index.org

A comparison of some structural models of private information arrival

Jefferson Duarte, Edwin Hu, and Lance Young

Paper

Latest version available at SSRN: https://ssrn.com/abstract=2564369

Previous Versions

• JFE #4 2018-12-17
• JFE #3 2017-11-24
• JFE #2 2017-09-07
• JFE #1 2017-03-10
• JF #2 2016-04-29
• JF #1 2015-05-01

Data

https://www.dropbox.com/s/45b42e89gaafg0n/cpie_data.zip?dl=1

The data contains several files:

• cpie_daily.csv
• eo_yearly.csv
• dy_yearly.csv
• gpin_yearly.csv
• owr_yearly.csv

Daily

cpie_daily.csv is sorted by permno and date:

permno	year	date	cpie_pin	cpie_dy	cpie_gpin	cpie_owr	cpie_mech	ret_o	ret_d	y_e	n_buys	n_sells	turn
10057	1993	19930104	0.0050	0.608	0.011	0.85	0	-0.007	-0.014	-0.35	4	7	11
10057	1993	19930105	0.0000	0.003	0.071	0.88	0	0.011	0.012	0.93	2	0	2
10057	1993	19930106	0.0007	0.212	0.014	0.91	0	-0.016	-0.014	-0.38	3	5	8
10057	1993	19930107	0.0257	0.131	0.433	0.82	0	0.001	0.015	0.26	8	4	12
10057	1993	19930108	0.0962	0.946	0.003	0.72	1	0.001	-0.007	-0.57	4	10	14
10057	1993	19930111	0.0011	0.009	0.281	0.69	0	0.007	-0.003	0.41	5	1	6
10057	1993	19930112	0.9448	0.226	0.528	0.97	1	0.016	0.022	0.11	14	12	26
10057	1993	19930113	0.4594	0.973	0.001	0.82	1	0.008	-0.020	-0.62	4	12	16
10057	1993	19930114	0.0006	0.101	0.044	0.88	0	-0.008	-0.015	-0.33	4	4	8

cpie_pin corresponds to $CPIE_{PIN}$ in the paper (the PIN model).

cpie_dy corresponds to $CPIE_{DY}$ in the paper (the DY model).

cpie_gpin corresponds to $CPIE_{GPIN}$ in the paper (the GPIN model).

cpie_owr corresponds to $CPIE_{OWR}$ (the Odders-White and Ready (2008) model).

cpie_mech is the $CPIE_{Mech}$, which is a dummy variable defined as: \begin{equation} CPIE_{Mech,j,t}=% \begin{cases} 0, & \text{ if }turn_{j,t}<\overline{turn}_{j}
1, & \text{ if }turn_{j,t}\geq \overline{turn}_{j},% \end{cases} \end{equation}

ret_d, ret_o, and y_e correspond to $(r_d,r_o,y_e)$.

n_buys and n_sells corresponds to the $B$ and $S$, and turn is the sum, corresponding to $turn$.

PIN Model

pin_yearly.csv is sorted by permno and year:

permno	year	a	eb	es	u	d
10057	1993	0.2301	5.4697	5.7433	10.5068	0.6052
10057	1994	0.0810	6.9449	6.6969	34.4076	0.3984
10057	1995	0.2692	14.2371	16.8493	33.2753	0.8156
10064	1993	0.2502	45.7045	41.3353	71.4708	0.6673
10064	1994	0.2835	25.8929	27.7240	40.0622	0.6076
10064	1995	0.1656	32.7579	38.9675	94.3707	0.8213
10064	1996	0.1910	27.7305	39.9372	94.2733	0.8373
10071	1993	0.2755	15.2707	14.3848	22.8094	0.6077
10071	1994	0.2000	12.8310	14.1135	24.6615	0.6733

a is $α$, eb is $ε_B$, es is $ε_S$, u is $μ$, and d is $δ$.

