Non-financial Liabilities and Restructuring

Empirical support from 2005 Reform

Author

Hongyi Xu

Published

January 10, 2025

Modified

February 28, 2025

This analysis is suggested by discussant Samuel Antill in 2025 AFA meeting.

S1. Download and Load the Raw Dataset

We download the firm-level capital structure = from Compustat-CapitalIQ North America. The first measure of executory contracts is accounting data which (under U.S. GAAP) cover leases and rental contracts. We use the variables “Debt Equivalent of Operating Leases”, “Capital Leases” and “Net Rental Expenses”.

Debt - Capitalized Lease Obligations (dclo): This item represents long-term obligations due in respect of lease finances and hire purchase arrangements. This item includes:
- Long-term finance lease obligations
- Long-term lease financing
- Long-term hire purchase obligations
- Non-current obligations under capital leases
Rental Expense (xrent): This item represents all costs charged to operations for rental, lease, or hire of space and/or equipment. This item includes:
- Airlines’ landing fees
- Contingent rentals associated with capitalized leases
- Lease charges and plant hire
This xrent variable is used in Section 5.1. In the main test, we construct a new variable called total rent, which is defined as the total rental commitment mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct reported in year t.
Asset - Total (at): Book value of total assets.
Debt in Current Liabilities - Total (dlc) is for the short-term debt and Long-term debt - Total (dltt) is for the long-term debt.
Earnings Before Interest (EBITDA): EBITDA.
Industry classifications: naics, sic and spcindcd + (Fama-French 30 Industry Classifiers, To Be Added / Definition).

We download all annual observations with Data Date (datadate) between Jan 2000 and Dec 2007 and restrict to firms either incorporated or having headquarters in US (i.e. fic == "USA" | loc == "USA"). Given that the policy shock happens under the US jurisdiction, we expect more significant treatment effect to be detected in this sample. The period used in Samuel’s slides is between 2002 and 2007, which is after the doc.com bubble and after the GFC. The variables in the dataset are:

Show the code

## load the compustat_na data 
# compustat_na <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/g4f68dkn3beu1iag.csv") )
compustat_na <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/azsyeo1yjmrb2q1v.csv") ) %>%
  filter(fic == "USA" | loc == "USA" ) ## keep only US related observations (| fic == "CAN" | loc == "CAN")

## data file record: 
# | Feb 24, 2025 > azsyeo1yjmrb2q1v.csv 
# | Jan 14, 2025 > g4f68dkn3beu1iag.csv 

## check the variable names 
cat("Variables include:\n\n")
sort(names(compustat_na))

Variables include:

 [1] "at"       "cik"      "conm"     "consol"   "costat"   "curcd"   
 [7] "cusip"    "datadate" "datafmt"  "dclo"     "dlc"      "dltt"    
[13] "ebitda"   "fic"      "fyear"    "fyr"      "fyrc"     "gvkey"   
[19] "indfmt"   "loc"      "mrc1"     "mrc2"     "mrc3"     "mrc4"    
[25] "mrc5"     "mrct"     "naics"    "popsrc"   "sic"      "spcindcd"
[31] "tic"      "xrent"

S2. Data Cleaning and Prep

Two dependent variables are constructed to measure firm’s leverage: (cf. p18 and Table 2 of the paper)

Book leverage: we define leverage as the ratio of the sum of short-term debt and long-term debt to assets (we drop observations where this is outside the [0,1] range)*.
Debt/EBITDA: we define debt-to-EBITDA as the ratio of the long- and short-term debt to EBITDA (we drop observations where this is outside the [0,15] range)*.

*These restrictions are imposed only in analyses in Section 5.2.

Note 1

What should be the best way to construct this PV of Rent / Net Rental Expenses variable in this test?

