mergeron 2024.739099.2__py3-none-any.whl → 2024.739105.2__py3-none-any.whl

mergeron/gen/enforcement_stats.py

@@ -21,7 +21,6 @@ from scipy.stats import beta, norm  # type: ignore
 from .. import (  # noqa: TID252
     _PKG_NAME,
     DATA_DIR,
-    TI,
     VERSION,
     ArrayBIGINT,
     ArrayDouble,
@@ -34,7 +33,7 @@ __version__ = VERSION
 
 
 @enum.unique
-class INDGRPConstants(enum.StrEnum):
+class IndustryGroup(enum.StrEnum):
     ALL = "All Markets"
     GRO = "Grocery Markets"
     OIL = "Oil Markets"
@@ -48,7 +47,7 @@ class INDGRPConstants(enum.StrEnum):
 
 
 @enum.unique
-class EVIDENConstants(enum.StrEnum):
+class OtherEvidence(enum.StrEnum):
     HD = "Hot Documents Identified"
     CC = "Strong Customer Complaints"
     NE = "No Entry Evidence"
@@ -165,32 +164,6 @@ hhi_delta_ranger, hhi_zone_post_ranger = (
     for _f in (HHI_DELTA_KNOTS, HHI_POST_ZONE_KNOTS)
 )
 
-HMG_PRESUMPTION_ZONE_DICT = {
-    HHI_POST_ZONE_KNOTS[0]: {
-        HHI_DELTA_KNOTS[0]: (0, 0, 0),
-        HHI_DELTA_KNOTS[1]: (0, 0, 0),
-        HHI_DELTA_KNOTS[2]: (0, 0, 0),
-    },
-    HHI_POST_ZONE_KNOTS[1]: {
-        HHI_DELTA_KNOTS[0]: (0, 1, 1),
-        HHI_DELTA_KNOTS[1]: (1, 1, 2),
-        HHI_DELTA_KNOTS[2]: (1, 1, 2),
-    },
-    HHI_POST_ZONE_KNOTS[2]: {
-        HHI_DELTA_KNOTS[0]: (0, 2, 1),
-        HHI_DELTA_KNOTS[1]: (1, 2, 3),
-        HHI_DELTA_KNOTS[2]: (2, 2, 4),
-    },
-}
-
-ZONE_VALS = np.unique(
-    np.vstack([
-        tuple(HMG_PRESUMPTION_ZONE_DICT[_k].values())
-        for _k in HMG_PRESUMPTION_ZONE_DICT
-    ]),
-    axis=0,
-)
-
 ZONE_STRINGS = {
     0: R"Green Zone (Safeharbor)",
     1: R"Yellow Zone",
@@ -234,12 +207,37 @@ ZONE_DETAIL_STRINGS_DELTA_LATEX = {
     4: Rf"\Delta HHI \geqslant \text{{{HHI_DELTA_KNOTS[2]} pts.}}",
 }
 
+HMG_PRESUMPTION_ZONE_MAP = {
+    HHI_POST_ZONE_KNOTS[0]: {
+        HHI_DELTA_KNOTS[0]: (0, 0, 0),
+        HHI_DELTA_KNOTS[1]: (0, 0, 0),
+        HHI_DELTA_KNOTS[2]: (0, 0, 0),
+    },
+    HHI_POST_ZONE_KNOTS[1]: {
+        HHI_DELTA_KNOTS[0]: (0, 1, 1),
+        HHI_DELTA_KNOTS[1]: (1, 1, 2),
+        HHI_DELTA_KNOTS[2]: (1, 1, 2),
+    },
+    HHI_POST_ZONE_KNOTS[2]: {
+        HHI_DELTA_KNOTS[0]: (0, 2, 1),
+        HHI_DELTA_KNOTS[1]: (1, 2, 3),
+        HHI_DELTA_KNOTS[2]: (2, 2, 4),
+    },
+}
+
+ZONE_VALS = np.unique(
+    np.vstack([
+        tuple(HMG_PRESUMPTION_ZONE_MAP[_k].values()) for _k in HMG_PRESUMPTION_ZONE_MAP
+    ]),
+    axis=0,
+)
+
 