DY Model

dy_yearly.csv is sorted by permno and year:

permno	year	a	eb	es	ub	us	d	tn	sb	ss
10057	1993	0.38	4.67	3.55	10.63	5.00	0.21	0.20	7.67	7.74
10057	1994	0.30	4.52	4.84	10.81	9.21	0.51	0.07	26.43	32.65
10057	1995	0.43	11.92	9.46	26.59	10.60	0.40	0.24	21.59	25.63
10064	1993	0.36	35.98	37.14	31.10	69.81	0.76	0.15	82.19	30.60
10064	1994	0.49	16.74	24.25	21.83	45.73	0.88	0.24	27.10	21.28
10064	1995	0.37	27.68	31.84	39.64	43.77	0.59	0.09	103.78	39.22
10064	1996	0.21	26.46	38.54	78.89	124.34	0.87	0.02	226.61	623.26
10071	1993	0.36	11.91	11.46	11.65	17.20	0.60	0.20	23.35	14.43
10071	1994	0.44	9.46	11.76	11.65	15.28	0.74	0.10	30.51	21.51

a is $α$, eb is $ε_B$, es is $ε_S$, ub is $μ_B$, us is $μ_S$, d is $δ$, tn is $θ$, sb is $Δ_B$, and ss is $Δ_S$.

GPIN Model

gpin_yearly.csv is sorted by permno and year:

permno	year	a	r	p	eta	d	th
10057	1993	0.30	6.51	0.62	1.00	1.00	0.44
10057	1994	0.17	2.27	0.85	1.00	0.40	0.51
10057	1995	0.28	2.89	0.92	1.00	0.48	0.54
10064	1993	0.19	7.25	0.93	0.75	0.57	0.53
10064	1994	0.16	8.48	0.87	0.78	0.48	0.50
10064	1995	0.18	5.88	0.93	0.77	0.76	0.44
10064	1996	0.11	2.41	0.97	0.90	0.59	0.44
10071	1993	0.21	6.46	0.83	0.79	0.60	0.52
10071	1994	0.13	5.01	0.85	0.88	0.80	0.44

a is $α$, r is $r$, p is $p$, eta is $η$, d is $δ$, and th is $θ$.

OWR Model

owr_yearly.csv is sorted by permno and year:

permno	year	a	su	sz	si	spd	spo
10057	1993	0.7808	0.1886	0.4790	0.0226	0.0059	0.0101
10057	1994	0.4135	0.1548	0.5136	0.0298	0.0074	0.0108
10057	1995	0.7746	0.1368	0.4319	0.0251	0.0077	0.0000
10064	1993	0.2991	0.1227	0.3457	0.0248	0.0085	0.0072
10064	1994	0.6279	0.1516	0.3194	0.0222	0.0070	0.0047
10064	1995	0.7526	0.1865	0.3440	0.0241	0.0073	0.0009
10064	1996	0.6543	0.1391	0.3481	0.0299	0.0064	0.0000
10071	1993	0.7885	0.1390	0.4146	0.0148	0.0045	0.0054
10071	1994	0.6455	0.1417	0.4427	0.0132	0.0061	0.0050

a is $α$, su is $σ_u$, sz is $σ_z$, si is $σ_i$, spd is $σ_{pd}$, and spo is $σ_{po}$.

**New data**

2013–2019 GPIN and OWR estimates:
https://www.dropbox.com/s/6xaa3x5zbmvyyq1/pin-est-1319.zip?dl=1

Based on requests from other researchers we have updated our estimates beyond the sample period in our paper. These estimates may be used as starting points for your own estimation, or used as-is. We have done some basic quality checks with the estimates, but not to the full extent of the 1993–2012 sample from the paper. In the paper we also only used NYSE-listed stocks. If you have any questions, comments, suggestions, or find any issues please feel free to contact me. If you use these estimates in your work, please cite to our paper and website so that future researchers can find our work.

• Based on WRDS DTAQ Intraday Indicators.
• CRSP shrcd 10, 11 and exchcd 1, 2, 3, 4.
• Unlike the paper we do not remove distribution/event days for the OWR.
• Estimates are based on up to five random starting points.

Code

Python Environment: https://www.dropbox.com/scl/fi/m3u1i5aoejf7ltoo30tl6/environment.sh?rlkey=s44j5sbqn5m7ri5hlxhk67xlw&st=fbxwbvxu&dl=1

Code is now available on its own GitHub page. A detailed description of the code is also available here.