Show the code

## set the base year 
base_year <- 2005 

## create leverage variables 
data <- compustat_na %>% 
  # filter(at > 0) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    # contract_intensity = (totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year 
  ) 

## record the variable names 
cat("Variables include:\n\n")
sort(names(data))

Variables include:

 [1] "at"                 "book_leverage"      "cik"               
 [4] "conm"               "consol"             "contract_intensity"
 [7] "costat"             "curcd"              "cusip"             
[10] "datadate"           "datafmt"            "dclo"              
[13] "debt_ebitda"        "dlc"                "dltt"              
[16] "dyear"              "ebitda"             "fic"               
[19] "fyear"              "fyr"                "fyrc"              
[22] "gvkey"              "indfmt"             "loc"               
[25] "mrc1"               "mrc2"               "mrc3"              
[28] "mrc4"               "mrc5"               "mrct"              
[31] "naics"              "popsrc"             "sic"               
[34] "spcindcd"           "tic"                "totalrent"         
[37] "xrent"

The main independent variable in the test is executory contract intensity, which is named as contract_intensity in our dataset. We follow Samuel’s definition w/o purchase contracts information for firm i at year t:

\begin{equation} \begin{split} & \text{Executory Contract Intensity}_{i,t} = \frac{\text{Capitalized Leases}_{i,t} + \text{PV of Rent}_{i,t} }{\text{Book Asset}_{i,t}} \\ & \Rightarrow \text{contract\_intensity}_{i,t} = \begin{cases} \frac{\text{dclo}_{i,t} + \text{xrent}_{i,t} \times 3 }{\text{at}_{i,t}} \\ \frac{\text{dclo}_{i,t} + \text{totalrent}_{i,t} }{\text{at}_{i,t}}\\ \end{cases}, \\ & \quad \text{ with totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct} \\ \end{split} \end{equation}

S3. Difference-in-Difference Examination

Then, we construct the dataset for the difference-in-difference analysis.

In this section, we use all observations without missing values in subsequent analyses.

The original dataset (data) contains observations with data year (dyear) range in [1997, 2009] and we restrict the sample to the data year in [2002, 2007].

For results with restrictions on specific variables, please refer to Section 5.2.

Note 2

Does altering the time (i.e. 2000 to 1995) for measuring the treatment intensity alter the results?

s3.1. try-and-error with calender year `dyear`

In this section, we first try to use the year information dyear extracted from variable datadate as the periods.

On top of the base data, we add several new variables in dataset data_did:

contract_intensity_2000: the contract intensity for each firm in year dyear == 2000.
post_2005: an indicator equals to 1 if dyear > 2005 and 0 otherwise.

Show the code

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classifiaction 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) # %>% 
  # drop values outside the range:
  # mutate(
  #   book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA),
  #   debt_ebitda = ifelse(test = (debt_ebitda >= 0 & debt_ebitda <= 15), yes = debt_ebitda, no = NA )
  # )

DiD Analysis:

Note 3

Q: Do we need to remove IPO and delisting firms?

Important 1: More Fixed-effects

Adding the industry-year fixed effects does not change the regression results.

Show the code

data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    ind = as.factor(ff_30ind_name), # industry 
    ind_time = as.factor(paste(ff_30ind_name, dyear, sep = " + ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  ## adding industry*year FE does not change the results: + ind_time 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  fixest::etable()

                                                      .
Dependent Var.:                           book_leverage
                                                       
contract_intensity_2000 x post_2005 -0.4184*** (0.1028)
Fixed-Effects:                      -------------------
gvkey                                               Yes
dyear                                               Yes
___________________________________ ___________________
S.E.: Clustered                       by: gvkey & ff_3.
Observations                                     22,803
R2                                              0.39586
Within R2                                       0.00336
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, indexyear, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 1998 & dyear <= 2007) %>%
  mutate(
    gvkey = as.factor(gvkey) #, 
    # dyear = as.factor(dyear), 
    # indexyear = relevel(factor(indexyear), ref = "0") 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ 
                  # i(dyear):contract_intensity_2000
                  i(indexyear, contract_intensity_2000, ref = -1)
                | gvkey + indexyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  iplot(., 
        xlab = paste("Time to treatment (0 is Year ", base_year, ")", sep = ""),
        main = "Treatment Effect (TWFE): calendar year", 
        ref.line = FALSE,
        ref.line.par = list(col = "red", lty = 2),
        pt.join = FALSE
        ); axis(1, at = -7:2); abline(v = 0.5, col = "red", lty = 2, lwd = 3)

Show the code

  # fixest::coefplot(sub = "base year: dyear = 2004")

s3.2. try-and-error with calender year `fyear`

In this section, we first try to use the year information fyear extracted from variable datadate as the periods.