 def enf_stats_output(
     _data_array_dict: fid.INVData,
     _data_period: str = "1996-2003",
-    _table_ind_group: INDGRPConstants = INDGRPConstants.ALL,
-    _table_evid_cond: EVIDENConstants = EVIDENConstants.UR,
+    _table_ind_group: IndustryGroup = IndustryGroup.ALL,
+    _table_evid_cond: OtherEvidence = OtherEvidence.UR,
     _stats_group: StatsGrpSelector = StatsGrpSelector.FC,
     _enf_spec: INVResolution = INVResolution.CLRN,
     /,
@@ -295,8 +293,8 @@ def enf_stats_output(
 def enf_stats_listing_by_group(
     _invdata_array_dict: Mapping[str, Mapping[str, Mapping[str, fid.INVTableData]]],
     _study_period: str,
-    _table_ind_grp: INDGRPConstants,
-    _table_evid_cond: EVIDENConstants,
+    _table_ind_grp: IndustryGroup,
+    _table_evid_cond: OtherEvidence,
     _stats_group: StatsGrpSelector,
     _enf_spec: INVResolution,
     /,
@@ -331,8 +329,8 @@ def enf_stats_listing_by_group(
 def enf_cnts_listing_byfirmcount(
     _data_array_dict: Mapping[str, Mapping[str, Mapping[str, fid.INVTableData]]],
     _data_period: str = "1996-2003",
-    _table_ind_group: INDGRPConstants = INDGRPConstants.ALL,
-    _table_evid_cond: EVIDENConstants = EVIDENConstants.UR,
+    _table_ind_group: IndustryGroup = IndustryGroup.ALL,
+    _table_evid_cond: OtherEvidence = OtherEvidence.UR,
     _enf_spec: INVResolution = INVResolution.CLRN,
     /,
 ) -> ArrayBIGINT:
@@ -367,8 +365,8 @@ def enf_cnts_listing_byfirmcount(
 def enf_cnts_listing_byhhianddelta(
     _data_array_dict: Mapping[str, Mapping[str, Mapping[str, fid.INVTableData]]],
     _data_period: str = "1996-2003",
-    _table_ind_group: INDGRPConstants = INDGRPConstants.ALL,
-    _table_evid_cond: EVIDENConstants = EVIDENConstants.UR,
+    _table_ind_group: IndustryGroup = IndustryGroup.ALL,
+    _table_evid_cond: OtherEvidence = OtherEvidence.UR,
     _enf_spec: INVResolution = INVResolution.CLRN,
     /,
 ) -> ArrayBIGINT:
@@ -402,8 +400,8 @@ def enf_cnts_listing_byhhianddelta(
 
 def table_no_lku(
     _data_array_dict_sub: Mapping[str, fid.INVTableData],
-    _table_ind_group: INDGRPConstants = INDGRPConstants.ALL,
-    _table_evid_cond: EVIDENConstants = EVIDENConstants.UR,
+    _table_ind_group: IndustryGroup = IndustryGroup.ALL,
+    _table_evid_cond: OtherEvidence = OtherEvidence.UR,
     /,
 ) -> str:
     if _table_ind_group not in (
@@ -475,7 +473,7 @@ def enf_cnts_byconczone(_cnts_array: ArrayBIGINT, /) -> ArrayBIGINT:
                 else (_hhi_delta_ranged == _hhi_zone_delta_lim)
             )
 
-            _zone_val = HMG_PRESUMPTION_ZONE_DICT[_hhi_zone_post_lim][
+            _zone_val = HMG_PRESUMPTION_ZONE_MAP[_hhi_zone_post_lim][
                 _hhi_zone_delta_lim
             ]
 
@@ -723,8 +721,8 @@ def stats_print_rows(
 
 
 def propn_ci(
-    _npos: ArrayINT[TI] | int = 4,
-    _nobs: ArrayINT[TI] | int = 10,
+    _npos: ArrayINT | int = 4,
+    _nobs: ArrayINT | int = 10,
     /,
     *,
     alpha: float = 0.05,

mergeron/gen/upp_tests.py

@@ -11,19 +11,16 @@ from typing import Literal, TypeAlias, TypedDict
 
 import numpy as np
 import tables as ptb  # type: ignore
-from joblib import Parallel, cpu_count, delayed  # type: ignore
 from numpy.random import SeedSequence
 
-from .. import (  # noqa: TID252
-    TF,
-    TI,
+from .. import (  # noqa
     VERSION,
     ArrayBIGINT,
     ArrayBoolean,
     ArrayDouble,
     ArrayFloat,
     ArrayINT,
-    RECConstants,
+    RECForm,
     UPPAggrSelector,
 )
 from ..core import guidelines_boundaries as gbl  # noqa: TID252
@@ -32,12 +29,10 @@ from . import (
     DataclassInstance,
     INVResolution,
     MarketDataSample,
-    MarketSpec,
     UPPTestRegime,
     UPPTestsCounts,
     UPPTestsRaw,
 )
-from . import data_generation as dgl
 from . import enforcement_stats as esl
 
 __version__ = VERSION
@@ -58,169 +53,37 @@ class INVRESCntsArgs(TypedDict, total=False):
     saved_array_name_suffix: str
 
