- The material here refers to the February, 2005 draft of the paper (Trading Costs and Returns for US Equities: The Evidence from Daily Data), and is being maintained for the convenience of researchers who are currently using the estimates from that paper.
- The current version of the paper is Trading Costs and Returns for US Equities: Estimating Effective Costs from Daily Data (August, 2006). The estimates developed in the newer draft are available here.
- I make the estimates available for noncommercial research purposes.

Gibbs v01.zip (April 7, 2005, about 12 Kbytes) contains Matlab routines to estimate the Roll model using the Gibbs sampler.

Liquidity v01.zip (April 7, 2005, about 43 Megabytes) contains daily and monthly TAQ-based liquidity estimates for the comparison sample used in the paper (250 firms per year, 1993-2003). NOTE: The price impact and related measures (in the est file) are missing for August 1996. Thanks to Joost Driessen for catching this.

Note:

dsfLiquidity2003v01.sas7bdat (74 Megabytes) contains effective cost, liquidity, illiquidity and gamma estimates for the CRSP daily sample (1962-2003).

If you use these data, please drop me an email at jhasbrou@stern.nyu.edu. I'll notify you of any updates. I would, as well, be grateful to know about any problems or other things you notice about the data. I make this dataset available on a "best efforts" basis. I believe, but cannot ensure, that the calculations are correct.

The dataset maps to the CRSP daily file: (almost) all stocks, from 1962 to 2003. For each permno, there is an (approximately) annual series of the liquidity measures described in the paper, including Gibbs and moment estimates of the effective cost, the liquidity ratio, the illiquidity ratio, the Pastor-Stambaugh reversal measure.

The estimates are based on daily data. I start with the first day a firm (permno) is in the daily file. Going forward, a sample break occurs when a year end is reached, or there is a change of listing venue, share code, share class, or if the CRSP cumulative price adjustment factor changes by more than 10%. (Most of the time there will be one record per firm per year).

The dataset contains estimates computed even when the number of observations is very low (and estimation error would be presumed to be large). In all cases I include in the dataset the number of observations used. In my own work, I generally ignore (set to missing) estimates when the number of observations is under 50. Use your own judgement.

Alphabetic List of Variables and Attributes | |||||
---|---|---|---|---|---|

# | Variable | Type | Len | Format | Label |

12 | I1 |
Num | 8 | Amihud illiquidity ratio | |

13 | I2 |
Num | 8 | Square-root variant of Amihud illiquidity ratio | |

10 | L1 |
Num | 8 | Liquidity ratio | |

11 | L2 |
Num | 8 | Square-root variant of liquidity ratio | |

1 | PERMNO |
Num | 4 | CRSP permno | |

34 | PRC |
Num | 4 | CRSP, most recent valid closing price as of sampleEndDate | |

30 | cLevelMean |
Num | 8 | Gibbs est of C, $/share | |

31 | cLogMean |
Num | 8 | Gibbs est of c, log | |

4 | cMdmLevel |
Num | 8 | Moment estimate of C, $/share, infeasible set to 'missing' | |

5 | cMdmLevelz |
Num | 8 | Moment estimate of C, $/share, infeasible set to zero | |

8 | cMdmLog |
Num | 8 | Moment estimate of c, (log eff cost), infeasible set to 'missing' | |

9 | cMdmLogz |
Num | 8 | Moment estimate of c, (log eff cost), infeasible set to zero | |

24 | comnam |
Char | 32 | CRSP comnam as of sampleEndDate | |

26 | dFacpr |
Num | 8 | CRSP cfacpr on sampleStartDate / most recent prior cfacpr | |

21 | exchcd |
Num | 6 | CRSP exchcd, as of sampleEndDate | |

16 | gamma |
Num | 8 | Pastor and Stambaugh gamma | |

35 | logMktCap |
Num | 8 | log equity market capitalization ($ Million) as of sampleEndDate | |

17 | nGamma |
Num | 8 | # observations used to compute gamma | |

15 | nI |
Num | 8 | Illiquidity ratios, # days used to compute averages | |

14 | nL |
Num | 8 | Liquidity ratios, # days used to compute averages | |

2 | nMLevel |
Num | 8 | # obs used to estimate cMdmLevel and stdLevel | |

6 | nMLog |
Num | 8 | # obs used to estimate cMdmLog and stdLog | |

29 | nMidGibbs |
Num | 4 | # days for which the reported price was a quote midpoint | |

28 | nMissingGibbs |
Num | 4 | # days for which price is missing | |

27 | nObsGibbs |
Num | 4 | # days used in Gibbs estimate, incl missing and midpoint | |

19 | sampleEndDate |
Num | 8 | DATE9. | Last date used to estimate liquidity measures in this record |

18 | sampleStartDate |
Num | 8 | DATE9. | First date used to estimate liquidity measures in this record |

32 | sduLevelMean |
Num | 8 | Gibbs est, std dev of random walk component, $/share | |

33 | sduLogMean |
Num | 8 | Gibbs est, std dev of log random walk component | |

22 | shrcd |
Num | 6 | CRSP shrcd as of sampleEndDate | |

23 | shrcls |
Char | 1 | CRSP shrcls as of sampleEndDate | |

25 | shrout |
Num | 6 | CRSP shrout as of sampleEndDate | |

3 | stdLevel |
Num | 8 | D12.3 | Std deviation of daily price changes, $/share |

7 | stdLog |
Num | 8 | D12.3 | Std deviation of daily log price changes |

20 | ticker |
Char | 5 | CRSP ticker as of sampleEndDate |

Other properties of the dataset:

Data Set Name | SASDS.DSFLIQUIDITY2003V01 | Observations | 285396 |
---|---|---|---|

Member Type | DATA | Variables | 35 |

Engine | V9 | Indexes | 0 |

Created | Thu, Feb 03, 2005 04:24:02 PM | Observation Length | 272 |

Last Modified | Thu, Feb 03, 2005 04:24:02 PM | Deleted Observations | 0 |

Protection | Compressed | NO | |

Data Set Type | Sorted | NO | |

Label | |||

Data Representation | WINDOWS_32 | ||

Encoding | wlatin1 Western (Windows) |