On top of the base data, we add several new variables in dataset data_did:

contract_intensity_2000: the contract intensity for each firm in year fyear == 2000.
post_2005: an indicator equals to 1 if fyear > 2005 and 0 otherwise.

Show the code

data_firm <- data %>% 
  filter(fyear == 2000) %>% # use `fyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) 
  
FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classifiaction 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (fyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(fyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  )

DiD Analysis:

Show the code

data_did %>% 
  select(gvkey, fyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(fyear >= 2002 & fyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    fyear = as.factor(fyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + fyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  fixest::etable()

                                                      .
Dependent Var.:                           book_leverage
                                                       
contract_intensity_2000 x post_2005 -0.4250*** (0.1023)
Fixed-Effects:                      -------------------
gvkey                                               Yes
fyear                                               Yes
___________________________________ ___________________
S.E.: Clustered                       by: gvkey & ff_3.
Observations                                     22,689
R2                                              0.30332
Within R2                                       0.00113
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

data_did %>% 
  select(gvkey, fyear, contract_intensity_2000, post_2005, indexyear, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(fyear >= 1998 & fyear <= 2007) %>%
  mutate(
    gvkey = as.factor(gvkey)  
    # fyear = as.factor(fyear), 
    # indexyear = relevel(factor(indexyear), ref = "0") 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ i(indexyear, contract_intensity_2000, -1) | gvkey + indexyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  iplot(., 
        xlab = paste("Time to treatment (0 is Year ", base_year, ")", sep = ""),
        main = "Treatment Effect (TWFE): fiscal year", 
        ref.line = FALSE,
        ref.line.par = list(col = "red", lty = 2),
        pt.join = FALSE
        ); axis(1, at = -7:2); abline(v = 0.5, col = "red", lty = 2, lwd = 3)

Show the code

  # fixest::coefplot(sub = "base year: fyear = 2000")

S4. [Deprecated] Industry Level DiD

We start by constructing industry-level information.

Show the code

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>%
  group_by(gvkey) %>%
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>%
  ungroup()

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

## industry level measures
data_did_ind <- data %>%
  ## merge the grouping variable
  left_join(y = data_firm, by = join_by(gvkey) ) %>%
  ## merge the FF 30 industry classifiaction
  left_join(y = FF_30ind, by = join_by(sic)) %>%
  select(gvkey, dyear, contract_intensity, book_leverage, debt_ebitda, sic, ff_30ind, ff_30ind_desc) %>%
  mutate(ff_30_ind_full = paste(ifelse(ff_30ind < 10,
                                       yes = paste(" ", ff_30ind, sep = ""),
                                       no = as.character(ff_30ind)),
                                ff_30ind_desc, sep = ": ") ) %>%

  ## for each industry-year:
  group_by(dyear, ff_30_ind_full, ff_30ind, ff_30ind_desc) %>%
  summarise(
    contract_intensity_ind = median(contract_intensity, na.rm = TRUE),
    book_leverage_ind = median(book_leverage, na.rm = TRUE),
    debt_ebitda_ind = median(debt_ebitda, na.rm = TRUE)
  ) %>%
  ungroup()