 
-def sim_enf_cnts_ll(
-    _mkt_sample_spec: MarketSpec,
-    _enf_parm_vec: gbl.HMGThresholds,
-    _sim_test_regime: UPPTestRegime,
+def enf_cnts(
+    _market_data_sample: MarketDataSample,
+    _upp_test_parms: gbl.HMGThresholds,
+    _upp_test_regime: UPPTestRegime,
     /,
-    *,
-    sample_size: int = 10**6,
-    seed_seq_list: list[SeedSequence] | None = None,
-    nthreads: int = 16,
-    save_data_to_file: SaveData = False,
-    saved_array_name_suffix: str = "",
 ) -> UPPTestsCounts:
-    """A function to parallelize data-generation and testing
-
-    The parameters `_sim_enf_cnts_kwargs` are passed unaltered to
-    the parent function, `sim_enf_cnts()`, except that, if provided,
-    `seed_seq_list` is used to spawn a seed sequence for each thread,
-    to assure independent samples in each thread, and `nthreads` defines
-    the number of parallel processes used. The number of draws in
-    each thread may be tuned, by trial and error, to the amount of
-    memory (RAM) available.
+    """Estimate enforcement and clearance counts from market data sample
 
     Parameters
     ----------
+    _market_data_sample
+        Market data sample
 
-    _enf_parm_vec
-        Guidelines thresholds to test against
-
-    _mkt_sample_spec
-        Configuration to use for generating sample data to test
-
-    _sim_test_regime
-        Configuration to use for testing
-
-    saved_array_name_suffix
-        Suffix to add to the array names in the HDF5 file
-
-    save_data_to_file
-        Whether to save data to an HDF5 file, and where to save it
-
-    sample_size
-        Number of draws to simulate
-
-    seed_seq_list
-        List of seed sequences, to assure independent samples in each thread
+    _upp_test_parms
+        Threshold values for various Guidelines criteria
 
-    nthreads
-        Number of parallel processes to use
+    _upp_test_regime
+        Specifies whether to analyze enforcement, clearance, or both
+        and the GUPPI and diversion ratio aggregators employed, with
+        default being to analyze enforcement based on the maximum
+        merging-firm GUPPI and maximum diversion ratio between the
+        merging firms
 
     Returns
     -------
-        Arrays of UPPTestCounts
+    UPPTestsCounts
+        Enforced and cleared counts
 
     """
-    _sample_sz = sample_size
-    _subsample_sz = 10**6
-    _iter_count = int(_sample_sz / _subsample_sz) if _subsample_sz < _sample_sz else 1
-    _thread_count = cpu_count()
-
-    if (
-        _mkt_sample_spec.share_spec.recapture_form != RECConstants.OUTIN
-        and _mkt_sample_spec.share_spec.recapture_rate != _enf_parm_vec.rec
-    ):
-        raise ValueError(
-            "{} {} {}".format(
-                f"Recapture rate from market sample spec, {_mkt_sample_spec.share_spec.recapture_rate}",
-                f"must match the value, {_enf_parm_vec.rec}",
-                "the guidelines thresholds vector.",
-            )
-        )
-
-    _rng_seed_seq_list = [None] * _iter_count
-    if seed_seq_list:
-        _rng_seed_seq_list = list(
-            zip(*[g.spawn(_iter_count) for g in seed_seq_list], strict=True)  # type: ignore
-        )
-
-    _sim_enf_cnts_kwargs: INVRESCntsArgs = INVRESCntsArgs({
-        "sample_size": _subsample_sz,
-        "save_data_to_file": save_data_to_file,
-        "nthreads": nthreads,
-    })
-
-    _res_list = Parallel(n_jobs=_thread_count, prefer="threads")(
-        delayed(sim_enf_cnts)(
-            _mkt_sample_spec,
-            _enf_parm_vec,
-            _sim_test_regime,
-            **_sim_enf_cnts_kwargs,
-            saved_array_name_suffix=f"{saved_array_name_suffix}_{_iter_id:0{2 + int(np.ceil(np.log10(_iter_count)))}d}",
-            seed_seq_list=_rng_seed_seq_list_ch,
-        )
-        for _iter_id, _rng_seed_seq_list_ch in enumerate(_rng_seed_seq_list)
-    )
-
-    _res_list_stacks = UPPTestsCounts(*[
-        np.stack([getattr(_j, _k) for _j in _res_list])
-        for _k in ("by_firm_count", "by_delta", "by_conczone")
-    ])
-    upp_test_results = UPPTestsCounts(*[
-        np.column_stack((
-            (_gv := getattr(_res_list_stacks, _g))[0, :, :_h],
-            np.einsum("ijk->jk", np.int64(1) * _gv[:, :, _h:]),
-        ))
-        for _g, _h in zip(
-            _res_list_stacks.__dataclass_fields__.keys(), [1, 1, 3], strict=True
-        )
-    ])
-    del _res_list, _res_list_stacks
-
-    return upp_test_results
-
-
-def sim_enf_cnts(
-    _mkt_sample_spec: MarketSpec,
-    _upp_test_parms: gbl.HMGThresholds,
-    _sim_test_regime: UPPTestRegime,
-    /,
-    *,
-    sample_size: int = 10**6,
-    seed_seq_list: list[SeedSequence] | None = None,
-    nthreads: int = 16,
-    save_data_to_file: SaveData = False,
-    saved_array_name_suffix: str = "",
-) -> UPPTestsCounts:
-    # Generate market data
-    _market_data_sample = dgl.gen_market_sample(
-        _mkt_sample_spec,
-        sample_size=sample_size,
-        seed_seq_list=seed_seq_list,
-        nthreads=nthreads,
-    )
-
-    _invalid_array_names = (
-        ("fcounts", "choice_prob_outgd", "nth_firm_share", "hhi_post")
-        if _mkt_sample_spec.share_spec.dist_type == "Uniform"
-        else ()
-    )
-
-    save_data_to_hdf5(
-        _market_data_sample,
-        saved_array_name_suffix=saved_array_name_suffix,
-        excluded_attrs=_invalid_array_names,
-        save_data_to_file=save_data_to_file,
-    )
-
-    _upp_test_arrays = enf_cnts(_market_data_sample, _upp_test_parms, _sim_test_regime)
-
-    save_data_to_hdf5(
-        _upp_test_arrays,
-        saved_array_name_suffix=saved_array_name_suffix,
-        save_data_to_file=save_data_to_file,
-    )
-
-    return _upp_test_arrays
-
 
