Using from Python

To install via pip, from the command line run:

pip install pytwoway

To install via Conda, from the command line run:

conda install -c tlamadon pytwoway

Sample data: download

To run in Python:

• If you want to estimate AKM and its bias corrections:

import pandas as pd
import bipartitepandas as bpd
import pytwoway as tw
# Load data into Pandas DataFrame
df = pd.read_csv(filepath)
# Convert into BipartitePandas DataFrame
bdf = bpd.BipartiteDataFrame(i=df['i'], j=df['j'], y=df['y'], t=df['t'])
# Clean data and collapse it at the worker-firm spell level
clean_params = bpd.clean_params({'connectedness': 'leave_out_spell', 'collapse_at_connectedness_measure': True, 'drop_single_stayers': True})
bdf = bdf.clean(clean_params)
# Initialize FE estimator
fe_params = tw.fe_params({'he': True})
fe_estimator = tw.FEEstimator(bdf, fe_params)
# Fit FE estimator
fe_estimator.fit()
# Investigate results
print(fe_estimator.summary)

• If you want to estimate CRE:

import pandas as pd
import bipartitepandas as bpd
import pytwoway as tw
# Load data into Pandas DataFrame
df = pd.read_csv(filepath)
# Convert into BipartitePandas DataFrame
bdf = bpd.BipartiteDataFrame(i=df['i'], j=df['j'], y=df['y'], t=df['t'])
# Clean data
clean_params = bpd.clean_params({'connectedness': 'connected'})
bdf = bdf.clean(clean_params)
# Collapse data at the worker-firm spell level
bdf = bdf.collapse()
# Cluster
bdf = bdf.cluster()
# Convert to cross-section format
bdf = bdf.to_eventstudy().get_cs()
# Initialize CRE estimator
cre_estimator = tw.CREEstimator(bdf)
# Fit CRE estimator
cre_estimator.fit()
# Investigate results
print(cre_estimator.summary)

• If you want to estimate BLM:

import pandas as pd
import bipartitepandas as bpd
import pytwoway as tw
# Load data into Pandas DataFrame
df = pd.read_csv(filepath)
# Convert into BipartitePandas DataFrame
bdf = bpd.BipartiteDataFrame(i=df['i'], j=df['j'], y=df['y'], t=df['t'])
# Clean data
bdf = bdf.clean()
# Collapse data at the worker-firm spell level
bdf = bdf.collapse()
# Cluster
n_firm_types = 6
cluster_params = bpd.cluster_params({'grouping': bpd.grouping.KMeans(n_clusters=n_firm_types)})
bdf = bdf.cluster(cluster_params)
# Convert to event study format
bdf = bdf.to_eventstudy()
# Separate movers and stayers
movers = bdf.get_worker_m()
jdata = bdf.loc[movers, :]
sdata = bdf.loc[~movers, :]
# Initialize BLM estimator
blm_estimator = tw.BLMEstimator(tw.blm_params({'nk': n_firm_types}))
# Fit BLM estimator
blm_estimator.fit(jdata, sdata)
# Plot results
blm_estimator.plot_log_earnings()
blm_estimator.plot_type_proportions()

• If you want to estimate Sorkin:

import pandas as pd
import bipartitepandas as bpd
import pytwoway as tw
# Load data into Pandas DataFrame
df = pd.read_csv(filepath)
# Convert into BipartitePandas DataFrame
bdf = bpd.BipartiteDataFrame(i=df['i'], j=df['j'], y=df['y'], t=df['t'])
# Clean data and collapse it at the worker-firm spell level
clean_params = bpd.clean_params({'connectedness': 'strongly_connected'})
bdf = bdf.clean(clean_params).collapse()
# Convert to event study format
bdf = bdf.to_eventstudy()
# Initialize Sorkin estimator
sorkin_estimator = tw.SorkinEstimator()
# Fit Sorkin estimator
sorkin_estimator.fit(bdf)
# Investigate results
print(sorkin_estimator.V_EE)

• If you want to estimate Borovickova-Shimer:

import pandas as pd
import bipartitepandas as bpd
import pytwoway as tw
# Load data into Pandas DataFrame
df = pd.read_csv(filepath)
# Convert into BipartitePandas DataFrame
bdf = bpd.BipartiteDataFrame(i=df['i'], j=df['j'], y=df['y'], t=df['t'])
# Clean data and collapse it at the worker-firm spell level (make sure to drop returns)
clean_params = bpd.clean_params({'drop_returns': 'returns'})
bdf = bdf.clean(clean_params).collapse()
# Make sure all workers and firms have at least 2 observations
bdf = bdf.min_joint_obs_frame()
# Clean up ids
bdf = bdf.clean()
# Initialize Borovickova-Shimer estimator
bs_estimator = tw.BSEstimator()
# Fit Borovickova-Shimer estimator
bs_estimator.fit(bdf)
# Investigate results
print(bs_estimator.res)

Check out the notebooks for more detailed examples!