data_did_ind %>%
  filter(dyear == 2000 & ff_30ind != 20) %>%
  arrange(-contract_intensity_ind) %>%
  select(-dyear, -ff_30_ind_full,
         `FF30 Code` = ff_30ind, # the industry code
         `Fama-French 30 Industry` = ff_30ind_desc, # name of the
         `Contract Intensity` = contract_intensity_ind,
         `Book Leverage` = book_leverage_ind,
         `Debt-to-EBITDA` = debt_ebitda_ind 
         ) %>%
  gt(
    rowname_col = "FF30 Code"
  ) %>%
  fmt_number(
    columns = c(`Contract Intensity`, `Book Leverage`, `Debt-to-EBITDA`),
    decimals = 3
  ) %>%
  tab_header(
    title = "Table: Contract Intensity by Industry in 2000"
  ) %>%
  tab_style(
    style = cell_text(size = "small"),  # Reduce font size
    locations = cells_column_labels(columns = everything()) # reduce the size of tab headers
  ) %>%
  tab_style(
    style = cell_text(size = "small"),  # Reduce font size
    locations = cells_body()
  )

Table: Contract Intensity by Industry in 2000
	Fama-French 30 Industry	Contract Intensity	Book Leverage	Debt-to-EBITDA
28	Restaurants, Hotels, Motels	0.566	0.348	2.227
27	Retail	0.538	0.228	1.058
22	Personal and Business Services	0.195	0.061	0.000
25	Transportation	0.188	0.321	2.128
7	Apparel	0.167	0.255	0.735
5	Printing and Publishing	0.137	0.276	1.540
8	Healthcare, Medical Equipment, Pharmaceutical Products	0.131	0.098	0.000
30	Everything Else	0.118	0.276	0.000
26	Wholesale	0.115	0.302	1.921
2	Beer & Liquor	0.115	0.312	0.917
23	Business Equipment	0.114	0.064	0.000
4	Recreation	0.106	0.374	0.940
18	Coal	0.095	0.521	3.799
6	Consumer Goods	0.094	0.307	1.343
13	Fabricated Products and Machinery	0.083	0.260	1.267
10	Textiles	0.079	0.476	3.529
21	Communication	0.079	0.377	0.014
14	Electrical Equipment	0.076	0.216	0.461
16	Aircraft, ships, and railroad equipment	0.075	0.275	1.752
9	Chemicals	0.071	0.313	1.773
24	Business Supplies and Shipping Containers	0.071	0.396	2.564
15	Automobiles and Trucks	0.067	0.363	2.514
1	Food Products	0.065	0.378	2.414
11	Construction and Construction Materials	0.058	0.311	1.754
12	Steel Works Etc	0.056	0.340	3.191
3	Tobacco Products	0.049	0.256	1.208
29	Banking, Insurance, Real Estate, Trading	0.048	0.150	3.215
17	Precious Metals, Non-Metallic, and Industrial Metal Mining	0.042	0.181	0.000
19	Petroleum and Natural Gas	0.028	0.254	0.958

Show the code

## Plot the top and bottom parts:

# data_did_ind %>%

#   ggplot(aes(x = dyear, y = contract_intensity_ind, colour = ff_30_ind_full)) +

#   geom_line() +

#   geom_point(aes(shape = ff_30_ind_full)) +

#   scale_x_continuous(n.breaks = 10, name = NULL)

Then, we perform DiD analysis using the industry-level contract intensity. The hypothesis in this test is that, after the option value of rejecting non-financial obligations decreases, the debt capacity of the firm also decreases. By replacing the contract intensity at the firm level by that at the industry level in 2000, we re-run the standard TWFE regression as in Section 3.

\begin{equation} \text{Book Leverage}_{i,j,t} = \theta_{i} + \gamma_t + \beta \text{Contract Intensity}^{2000}_j \times \text{Post}^{2005}_t + \varepsilon_{i,j,t} \end{equation}

