Inflation Bias in the Czech Republic Randall K

Inflation Bias in the Czech Republic Randall K. Filer, CUNY and CERGEEI Jan Hanousek, CERGE-EI

Motivation: Real GDP during transition

Motivation 2: Inflation versus GDP growth

True measure of inflation • Cost of Living Index: change in prices with a fixed level of consumer utility • Consumer price index: change in prices with a fixed consumer basket

Practice of measuring inflation • Advantages of CPI (Laspeyres indx) – Low data requirements, enabling timely and routine calculation. – Consistent at various levels of aggregation • Disadvantages: CLI (fixed utility) is a more relevant concept for public policy purposes

Inflation bias - U. S. studies • > 1 percentage point p. a. : Gordon (1995), Boskin at al (1995), Darby (1995) and Diewert (1995) • = 1 percentage point p. a. : Shapiro and Wilcox (1996), Jorgenson (1995), Advisory commission (1995) and Gliliches (1995) • < 1 percentage point p. a. : Wynne and Sigalla (1994), Lebow, Roberts and Stockton (1994), Pakes (1995), CBO(1995)

Bias studies outside the U. S • • U. K. [Oulton (1996)] Sweden [Dahlen (1994)] Australia [Woolford (1994)] Germany [Hoffmann (1999)]

Effect of Bias on GDP Growth

The main sources of inflationary bias • • Consumer Substitution Bias Outlet Substitution Bias New Goods Bias Quality improvement bias

Formula bias Relative price change between period T and T-1 where

BLS “seasoning” method Relative price change between period T and T-1 where

Summary of the BLS studies • For link method: – Higher variance in P(j, L) is expected to cause a greater level of formula bias (for example by increasing covariance between 1/P(j, L) and P(j, L). – Bias is positive: For a given P(j, T) and P(j, B) and increase in P(j, L) decreases P(j, T)/P(j, L) while increasing P(j, L)/P(j, B)

Laspeyres and Paasche price indexes

Example 1 - Substitution

Example 2 - External shocks

Czech CPI weights changes

CPI 1995: 1993 vs. 1995 w.

CPI 1993 vs. 1999 w

Substitution bias: results

The main sources of inflationary bias • • Consumer Substitution Bias Outlet Substitution Bias New Goods Bias Quality improvement bias

Estimates of outlet bias • When discount outlets are increasing in importance, this bias will be: BD = (1 + I) s d, • where s is the increase in the share of discount stores while d is the average percentage discount in such stores.

Evolution of distribution channels

Price Differences across Outlets

Bias due to Hypermarkets

Bias Due to Discount Stores

The main sources of inflationary bias • • Consumer Substitution Bias Outlet Substitution Bias New Goods Bias Quality improvement bias

Quality Improvement • specification of quality is much more difficult than specification of quantities • natural tendency in command economies to reduce quality while still meeting numerical quantity goals • important in transition economies, largely because initial quality levels were so low

Examples of Quality Changes • 1990 Czech "fresh" milk came in a plastic container giving the milk a chemical smell and resulted in the milk spoiling in less than two days. • 1996 fresh milk from the same dairy came in paper-pack that assured a natural smell and durability that was guaranteed for four days

Adjusting Quality Changes w. r. t Price Changes Mostly used methods: • Overlap pricing method • Link method • Hedonic regressions • direct quality adjustment

Quality Adjustment for Cars • Skoda Favorit replaced by Skoda Felicia – the CSO reported a 5. 01 % improvement in utility – “link method” gives 6. 61% (32% higher estimate then CSO) We should use Hedonic regression

Hedonic regression: Czech Cars • Using data from catalogues, magazines. • Engine power and volume, Doors, Cylinders, Tiresdiameters, ABS, wheel base, dimensions (front and back tracks, length, width, height), weight, maximum weight, acceleration time from 0 to 100 km/h, maximum velocity, volume of tank, mileage, etc.

Data: Prices and Attributes • 1993: 684 prices and 834 attribute sets (256 data points). • 1994: 780 prices and 944 attribute sets (552, 424 data points). • 1995: 1013 prices and 998 attribute sets (441 complete, 428 data points).

Results • Using all attributes (fishing for the highest adj. R 2) we obtain quality change equal to 11. 56 %. • Using restricted model (highly correlated attributes are excluded - tests for stability) we obtain the resulting quality change 16. 61 %.

Hedonic regression: Czech Electronics Official Price Change Hedonic Regression Video Cameras VCRs (1996 -1998) -7. 9% (1996 -2000) -15. 6% -19. 3% -37. 6%

Problems in Using Classical Methods • Every-day products such as milk do not have the extensive set of measurable characteristics needed for hedonic regressions • Small continuous changes in products do not lead to observable changes needed for use of the overlap or temporary exclusion methods.

Proposed Approach • The best way to ascertain the extent to which consumers believe the quality of the products they purchase has changed is to ask the consumers themselves. FOCUS GROUP RESEARCH

Focus Group Layout • Performed by AISA during March and May 2001. • Men and women aged 33 to 55 who were the head or spouse of the head of a household in 1990 and who had secondary or higher education. • Six individuals per a group evaluated approximately ten different products.

Session Script • The moderator presents 1990 product. • 2001 product is presented and the group discusses its characteristics. • Participants are asked for a fair and appropriate price for the 1990 version, given 2001 version price. • Individual choices are discussed among the group.

Session Script (Continue) • At the end of the session, after discussing about ten different products, respondents are asked to again individually evaluate and record their relative evaluations for each product. (a chance to revise their evaluations in light of the group discussion and the several products being evaluated)

Problems in valuations: • It is believed that consumers value items more highly when asked what they would require to give them up (willingness to accept) versus what they would be willing to pay for an item they do not now have.

Difference should be minimal • Difference is smaller when close substitutes exist for the product in question (Shogren (1994); Adamowicz, Bhardwaj and Macnab (1993)) • Difference is higher when consumers do not have experience with both products (Kolstad and Guzman 1999)

Results of Focus Group Pricing • Our focus groups evaluated 63 items that comprised 16. 2 per cent of the total weight in the consumer basket as of 1990. • Three items have been excluded from the analysis because “the 1990 quality was so low that the proper 2001 price would be zero or negative”

Results for Food (18 items)

Results for Clothes (6)

Results for Home Appliances

Results Recreational Products

Summary of Results

Real GDP and DEM denominated growth