-def enf_cnts(
-    _market_data_sample: MarketDataSample,
-    _upp_test_parms: gbl.HMGThresholds,
-    _upp_test_regime: UPPTestRegime,
-    /,
-) -> UPPTestsCounts:
+    _enf_cnts_sim_array = -1 * np.ones((6, 2), np.int64)
     _upp_test_arrays = gen_upp_test_arrays(
         _market_data_sample, _upp_test_parms, _upp_test_regime
     )
@@ -231,23 +94,23 @@ def enf_cnts(
 
     _stats_rowlen = 6
     # Clearance/enforcement counts --- by firm count
-    _firm_counts_weights = np.unique(_fcounts)
-    if _firm_counts_weights is not None and np.all(_firm_counts_weights >= 0):
-        _max_firm_count = len(_firm_counts_weights)
+    _firmcounts_list = np.unique(_fcounts)
+    if _firmcounts_list is not None and np.all(_firmcounts_list >= 0):
+        _max_firmcount = max(_firmcounts_list)
 
         _enf_cnts_sim_byfirmcount_array = -1 * np.ones(_stats_rowlen, np.int64)
-        for _firm_cnt in 2 + np.arange(_max_firm_count):
-            _firm_count_test = _fcounts == _firm_cnt
+        for _firmcount in np.arange(2, _max_firmcount + 1):
+            _firmcount_test = _fcounts == _firmcount
 
             _enf_cnts_sim_byfirmcount_array = np.vstack((
                 _enf_cnts_sim_byfirmcount_array,
                 np.array([
-                    _firm_cnt,
-                    np.einsum("ij->", 1 * _firm_count_test),
+                    _firmcount,
+                    np.einsum("ij->", 1 * _firmcount_test),
                     *[
                         np.einsum(
                             "ij->",
-                            1 * (_firm_count_test & getattr(_upp_test_arrays, _f)),
+                            1 * (_firmcount_test & getattr(_upp_test_arrays, _f)),
                         )
                         for _f in _upp_test_arrays.__dataclass_fields__
                     ],
@@ -353,12 +216,10 @@ def gen_upp_test_arrays(
         getattr(_upp_test_parms, _f) for _f in ("guppi", "divr", "cmcr", "ipr")
     )
 
-    _enf_resolution, _guppi_aggregator, _divr_aggregator = (
-        getattr(_sim_test_regime, _f)
-        for _f in ("resolution", "guppi_aggregator", "divr_aggregator")
+    _guppi_array, _ipr_array, _cmcr_array = (
+        np.empty_like(_market_data.price_array) for _ in range(3)
     )
 
-    _guppi_array = np.empty_like(_market_data.divr_array)
     np.einsum(
         "ij,ij,ij->ij",
         _market_data.divr_array,
@@ -367,18 +228,53 @@ def gen_upp_test_arrays(
         out=_guppi_array,
     )
 
