************US SIC-4 INDUSTRY BY YEAR UNCERTAINTY MEASURES***************** * This file runs some basic regression of output, sales, employment and investment on industry-by-year uncertainty measures. * It is designed simply to show users what this data is and provide some basic descriptive correlations. * The uncert_tfp and uncert_sales uncertainty data comes from the Census via "Really Uncertain Business Cycles", 2013, by Nick Bloom, Max Floetotto, Nir Jaimovich, Itay Saporta-Eksten & Stephen Terry) * The uncert_ret, uncert_tret and uncert_iret comes from CRSP, again via the "Really Uncetain Business Cycles" paper. These measure: * uncert_ret : the standard deviation of monthly firm stock returns within an industry year * uncert_tret: the average within-year standard deviation of monthly firm stock returns within each industry (the time component of uncert_ret) * uncert_iret: the standard-deviation of the average yearly return for each firm within an industry year (the cross-sectional component of uncert_ret) * Notes: * 1) This uncertainty data is slightly different from the paper version for disclosure reasons (some of our Census samples were too small so we had to smooth). * 2) BECAUSE OF THE CHANGE TO NAICS IN 1997 THE 1996 DATA IS TOO HIGH (THE SWITHC TO NAICS INTRODUCED AN UPWARD BIAS OF ABOUT 5% * 3) The output data comes from the Becker, Gray Marvakov NBER industry data http://www.nber.org/data/nberces5809.html * The data was produced by Itay Saporta-Eksten and Nick Bloom October 2013 * Any questions please contact: isaporta@stanford.edu or nbloom@stanford.edu ******************************************************************************************************************************************************************************************************** ***Below we just run some basic checks of the data to confirm at the industry level output and uncertainty are generally negatively correlated ******************************************************************************************************************************************************************************************************** set more off *************************** *** Industry (SIC4) census uncertainty measures have a strong predictive power for hiring, output, investment and other industry by year factors *************************** * Creating some basic measures cap log using RUBC,replace t u RUBC,replace tsset sic year gen rvship=vship/piship gen rvadd=vadd/piship gen drvship=2*(f1.rvship-rvship)/(f1.rvship+rvship) gen drvadd=2*(f1.rvadd-rvadd)/(f1.rvadd+rvadd) gen demp=2*(f1.emp-emp)/(f1.emp+emp) gen i_k=f1.invest/(equip+plant) gen linvest=log(invest) qui tab year, gen(yy) *Could easily use all these uncertainty measures in logs (rather than levels) and also clean on outliers (e.g. windorize at top and bottom 0.1%) *We do not do this here, but here are these variables for experimentation (results look generally similar) foreach var in uncert_tfp uncert_sales uncert_tret uncert_iret uncert_ret { gen l`var'=log(`var') egen pbottom`var'=pctile(`var'),p(0.1) egen ptop`var'=pctile(`var'),p(99.9) gen win_`var'=`var'*(`var'>pbottom`var')*(`var'=ptop`var'&`var'~=.) } * Showing our SD of TFP innovations uncertainty measure is negatively correlated wtih employement (in a variety of specifications, simple to general) reg demp uncert_tfp reg demp uncert_tfp ,cluster(sic) areg demp uncert_tfp,ab(sic) cluster(sic) areg demp uncert_tfp yy*,ab(sic) cluster(sic) areg demp uncert_tfp yy* [aw=indobs],ab(sic) cluster(sic) * Showing our SD of TFP innovations uncertainty measure is negatively correlated with other measures like value-added and investment areg drvship uncert_tfp yy* [aw=indobs],a(sic) cluster(sic) areg drvadd uncert_tfp yy* [aw=indobs],a(sic) cluster(sic) areg linvest uncert_tfp yy* [aw=indobs],a(sic) cluster(sic) areg i_k uncert_tfp yy* [aw=indobs],a(sic) cluster(sic) * Showing our SD of sales growth uncertainty measure is negatively correlated wtih employement (in a variety of specifications, simple to general) reg demp uncert_sales reg demp uncert_sales,cluster(sic) areg demp uncert_sales,ab(sic) cluster(sic) areg demp uncert_sales yy*,ab(sic) cluster(sic) areg demp uncert_sales yy* [aw=indobs],ab(sic) cluster(sic) * Showing our SD of sales growth uncertainty measure is negatively correlated with other measures like value-added and investment areg drvship uncert_sales yy* [aw=indobs],a(sic) cluster(sic) areg drvadd uncert_sales yy* [aw=indobs],a(sic) cluster(sic) areg linvest uncert_sales yy* [aw=indobs],a(sic) cluster(sic) areg i_k uncert_sales yy* [aw=indobs],a(sic) cluster(sic) ***************** *Showing our CRSP stock-returns uncertainty measures are correlated with the Census uncertainty measures ***************** pwcorr uncert*,sig areg uncert_tfp uncert_ret yy* [aw=indobs],a(sic) cluster(sic) areg uncert_tfp uncert_tret yy* [aw=indobs],a(sic) cluster(sic) areg uncert_tfp uncert_iret yy* [aw=indobs],a(sic) cluster(sic) areg uncert_sales uncert_ret yy* [aw=indobs],a(sic) cluster(sic) areg uncert_sales uncert_tret yy* [aw=indobs],a(sic) cluster(sic) areg uncert_sales uncert_iret yy* [aw=indobs],a(sic) cluster(sic) ***************** *Showing the CRSP measure have a good fit for investment but not for the other variables ***************** * CRSP-total - other measures areg demp uncert_ret yy* ,a(sic) cluster(sic) areg drvship uncert_ret yy* ,a(sic) cluster(sic) areg drvadd uncert_ret yy* ,a(sic) cluster(sic) areg linvest uncert_ret yy* ,a(sic) cluster(sic) areg i_k uncert_ret yy* ,a(sic) cluster(sic) * CRSP-ind - other measures areg demp uncert_iret yy* ,a(sic) cluster(sic) areg drvship uncert_iret yy* ,a(sic) cluster(sic) areg drvadd uncert_iret yy* ,a(sic) cluster(sic) areg linvest uncert_iret yy* ,a(sic) cluster(sic) areg i_k uncert_iret yy* ,a(sic) cluster(sic) * CRSP-time - other measures areg demp uncert_tret yy* ,a(sic) cluster(sic) areg drvship uncert_tret yy* ,a(sic) cluster(sic) areg drvadd uncert_tret yy* ,a(sic) cluster(sic) areg linvest uncert_tret yy* ,a(sic) cluster(sic) areg i_k uncert_tret yy* ,a(sic) cluster(sic) cap log close