Archive SIZ Format Update

• 2025-07-27

  Updated BETA in several cases:

  • Fixed bugs on special issues by changing pass to continue so unsatisfied cases are not included.
  • Changed matrix calculations from np.mat to np.array.
  • Extended market beta to FF5 beta in beta_ff5.py.
• 2025-04-22

  Fixed SUE again:

  • Switched sorting and cumulative count order to ensure cumulative counts are computed correctly.
  • Used EPS data from the past eight quarters, subtracting values from four quarters prior, then calculating standard deviation.
  • Required at least six non-missing quarterly observations among the prior eight quarters, including the most recent quarter.
• 2025-04-01

  Some EPS values did not change, but Python did not recognize this and assigned a standard deviation of zero, leading to a huge raw number for sue.

• 2025-02-28

  Fixed the calculation of acc (Operating Accruals). Following Sloan (1996), HXZ (2020), and GHZ (2017), annual income before extraordinary items (ib) minus operating cash flows (oancf) is divided by average total assets (at). If oancf is missing, it is set to change in act minus change in che minus change in lct plus change in dlc plus change in txp minus dp. Missing changes in income taxes payable are set to zero.

• 2025-02-23

  Fixed zerotrade (Number of zero-trading days rolling 3m). Following the footnote in Liu (2006), used a deflator of 11,000 to construct rolling 3-month zerotrade.

• 2025-02-23

  Corrected annual roa by using annual ib and one-year-lagged total assets (at_l1).

• 2025-02-23

  Unified several units to one consistent scale:

  • dolvol = log(vol * prc): multiply vol by 100 because monthly volume is in hundred shares.
  • turn: multiply vol by 100 and shrout by 1000 because monthly volume is in hundred shares and shares outstanding are in thousands.
  • std_turn: multiply shrout by 1000 because daily volume is in shares while shares outstanding are in thousands.
• 2025-02-23

  Corrected the unit of me (size) from thousand to million so it is consistent with accounting variables used in characteristic calculations. Influenced characteristics include bm, me_ia, cfp, ep, rsup, lev, sp, rdm, adm, dy, and alm.

• 2025-02-21

  Fixed a denominator issue in ill (illiquidity) by adding brackets around abs(prc) * vol.

• 2024-08-22

  Observations before the first annual records are already excluded. This is equivalent to setting cumcount >= 0. Setting cumcount >= 1 is the same as the FF3 replication code and excludes the first two years of annual records.

• 2024-08-08

  In step 3 (merge_chars.py), changed the merge method from outer to left so the permno list remains consistent with accounting-related characteristics and does not generate extra observations.

• 2024-07-10

  Added ticker and comnam from CRSP and adjusted fillna_ind, fillna_all, and standardize in functions.py so fundamental stock information does not influence data preparation.

• 2024-06-01

  For daily-return characteristics, added the constraint that values are set to missing if volume is empty for more than 40 trading days out of 60 days.