CAPÍTULO II. EL CICLO VITAL
2.6 PERFILES PROFESIONALES PARA DESARROLLAR UNA ADECUADA EDUCACIÓN CON PERSONAS MAYORES
We now examine how inflation and real consumption per capita constructed with our market-based mVPI compares to alternative methods. The mVPI varies from a conven-tional method in both its treatment of geography and product variety changes. We first consider geography’s role when accounting for product variety changes by following the standard approach in the literature and focusing on variety changes at the national level using an aggregate variable goods price index (aVPI).29We construct the aVPI by assum-ing one national market and settassum-ing M = 1 in the mVPI model from Section 1.4. Sec-ond, we compare the mVPI against the conventional aFPI from Section 1.2.2. The aFPI tracks economy-wide price and spending changes among goods continuously-available anywhere in the country, overlooking geography and variety changes. Lastly, we con-struct a market-based chained Fisher price index (mFPI) that incorporates geography by
29Notable papers include Feenstra (1994), Broda and Weinstein (2010), Aghion et al. (2017) and Redding and Weinstein (2017).
allowing price and spending changes on continuing goods to vary across markets, but not variety changes.30 The mFPI informs the role of geography in conventional price indices.
We start by comparing inflation under each method.
Aggregate approaches to accounting for product variety changes that overlook geo-graphic diversity in consumption baskets understate inflation; nonetheless, incorporating variety changes is important. In Figure 1·7, we plot four-quarter inflation rates between 2005 and 2014 from our market-based mVPI and the aggregate aVPI, along with the con-ventional aFPI and mFPI that also allows for geography. Our mVPI captures average an-nual cost of living decreases of -1.18%. Overlooking geography causes the aVPI to over-state declines by -2.67%-points a year relative to the mVPI. Section 1.6 explores the aVPI and mVPI’s disparities in more detail. On the other hand, omitting variety changes alto-gether leads to an overstatement of inflation. The conventional aFPI finds that inflation averaged1.63% a year. MVPI inflation is also twice as volatile as the aFPI and counter-cyclical, whereas the aFPI is slightly procyclical. Lastly, the similarity between the mFPI and aFPI indicate that geography plays a limited role when overlooking variety changes.
Ignoring geographic differences in what consumers purchase cause aggregate approaches to accounting for product variety changes to overestimate real consumption growth. In Figure 2·1, we plot quarterly real consumption per capita using each price index from Fig-ure 1·7 paired with the same nominal consumption series from the HMS. The disparities in average inflation the various methods imply produce stark differences in trend growth.
Our market-based mVPI finds average annual real consumption growth of2.99% between 2004 and 2014. In comparison, the aggregate-level aVPI estimates5.73% a year growth, a 92% overstatement. Using the conventional aFPI and ignoring variety changes altogether suggests slight declines of -0.04% annually. Our baseline results are robust to numerous
30Specifically, we create separate Fisher price indices for each market between 2004 and 2014, then ag-gregate to a national series to construct the mFPI.
Figure 1·7: mVPI captures lower & more variable inflation than the aFPI
Note: Figure 1·7 plots four-quarter inflation rates from the Aggregate Fisher PI (mean=1.63%, variance=2.34%), Aggregate VPI (mean=-3.85%, variance=3.19%), Market-Based Fisher PI (mean=1.91%, variance=3.00%), and Market-Based VPI (mean=-1.18%, variance=4.79%).
alternative specifications ranging from restricting the set of consumption to varying the elasticities of substitution (see Appendix A.4). Overall, accurately measuring the gains to consumers from new products and choice expansion requires correctly specifying the markets that benefit from them.
Including more local products and services (i.e., restaurants, hospitals, housing, schools, etc.) in our household spending data would likely increase the disparity in real consump-tion growth between our market-based mVPI and the aggregate aVPI; conversely, includ-ing more national products would probably shirk the gap. The real consumption growth rates we present are only directly representative of the segment of household spending the HMS covers, specifically on in-home consumer goods. However, two simple thought experiments shed light on how adding additional spending categories might shape our re-sults. First, adding more local products and services spending would make consumption baskets look even more different geographically. The smaller the overlap in baskets across regions, the more ignoring geography and focusing on an aggregate basket overstates the
choices available to consumers. Thus, broadened our data to include more inherently local categories of spending will likely result in a greater overstatement of real consumption growth by the aVPI method relative to our market-based mVPI than the2.75 percentage points we find in the HMS. Second, following similar logic, adding more national goods to our dataset will make consumption baskets look more homogeneous across markets.
The less unique baskets are geographically, the smaller the overstatement in choice from focusing on an aggregate basket. Thus, using a dataset with more national goods would likely shrink the disparity between the aVPI and our mVPI found using the HMS. How-ever, the exact impact on our results from a broader dataset is uncertain.
We find evidence in the data that innovations in product quality and choice drive part of the faster real consumption growth the market-based mVPI captures that the conven-tional aFPI misses. First, new products introduced in 2014 had a23.38% higher quality than new goods in 2005 (see Appendix A.7 for details). Second, the average market con-sumed a4.69% wider variety of products in 2014 than in 2004. Overlooking the benefits to consumers from quality and choice improvements cause conventional methods to under-state real consumption growth. Moreover, the small difference between the conventional and market-based mFPI and aFPI indicates that geography becomes important once one takes variety changes into account.
Accurately measuring real consumption volatility requires accounting for the pro-cyclicality of product variety changes at the market level. Our market-based mVPI finds that around-trend variance was 0.47%, nearly double the 0.25% the conventional aFPI suggests. Volatility rises because within-market procyclicality in variety makes reces-sions look worse and booms look better than when overlooking product variety changes altogether.31However, ignoring geographic diversity in consumption baskets impacts real
31For example, the coincidence of declines in variety and product entry with decreases in spending during a recession amplifies the contraction in real consumption while the opposite occurs during an expansion.
consumption differently during booms than in busts, by overvaluing local entry but un-dervaluing local exit. Quantitatively, the aVPI suggests that around-trend variance was 0.57%, or 21% higher than our mVPI indicates. Thus, correctly measuring real con-sumption volatility, just as with growth, necessitates incorporating both product variety changes and geography differences.
Figure 1·8: Cost of living adjustments omitting geography overstate growth
Note: Figure 2·1 plots quarterly real consumption per capita time series as indices with a base year of 2004 constructed using the Aggregate Fisher PI (aFPI), Market-Based Fisher PI (mFPI), Market-Based VPI (mVPI), and Aggregate VPI (aVPI). We also plot each series’ log-linear fitted trend. All series use the same nominal consumption per capita measure, differing only in their price indices. The graph begins at 2004.q4 as each quarter in 2004 provides separate base periods for the price indices, which we construct from four-quarter changes.
We also consider grouping households by educational attainment instead of markets and construct an analogous version of our market-based mVPI in Appendix A.4.4 to test whether markets are the correct type of disaggregation. We would expect the house-hold characteristics that most accurately reflect the set of purchased products to report the slowest relative real consumption growth under our disaggregated VPI method. An education-based household grouping suffers from an aggregation bias similar to the aVPI.
Grouping households jointly by education and Census region nearly recovers the mVPI-based real consumption series, instilling confidence that market-mVPI-based groupings provide the correct type of disaggregation.32
This section’s results indicate that accounting for product variety at the market level is crucial to avoid overstating real consumption growth and volatility. We next analyze the sources of bias an aggregate approach introduces into cost of living measures.