BLP-Demand (Deprecated)

All current development is being done with Jeff Gortmakers pyblp package. See (https://github.com/jeffgortmaker/pyblp)

This is recommended only for teaching purposes now -- please use pyBLP

This package contains a state of the art implementation of the Nested Fixed Point (NFP) approach estimating demand using approach of Berry Levinsohn and Pakes (1995) (BLP).

The primary features are:

Single User-configurable file allows for off diagonal random coefficients, non-normal random coefficients, etc.
Solves for marketshares in parallel (market by market)
Option to solve marketshares using Newton's Method
Option to solve marketshares using Modified Fixed Point Iteration approach of Reynaerts, Varadhan, Nash (2012). This is VERY fast! (Default)
Computes second stage weighting matrix and standard errors (needs better error checking!)
Can use either knitromatlab or fmincon (knitromatlab highly recommended)
Optimized for Matlab R2015a

Package Description

This package contains the following files

extract_params.m:

this is a user-configurable file that MUST be edited and determines the correct specification passed to the optimization routine. It is where you specify the random components of utility and the corresponding derivatives. Two examples of this file are provided: indep_normal.m and correlated_normal.m for independently distributed and correlated random coefficients, respectively.

solveRCBLPpar.m:

this is the main routine that optimizes the BLP-GMM objective function, computes the weighting matrix, and produces the standard errors. As an intermediary output it produces first-step.mat for diagnostic purposes. Only the first few lines of the file are considered user configurable.

example.m:

this is an example script that walks the user through the steps required to estimate the model. These steps are as follows:

import data in "table" format (see the object dtable in sample.mat)
set starting values for the random coefficient parameters (size of this vector is problem-specific)
produce draws for random coefficients (see the structure draws in sample.mat)
call the function solveRCBLPpar.m by providing the data (in "table" format), the draws, the starting values for the random coefficients, and a string containing the name of the file that extracts the parameters.
display results

These are not considered user-configurable files

solveAllShares.m:

This is the main routine which concentrates out the mean utility parameters by solving the share equations market by market. If the computer has multiple cores, it is capable of spreading this task in parallel over several cores. It takes on two user-configurable options 'newton' sovles for the shares using Newton's Method (fsolve) and 'fixed-point' which solves for the shares using a modified fixed-point algorithm.

solveNewton.m:

This solves the system of J share equations for a single market using Newton's method (fsolve) and requires computing the Jacobian of the shares with respect to the mean utilities at each iteration

fp_squarem.m:

This is a port of the SQUAREM fixed-point algorithm in the R package "SQUAREM". It does not have as many options and only handles the simplest cases. It is a generic implmentation of SQUAREM and is not BLP specific. In this context it is used to solve the system of share equations for the mean utilities market by market.

rc_share_safe.m:

This is an under/over-flow safe implementation of the random coefficients logit probability for a SINGLE market. This should be re-written in C++/mex for optimal performance.

RCBLP_Jacobian.m

This computes the derivative of the shares with respect to the mean utilities (deltas) and nonlinear parameters (theta) for a single market. It has two calling modes. In the 'mpec' mode it returns both the delta and theta derivatives. In the 'sterr' mode it returns the d s / d theta = [d s/ d delta]^-1 d s/d theta. For more information consult the appendix to Nevo (2000).

License details

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

The views and conclusions contained in the software and documentation are those of the authors and should not be interpreted as representing official policies, either expressed or implied, of the FreeBSD Project.

chrisconlon/blp-demand