a similar way by all age groups. With regard to gender-specific differences, the same applies to ‘Mobility’ and ‘Leisure’ (culture, sports, and travel). Here, too, men and women show similar perceptions. Of further interest is the finding that the factor ‘City Size’ plays a role only in the context of mobility. Specifically, perceived price increases in cities with more than 100,000 inhabitants are, on average, about 4 % points lower than in medium-sized and small cities. This may be due to a higher availability of public transportation in urban areas, for which prices have remained largely stable. Soaring prices for fuel, which are of higher significance for mobility in rural areas, have in turn been shown to unfold a particularly strong influence on inflation perceptions (Binder 2018).

4.4 The Role of Different Consumer Baskets

The results obtained in Section 3 show that perceived inflation has exceeded the official rate in April 2023 by a factor of about three. At face value, this deviation indicates that individuals perceive the rise in prices on average as three times larger than it is measured by official statistics. Since the calculation of the official inflation rate is based on a fixed basket of goods, this deviation may, however, also partly be driven by an imperfect match between the official basket of goods and current patterns of consumption (Zekaite 2020). In light of substantial shifts in purchasing behavior as a result of the COVID-19 pandemic reported by Hagenkort-Rieger (2020) and others, this may be in particular true for the period under investigation. In addition, Mai and Egner (2023) show that already small shifts in the composition of the basket can significantly alter the measured inflation rate. Contributing to this debate, we describe in a first step to whether individual consumption structures reported in the survey deviate from the official basket of goods. Thereafter, we examine to which extent such differences contribute to the deviation between official and perceived inflation rates by calculating various combinations of consumer baskets and category-specific inflation rates.

A comparison of columns (I) and (IV) in Table 4 shows that the deviation between subjective expenditure share and the official basket weight is largest for the categories ‘Housing’ (+10.8 % points), ‘Food’ (+10.7 % points), and ‘Leisure’ (−7.6 % points). At the same time, in line with the results from Section 4.3, the perceived inflation rates in column (II) exceed the official rates in column (V) across all categories. Based on these data, we examine the relative importance of deviations of actual consumption from basket weights on the one hand and differences in category-specific inflation rates on the other hand. To do so, we calculate all possible combinations of baskets and inflation rates and examine to which extent this alters the overall inflation rate. Table 5 provides the results. The first row (‘Scenario 1’) shows the inflation rate for April 2023 calculated from official numbers on price increases and weights by category. Importantly, replacing the official weights of the consumer basket with the expenditure shares by category as they are reported in the survey data yields almost no change in the inflation rate (comparison between Scenarios 1 and 2). A much larger shift of around 11 % points occurs, in contrast, when the official basket weights are combined with subjective inflation rates (‘Scenario 3’). Finally, combining subjective inflation rates with the expenditure shares from the survey data yields an overall subjective inflation rate of 21.1 % (‘Scenario 4’), which closely aligns with the average reported subjective inflation rate of 23.2 % (‘Scenario 5’). The key insight from these scenarios is that differences in the underlying baskets of goods have only a negligible influence on the gap between perceived and official inflation rates. Instead, the deviation between both rates is almost exclusively driven by consistently higher reported inflation rates across all spending categories as compared to officially measured numbers.