Show the code

fixest::etable(
  ## industry level difference-in-difference analysis
  ## firm-level dependent variable:
  data %>%
  ## merge the grouping variable
  left_join(y = data_firm, by = join_by(gvkey) ) %>%
  ## merge the FF 30 industry classifiaction
  left_join(y = FF_30ind, by = join_by(sic)) %>%
  select(gvkey, dyear, book_leverage, debt_ebitda, sic, ff_30ind) %>%
  left_join(y = filter(data_did_ind, dyear == 2000) %>%
              `names<-`(value = str_replace(string = names(.),
                                            pattern = "_ind$",
                                            replacement = "_ind2000")) %>%
              select(-dyear),
            by = join_by(ff_30ind)
  ) %>%
  filter(dyear >= 2002 & dyear <= 2007) %>%
  mutate(
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0), # whether it is after the 2005 reform
    gvkey = as.factor(gvkey),
    dyear = as.factor(dyear),
    ind = as.factor(ff_30ind), # industry
    ind_time = as.factor(paste(ff_30_ind_full, dyear, sep = " + "))
  ) %>%
  na.omit() %>%
  fixest::feols(fml = book_leverage ~ contract_intensity_ind2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind")),
  ## industry-level dependent variable
  data_did_ind %>%
  filter(dyear >= 2002 & dyear <= 2007) %>%
  left_join(y = filter(data_did_ind, dyear == 2000) %>%
              select(-ff_30_ind_full, -ff_30ind_desc) %>%
              `names<-`(value = str_replace(string = names(.),
                                            pattern = "_ind$",
                                            replacement = "_ind2000")) %>%
              select(-dyear),
            by = join_by(ff_30ind)
  ) %>%
  mutate(
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0), # whether it is after the 2005 reform
    ff_30ind = as.factor(ff_30ind),
    dyear = as.factor(dyear)
  ) %>%
  na.omit() %>%
  fixest::feols(fml = book_leverage_ind ~ contract_intensity_ind2000:post_2005 | ff_30ind + dyear, data = ., cluster = c("ff_30ind"))
) %>%
  knitr::kable(format = "html", caption = "DiD Results with Industry-level Contract Intensity", col.names = c("", paste("(", 1:2, ")", sep = ""))) %>%
  kableExtra::kable_styling(
    font_size = 8,       # Smaller font (default is 16px)
    full_width = FALSE,   # Table width fits content
    bootstrap_options = c("striped", "condensed")
  )

DiD Results with Industry-level Contract Intensity
	(1)	(2)
Dependent Var.:	book_leverage	book_leverage_ind

contract_intensity_ind2000 x post_2005	0.9357 (1.703)	0.0739 (0.0543)
Fixed-Effects:	--------------	-----------------
gvkey	Yes	No
dyear	Yes	Yes
ff_30ind	No	Yes
______________________________________	______________	_________________
S.E.: Clustered	by: gvkey & ff_3.	by: ff_30ind
Observations	43,490	180
R2	0.42987	0.87141
Within R2	2.64e-6	0.01510

However, the results show that the industry level evidence is not significant to support the hypothesis. In the regressions using industry-level contact intensities in year 2000 for grouping, the dependent variable is firm level book leverage in regression (1) and industry level median book leverage in regression (2). The coefficient estimates of the interaction term are both positive, which is the opposite of our main regressions. This indicates that the industry-level intensity may not be a good measure for our firm level analysis in Section 3.

For more firm-level analysis for different industries, please refer to Section 5.5.

Appendix

Here are additional tests results with alternative specifications.

A1. Results using different measures of “Net Rental Expenses”

Show the code

## create leverage variables 
data <- compustat_na %>% 
  # filter(at > 0) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms 
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    # totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    # contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year
  )

In this part of the analysis, I use the definition of “Net Rental Expenses” in our paper, namely Debt in Current Liabilities (Total) + 3 * xrent and use this to construct a different measure of contract intensity.

A1.1. try-and-error with calender year `dyear`

In this section, we first try to use the year information dyear extracted from variable datadate as the periods.

On top of the base data, we add several new variables in dataset data_did:

contract_intensity_2000: the contract intensity for each firm in year dyear == 2000.
post_2005: an indicator equals to 1 if dyear > 2005 and 0 otherwise.

