Dynamics of Consumer Demand for New Durable Goods - Lecture Slides | ECON 664, Study notes of Economics

Download Dynamics of Consumer Demand for New Durable Goods - Lecture Slides | ECON 664 and more Study notes Economics in PDF only on Docsity! “Dynamics of Consumer Demand for New Durable Goods” by Gowrisankaran and Rysman Work in Progress 2007 Diana Iercosan  Aggregate Data:  NPD Techworld, Online  378 camcorder models  11 brands  monthly  March 2000-May 2008)  market share (US Census) Gowrisankaran and Rysman WP 2007 Gowrisankaran and Rysman WP 2007 Figure 2: Average indicator characteristics over time co * seit “eo ae = _ T T T T Jan0O Jano2 Jano4 JanO6 Photo capable ————~— Night shot SES DRESS Lamp corcoooooo Media: tape Gowrisankaran and Rysman WP 2007 Figure 3: Prices and sales for camcorders fF 1000 - 800 i ~ F600 [> 400 600-7 z & = ‘o 40075 8 Cc a s o na S ¢ “, 2007 é o Ww oT T T T T Jando Jando2 Jano4 Jan06 Seasonally-adjusted sales | 9 ~---------- Sales ETT Price Price in Jan. 2000 dollars  Gowrisankaran and Rysman WP 2007 Herfindahl index Figure 4: Competition in the camcorder market 60007 s000- 40007 3000-7 2000 F100 [80 > 60 -40 20 Jando Janod2 Jano4 JanO6 Herfindahl index TTS Number of products Number of products Gowrisankaran and Rysman WP 2007  Demand Side:  Continuum of heterogeneous agents i  Infinitely lived  Endogenous prices  Rational expectations  Future discount rate  Repeated purchase (obtains utility from only one good)  Allows time and product substitution  Supply Side:  New consumer durable good at time t=0  Infinitely durable  Secondary market are insignificant (technological change)  Exogenous choice set: +1  Exogenous development in product characteristics  tJ Gowrisankaran and Rysman WP 2007  Consumer i at time t maximizes:  Net Utility Flow from Good j:  given Gross Utility Flow  Errors are distributed GEV 1 with median  Gowrisankaran and Rysman WP 2007  Or, Consumer i at time t maximizes:  Net Utility Flow from Outside Option:  given a past purchase at time the Gross Flow - or, given NO previous purchase Mean Utility t̂ Gowrisankaran and Rysman WP 2007 a» By Assumption 1 Inclusive Value Sufficiency (IVS) P (6424112) = P (6i,041|Q) if dit (24) = die (NF). = Then EV; (Shoe, 2) = EV; (Sirs die, (Siet1, Jitra,--- a)) =» Becomes EV, (Shor die, B [Sitets Sites «> \) - EV, (Stor, Sit) Gowrisankaran and Rysman WP 2007 Figure 7: Evolution of 6;; over time -15 -10 -20 \ \ -25 Janoo Jan02 Jan04 Jan06 TTT delta_it 20th percentile ae delta_it 80th percentile delta_it 50th percentile Gowrisankaran and Rysman WP 2007  We can rewrite the Bellman Equation: Gowrisankaran and Rysman WP 2007 Figure 6: Average estimation error (€j) by month O01 £005 -.005 -.01 T T T Janoo Jand2 Jano4 Jand6 Mean xsi orm" Zero base Gowrisankaran and Rysman WP 2007 Procedure for the Estimation of Parameters:  Search over parameters  Compute average utility flow over population  Predict market shares based on dynamic consumer optimization Gowrisankaran and Rysman WP 2007 Inner loop  Evaluates computed predicted market shares  starting with draw that gives  and sensible “guestimations” of  to calculate for consumer i ‘s market share at time t=0  aggregate  Repeat each time for new distribution Gowrisankaran and Rysman WP 2007  BLP standard instruments used for price (z variables):  Product characteristics in x  Mean product characteristics for a firm at the same time  Mean product characteristics for all firms at the same time  Count of products offered by the firm and by other firms  NO direct cost shifter  Affect price cost margin Table 2: Parameter estimates Gowrisankaran and Rysman WP 2007 nam . . namic Parameter Base dynamic mod with model without Static model ated to mode with model exira random micro coefficients repurchases year moments Mean coefficients (a) Constant 141 (044)* 0970195)" -087 (1.5) 8.90 (263) 4,08 (132) =.243 (213) Log price 2.66 (S76)* -2.74.,975)" — -.056 (72.6) 0247 (19.1) — 089 (145) 3.01 (582) * Log size “007 (001) = 007 (014) = 002 (Fesy* = —153 (068) —340 (20) = —019(.003)* Log pixel 095 (.050) 098 (028) * 002 (027) 2.56 (2.43) 4.52 (5.85) 241 (146) * Log zoom 007 (002)* OOF LG02)* O07 (Ge4)* 6584 CORK) * S61 (269) O10 (004) * LogLcD size 993 (001)* 003 C001)® = —Se-4 (9-4) =053 (105) =361 (.325) OL (004) * Media: DVD 928 (.005)* 027.000)" —001(.004) —.177 (344) 229 (1.35) O52 (017) * Media: tape 0007 (008) 007 (005) —007,(0003)* 763 (333)* = — 671 (1.05) O17 (O16) Media: HD 020 (006) .023 (008) * 008 (004) 873 (425) * 1.32 (1.359) 039.019) * Lamp 006 (OOLy*® = OOFLGOI*® — —.002 (001) = 209 (.130) -.351 (402) 002 (004) Night shot O09 (001)* OOSCO01I)*® O07 (6e-4)* 646 (OTS * 1.20¢.199) 022 (003) * Photo capable = -014 (002) * = —015 C008)" = =.008 (002) * = =.431 (205) * 432 (767) 022 (.007) * Standard deviation coefficients (Z'7) Constant ORG (025) * 058 (.130) 2e-5 (27) 07 (4e4) 036 (1e3) le-7 (082) Log price Te-6 (563) 043 (8.06) 0002 (817) O01 (267) COLL (67.7) 651 (233) * Log size 5e-09 (096) Log pixel 0015 (337) Standard errors in parentheses, statistical significance at 3% level indicated with *