-    _ipr_array = np.empty_like(_market_data.divr_array)
-    np.divide(_guppi_array, (1 - _market_data.divr_array[:, ::-1]), out=_ipr_array)
+    np.divide(
+        np.einsum("ij,ij->ij", _market_data.pcm_array, _market_data.divr_array),
+        1 - _market_data.divr_array,
+        out=_ipr_array,
+    )
 
-    _cmcr_array = np.empty_like(_market_data.divr_array)
     np.divide(_ipr_array, 1 - _market_data.pcm_array, out=_cmcr_array)
 
-    _test_measure_seq = (_market_data.divr_array, _guppi_array, _cmcr_array, _ipr_array)
+    (_divr_test_vector,) = _compute_test_value_seq(
+        (_market_data.divr_array,),
+        _market_data.frmshr_array,
+        _sim_test_regime.divr_aggregator,
+    )
+
+    (_guppi_test_vector, _cmcr_test_vector, _ipr_test_vector) = _compute_test_value_seq(
+        (_guppi_array, _cmcr_array, _ipr_array),
+        _market_data.frmshr_array,
+        _sim_test_regime.guppi_aggregator,
+    )
+    del _cmcr_array, _ipr_array, _guppi_array
 
+    if _sim_test_regime.resolution == INVResolution.ENFT:
+        _upp_test_arrays = UPPTestsRaw(
+            _guppi_test_vector >= _g_bar,
+            (_guppi_test_vector >= _g_bar) | (_divr_test_vector >= _divr_bar),
+            _cmcr_test_vector >= _cmcr_bar,
+            _ipr_test_vector >= _ipr_bar,
+        )
+    else:
+        _upp_test_arrays = UPPTestsRaw(
+            _guppi_test_vector < _g_bar,
+            (_guppi_test_vector < _g_bar) & (_divr_test_vector < _divr_bar),
+            _cmcr_test_vector < _cmcr_bar,
+            _ipr_test_vector < _ipr_bar,
+        )
+
+    return _upp_test_arrays
+
+
+def _compute_test_value_seq(
+    _test_measure_seq: tuple[ArrayDouble, ...],
+    _wt_array: ArrayDouble,
+    _aggregator: UPPAggrSelector,
+) -> tuple[ArrayDouble, ...]:
     _wt_array = (
-        _market_data.frmshr_array
-        / np.einsum("ij->i", _market_data.frmshr_array)[:, None]
-        if _guppi_aggregator
+        _wt_array / np.einsum("ij->i", _wt_array)[:, None]
+        if _aggregator
         in (
             UPPAggrSelector.CPA,
             UPPAggrSelector.CPD,
@@ -388,7 +284,7 @@ def gen_upp_test_arrays(
         else EMPTY_ARRAY_DEFAULT
     )
 
-    match _guppi_aggregator:
+    match _aggregator:
         case UPPAggrSelector.AVG:
             _test_value_seq = (
                 1 / 2 * np.einsum("ij->i", _g)[:, None] for _g in _test_measure_seq
@@ -428,30 +324,7 @@ def gen_upp_test_arrays(
             )
         case _:
             raise ValueError("GUPPI/diversion ratio aggregation method is invalid.")
-    del _cmcr_array, _guppi_array
-    (_divr_test_vector, _guppi_test_vector, _cmcr_test_vector, _ipr_test_vector) = (
-        _test_value_seq
-    )
-
-    if _divr_aggregator == UPPAggrSelector.MAX:
-        _divr_test_vector = _market_data.divr_array.max(axis=1, keepdims=True)
-
-    if _enf_resolution == INVResolution.ENFT:
-        _upp_test_arrays = UPPTestsRaw(
-            _guppi_test_vector >= _g_bar,
-            (_guppi_test_vector >= _g_bar) | (_divr_test_vector >= _divr_bar),
-            _cmcr_test_vector >= _cmcr_bar,
-            _ipr_test_vector >= _ipr_bar,
-        )
-    else:
-        _upp_test_arrays = UPPTestsRaw(
-            _guppi_test_vector < _g_bar,
-            (_guppi_test_vector < _g_bar) & (_divr_test_vector < _divr_bar),
-            _cmcr_test_vector < _cmcr_bar,
-            _ipr_test_vector < _ipr_bar,
-        )
-
-    return _upp_test_arrays
+    return tuple(_test_value_seq)
 
