The CPI is the headline measure of inflation in most economies and is used to index wages, pensions and tax brackets. Whether it is RELIABLE as a measure of the cost of living depends on how closely its construction mirrors the actual spending patterns of the household in question.
Case supporting the view — CPI is unreliable. Four well-documented biases pull the index away from a true cost-of-living measure. First, substitution bias: when the price of a good in the basket rises, households rationally shift to cheaper substitutes. The CPI's fixed basket does not capture this, so it over-states the cost impact — households are not actually paying the higher price because they have switched. Second, outdated weights: the expenditure survey underlying the weights is run infrequently, so the basket reflects spending patterns from several years earlier; emerging categories (streaming services, electric vehicles) are under-weighted and declining ones (fixed-line telephones, physical media) are over-weighted. Third, quality change: when a product improves at the same money price — faster processors, safer cars, better pharmaceuticals — the CPI records 'no change' when consumers are actually getting more for their money (an implicit price fall). Fourth, new products enter the basket only after a lag, so their rapid early-life price falls are missed. Each of these biases is small, but they typically push the measured inflation rate ABOVE the true cost-of-living change. On this reading, CPI is at best an imperfect proxy and at worst a systematic over-estimate.
Case against the view — CPI is adequately reliable. Four counter-considerations soften the critique. First, the magnitudes are small: careful studies (e.g. the US Boskin Commission) estimated the combined bias at around 1-1.5 percentage points per year, which is modest compared with the inflation rates of concern. Second, statistical agencies actively mitigate the biases: baskets and weights are revised periodically (annually in many countries), hedonic methods adjust for quality changes in computers and other technology products, and chain-linking techniques limit substitution bias. Third, the CPI is transparent and reproducible — its construction is well-documented, which makes it hard for governments to manipulate and allows comparisons across countries and over time. Fourth, no alternative is clearly BETTER for the headline macro purpose: the Harmonised Index (HICP), the GDP deflator and the Personal Consumption Expenditure deflator all share similar limitations, and household-specific indices sacrifice comparability.
Different households, different experiences. The strongest version of the critique is that the CPI measures the AVERAGE household's cost of living, but no actual household IS the average. A low-income household spends a larger share of its income on food and energy, which are often more volatile and often rise faster than the headline index; a retired household faces a different basket again. So even if the CPI is reliable for its construction purpose (an aggregate index), it may be unreliable as a guide to any particular household's experience. Many statistical agencies now publish sub-indices (CPI for pensioners, CPI for low-income households) to address this.
Reasoned judgement. The view that the CPI is 'unreliable' is too strong in its unqualified form. The CPI is a well-constructed index whose biases are small, well-understood and actively mitigated. It is the best available single measure of changes in the general price level and remains fit for its primary macroeconomic purpose. However, it is NOT a precise measure of ANY specific household's cost of living, because households differ in their spending patterns; nor is it immune to the structural biases listed, which do tend to over-state inflation on balance. The strongest formulation is therefore: the CPI is RELIABLE as an aggregate indicator of changes in the general price level and as a basis for indexation and policy, but LESS reliable as a cost-of-living measure for atypical households (low income, retired, high housing costs) — for whom supplementary sub-indices are more appropriate.