 
 def initialize_hd5(
@@ -494,11 +367,7 @@ def save_data_to_hdf5(
 
 
 def save_array_to_hdf5(
-    _array_obj: ArrayFloat[TF]
-    | ArrayINT[TI]
-    | ArrayDouble
-    | ArrayBIGINT
-    | ArrayBoolean,
+    _array_obj: ArrayFloat | ArrayINT | ArrayDouble | ArrayBIGINT | ArrayBoolean,
     _array_name: str,
     _h5_group: ptb.Group,
     _h5_file: ptb.File,

mergeron-2024.739105.2.dist-info/METADATA

@@ -0,0 +1,115 @@
+Metadata-Version: 2.1
+Name: mergeron
+Version: 2024.739105.2
+Summary: Merger Policy Analysis using Python
+License: MIT
+Keywords: merger policy analysis,merger guidelines,merger screening,policy presumptions,concentration standards,upward pricing pressure,GUPPI
+Author: Murthy Kambhampaty
+Author-email: smk@capeconomics.com
+Requires-Python: >=3.12,<4.0
+Classifier: Development Status :: 4 - Beta
+Classifier: Environment :: Console
+Classifier: Intended Audience :: End Users/Desktop
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3 :: Only
+Classifier: Programming Language :: Python :: Implementation :: CPython
+Requires-Dist: aenum (>=3.1.15,<4.0.0)
+Requires-Dist: attrs (>=23.2)
+Requires-Dist: bs4 (>=0.0.1)
+Requires-Dist: certifi (>=2023.11.17)
+Requires-Dist: google-re2 (>=1.1)
+Requires-Dist: jinja2 (>=3.1)
+Requires-Dist: joblib (>=1.3)
+Requires-Dist: matplotlib (>=3.8)
+Requires-Dist: mpmath (>=1.3)
+Requires-Dist: msgpack (>=1.0)
+Requires-Dist: msgpack-numpy (>=0.4)
+Requires-Dist: numpy (>=1.26,<2)
+Requires-Dist: scipy (>=1.12)
+Requires-Dist: sympy (>=1.12)
+Requires-Dist: tables (>=3.8)
+Requires-Dist: types-beautifulsoup4 (>=4.11.2)
+Requires-Dist: urllib3 (>=2.2.2,<3.0.0)
+Requires-Dist: xlrd (>=2.0.1,<3.0.0)
+Requires-Dist: xlsxwriter (>=3.1)
+Description-Content-Type: text/x-rst
+
+mergeron: Merger Policy Analysis using Python
+=============================================
+
+Analyze the sets of mergers conforming to concentration and diversion ratio bounds. Analyze intrinsic enforcement rates, and intrinsic clearance rates, under concentration, diversion ratio, GUPPI, CMCR, and IPR bounds using generated data with specified distributions of market shares, price-cost margins, firm counts, and prices, optionally imposing restrictions implied by statutory filing thresholds and/or Bertrand-Nash oligopoly with MNL demand. Download and analyze merger investigations data published by the U.S. Federal Trade Commission in various reports on extended merger investigations (Second Requests) during 1996 to 2011.
+
+Here, enforcement rates derived with merger enforcement as being exogenous to firm conduct are defined as intrinsic enforcement rates, and similarly intrinsic clearance rates. Depending on the merger enforcement regime, or merger control regime, intrinsic enforcement rates may also not be the complement of intrinsic clearance rates, i.e, it is not necessarily true that the intrinsic clearance rate estimate for a given enforcement regime is 1 minus the intrinsic enforcement rate. In contrast, observed enforcement rates reflect the deterrent effects of merger enforcement on firm conduct as well as the effects of merger screening on the level of enforcement; and, by definition, the observed clearance rate is 1 minus the observed enforcement rate.
+
+Introduction
+------------
+
+Module :code:`.core.guidelines_boundaries` includes classes for specifying concentration bounds (:code:`..core.guidelines_boundaries.ConcentrationBoundary`) and diversion-ratio bounds (:code:`..core.guidelines_boundaries.DiversionRatioBoundary`), with automatic generation of boundary (as an array of share-pairs) and area. This module also includes a function for generating plots of concentration and diversion-ratio boundaries, and functions for mapping GUPPI standards to concentration (ΔHHI) standards, and vice-versa.
+
+Module :code:`.gen.data_generation` includes the :code:`.gen.data_generation.MarketSample` which provides for a rich specification of shares and diversion ratios (:code:`.gen.data_generation.MarketSample.share_spec`), margins (:code:`.gen.data_generation.MarketSample.pcm_spec`, prices (:code:`.gen.data_generation.MarketSample.price_spec`), and HSR filing requirements (:code:`.gen.data_generation.MarketSample.hsr_filing_test_type`), and with methods for, (i) generating sample data (:code:`.gen.data_generation.MarketSample.generate_sample`), and (ii) estimating enforcement or clearance rates under specified enforcement regimes given a method of aggregating diversion ratio or GUPPI estimates for the firms in a merger (:code:`.gen.data_generation.MarketSample.estimate_enf_counts`). While the latter populate the properties, :code:`.gen.data_generation.MarketSample.data`
+and :code:`.gen.data_generation.MarketSample.enf_counts`, respectively, the underlying methods for generating standalone :code:`MarketDataSample` and :code:`UPPTestCounts` objects are included in the class definition, with helper functions defined in the modules, :code:`.gen.data_generation_functions` and :code:`.gen.upp_tests`. Notably, market shares are generated for a sample of markets with firm-count distributed as specified in :code:`.gen.data_generation.MarketSample.ShareSpec.firm_count_weights`, with defaults as discussed below (also see, :code:`.gen.ShareSpec.firm_count_weights`.
+
+By default, merging-firm shares are drawn with uniform distribution over the space :math:`s_1 + s_2 \leqslant 1` for an unspecified number of firms. Alternatively, shares may be drawn from the Dirichlet distribution, with specified shape parameters (see :code:`.gen.data_generation.MarketSample.ShareSpec`, and, specifically, :code:`.gen.SHRDistribution`). When drawing shares from the Dirichlet distribution, the user passes, using :code:`.gen.data_generation.MarketSample.ShareSpec.firm_count_weights`, a vector of weights specifying the frequency distribution over sequential firm counts, e.g., :code:`[133, 184, 134, 52, 32, 10, 12, 4, 3]` to specify shares drawn from Dirichlet distributions with 2 to 10 pre-merger firms distributed as in data for FTC merger investigations during 1996--2003 (See, for example, Table 4.1 of `FTC, Horizontal Merger Investigations Data, Fiscal Years 1996--2003 (Revised: August 31, 2004) <https://www.ftc.gov/sites/default/files/documents/reports/horizontal-merger-investigation-data-fiscal-years-1996-2003/040831horizmergersdata96-03.pdf>`_). If :code:`.gen.data_generation.MarketSample.ShareSpec.firm_count_weights` is not explicitly assigned a value when defining :code:`.gen.data_generation.MarketSample.ShareSpec`, the default values is used, which results in a sample of markets with 2 to 7 firms with relative frequency in inverse proportion to firm-count, with 2-firm markets being 6 times as likely to be drawn as 7-firm markets.
+
+Recapture rates can be specified as, "proportional", "inside-out", "outside-in" (see :code:`.RECForm`. The "inside-out" specification (:code:`.gen.data_generation.MarketSample.ShareSpec.recapture_form`:code:` = `:code:`.RECForm.INOUT`) results in recapture ratios consistent with merging-firms' in-market shares and a default recapture rate. The "outside-in" specification (:code:`.gen.data_generation.MarketSample.ShareSpec.recapture_form`:code:` = `:code:`.RECForm.INOUT`) yields diversion ratios from purchase probabilities drawn at random for :math:`N+1` goods, from which are derived market shares and recapture rates for the :math:`N` goods in the putative market (see, :code:`.gen.ShareSpec`). The "outside-in" specification is invalid when the distribution of markets over firm-count is unspecified, i.e., when :code:`.gen.data_generation.MarketSample.ShareSpec.dist_type`:code:` ==`:code:`.gen.ShareDistributions.UNI`, thus raising a :code:`ValueError` exception. The "proportional" form (:code:`.gen.data_generation.MarketSample.ShareSpec.recapture_form`:code:` = `:code:`.RECForm.FIXED`) is often used in the literature, as an approximation to the "inside-out" form. See, for example, Coate (2011).
+
+Price-cost-margins may be specified as having uniform distribution, Beta distribution (including a bounded Beta distribution with specified mean and variance), or an empirical distribution (see, :code:`.gen.PCMSpec`). The empirical margin distribution is based on resampling margin data published by Prof. Damodaran of NYU Stern School of Business (see Notes), using an estimated Gaussian KDE. The second merging firm's margin (:code:`.gen.data_generation.MarketSample.PCMSpec.firm2_pcm_constraint`) may be specified as symmetric, i.i.d., or subject to equilibrium conditions for (profit-maximization in) Bertrand-Nash oligopoly with MNL demand (:code:`.gen.FM2Constraint`).
+
+Prices may be specified as symmetric or asymmetric, and in the latter case, the direction of correlation between merging firm prices, if any, can also be specified (see, :code:`.gen.PriceSpec`). Prices may also be defined by imposing cost symmetry on firms in the sample, with fixed unit marginal costs normalized to 1 unit, such that price equal :math:`1 / (1 - \pmb{m})`, where :math:`\pmb{m}` represents the array of margins for firms in the sample.
+
+The market sample may be restricted to mergers meeting the HSR filing requirement under two alternative approaches: in the one, the smaller of the two merging firms meets the lower HSR size threshold ($10 million, as adjusted) and the larger of the two merging firms meets the size test if it's share is no less than 10 times the share of the smaller firm. In the other, the :math:`n`-th firm's size is maintained as $10 million, as adjusted (see, :code:`.gen.SSZConstant`), and a merger meets the HSR filing test if either, (a.) the smaller merging firm is no smaller than the n-th firm and the larger merging firm is at 10-times as large as the n-th firm, or (b.) the smaller merging firm's market share is in excess of 10%; in effect this version of the test maintains that if the smaller merging firm's market share exceeds 10%, the value of the transaction exceeds $200 million, as adjusted, and the size-of-person test is eliminated (see, FTC (2008, p. 12); the above are simplifications of the statutory HSR filing requirements). The second assumption avoids the unfortunate assumption in the first that, within the resulting sample, the larger merging firm be at least 10 times as large as the smaller merging firm, as a consequence of the full definition of the HSR filing requirement.
+
+The full specification of a market sample is given in a :code:`.gen.data_generation.MarketSample` object, including the above parameters. Data are drawn by invoking :code:`.gen.data_generation.MarketSample.generate_sample` which adds a :code:`data` property of class, :code:`.gen.MarketDataSample`. Enforcement or clearance counts are computed by invoking :code:`.gen.data_generation.MarketSample.estimate_enf_counts`, which adds an :code:`enf_counts` property of class :code:`.gen.UPPTestsCounts`. For fast, parallel generation of enforcement or clearance counts over large market data samples that ordinarily would exceed available limits on machine memory, the user can invoke the method :code:`.gen.data_generation.MarketSample.estimate_enf_counts` on a :code:`.gen.data_generation.MarketSample` object without first invoking :code:`.gen.data_generation.MarketSample.generate_sample`. Note, however, that this strategy does not retain the market sample in memory in the interests of conserving memory and maintaining high performance (the user can specify that the market sample and enforcement statistics be stored to permanent storage; when saving to current PCIe NVMe storage, the performance penalty is slight, but can be considerable if saving to SATA storage).
+
+Enforcement statistics based on FTC investigations data and test data are printed to screen or rendered to LaTex files (for processing into publication-quality tables) using methods provided in :code:`.gen.enforcement_stats`.
+
+Programs demonstrating the use of this package are included in the sub-package, :code:`.demo`.
+
+This package includes  a class, :code:`.core.pseudorandom_numbers.MulithreadedRNG` for generating random numbers with selected continuous distribution over specified parameters, and with CPU multithreading on machines with multiple virtual, logical, or physical CPU cores. This class is an adaptation from the documentation of the :code:`numpy` package, from the discussion on `multithreaded random-number generation <https://numpy.org/doc/stable/reference/random/multithreading.html>_`; the version included here permits selection of the distribution with pre-tests to catch and inform on common errors. To access these directly:
+
+.. code-block:: python
+
+    import mergeron.core.pseudorandom_numbers as prng
+
+Documentation for this package is in the form of the API Reference. Documentation for individual functions and classes is accessible within a python shell. For example:
+
+.. code-block:: python
+
+    import mergeron.core.market_sample as market_sample
+
+    help(market_sample.MarketSample)
+
+.. rubric:: References
+
+.. _coate2011:
+
+Coate, M. B. (2011). Benchmarking the upward pricing pressure model with Federal Trade
+Commission evidence. Journal of Competition Law & Economics, 7(4), 825--846. URL: https://doi.org/10.1093/joclec/nhr014.
+
+.. _ftc_premerger_guide2:
+
+FTC Premerger Notification Office. “To File or Not to File: When You Must File a Premerger Notification Report Form”. 2008 (September, revised). URL: https://www.ftc.gov/sites/default/files/attachments/premerger-introductory-guides/guide2.pdf
+
+
+.. image:: https://img.shields.io/endpoint?url=https://python-poetry.org/badge/v0.json
+   :alt: Poetry
+   :target: https://python-poetry.org/
+
+.. image:: https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/astral-sh/ruff/main/assets/badge/v2.json
+   :alt: Ruff
+   :target: https://github.com/astral-sh/ruff
+
+.. image:: https://www.mypy-lang.org/static/mypy_badge.svg
+   :alt: Checked with mypy
+   :target: https://mypy-lang.org/
+
+.. image:: https://img.shields.io/badge/License-MIT-yellow.svg
+   :alt: License: MIT
+   :target: https://opensource.org/licenses/MIT
+
+