Show the code

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classifiaction 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) # %>% 
  # drop values outside the range:
  # mutate(
  #   book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA),
  #   debt_ebitda = ifelse(test = (debt_ebitda >= 0 & debt_ebitda <= 15), yes = debt_ebitda, no = NA )
  # )

DiD Analysis:

Show the code

data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  fixest::etable()

                                                    .
Dependent Var.:                         book_leverage
                                                     
contract_intensity_2000 x post_2005 -0.9712* (0.4132)
Fixed-Effects:                      -----------------
gvkey                                             Yes
dyear                                             Yes
___________________________________ _________________
S.E.: Clustered                     by: gvkey & ff_3.
Observations                                   24,307
R2                                            0.37021
Within R2                                     0.00335
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, indexyear, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2000 & dyear <= 2007) %>%
  mutate(
    gvkey = as.factor(gvkey) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ i(indexyear, contract_intensity_2000, -1) | gvkey + indexyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  iplot(., 
        xlab = paste("Time to treatment (0 is Year ", base_year, ")", sep = ""),
        main = "Treatment Effect (TWFE): calendar year", 
        ref.line = FALSE,
        ref.line.par = list(col = "red", lty = 2),
        pt.join = FALSE
        ); axis(1, at = -7:2); abline(v = 0.5, col = "red", lty = 2, lwd = 3)

Show the code

  # fixest::coefplot()

A1.2. try-and-error with calender year `fyear`

In this section, we first try to use the year information fyear extracted from variable datadate as the periods.

On top of the base data, we add several new variables in dataset data_did:

contract_intensity_2000: the contract intensity for each firm in year fyear == 2000.
post_2005: an indicator equals to 1 if fyear > 2005 and 0 otherwise.

Show the code

data_firm <- data %>% 
  filter(fyear == 2000) %>% # use `fyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) 
  
FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classifiaction 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (fyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(fyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) # %>% 
  # drop values outside the range:
  # mutate(
  #   book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA),
  #   debt_ebitda = ifelse(test = (debt_ebitda >= 0 & debt_ebitda <= 15), yes = debt_ebitda, no = NA )
  # )

DiD Analysis:

Show the code

data_did %>% 
  select(gvkey, fyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(fyear >= 2002 & fyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    fyear = as.factor(fyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + fyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  fixest::etable()

                                                    .
Dependent Var.:                         book_leverage
                                                     
contract_intensity_2000 x post_2005 -0.9878* (0.4243)
Fixed-Effects:                      -----------------
gvkey                                             Yes
fyear                                             Yes
___________________________________ _________________
S.E.: Clustered                     by: gvkey & ff_3.
Observations                                   24,133
R2                                            0.29925
Within R2                                     0.00130
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

data_did %>% 
  select(gvkey, fyear, contract_intensity_2000, post_2005, indexyear, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(fyear >= 2000 & fyear <= 2007) %>%
  mutate(
    gvkey = as.factor(gvkey) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ i(indexyear, contract_intensity_2000, -1) | gvkey + indexyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  iplot(., 
        xlab = paste("Time to treatment (0 is Year ", base_year, ")", sep = ""),
        main = "Treatment Effect (TWFE): fiscal year", 
        ref.line = FALSE,
        ref.line.par = list(col = "red", lty = 2),
        pt.join = FALSE
        ); axis(1, at = -7:2); abline(v = 0.5, col = "red", lty = 2, lwd = 3)

Show the code

  # fixest::coefplot()

A2. Results using data filters in the main paper

In this section, we restrict to observations with Book leverage inside the [0,1] range over the full sample.

Show the code

## create leverage variables 
data <- compustat_na %>% 
  # filter(at > 0) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms 
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year
  ) 

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classifiaction 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) %>% 
  # drop values outside the range:
  mutate(
   book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA)
  )

First, we look at some summary statistics for the book_leverage variable.

Summary statistics for the book_leverage variable:

Show the code

data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  na.omit() %>% 
  select(book_leverage) %>% 
  # mutate(
  #   book_leverage_1 = ifelse(book_leverage >= 0 & book_leverage <= 1, yes = book_leverage, no = NA), 
  #   book_leverage_3 = ifelse(book_leverage >= 0 & book_leverage <= 3, yes = book_leverage, no = NA), 
  #   book_leverage_5 = ifelse(book_leverage >= 0 & book_leverage <= 5, yes = book_leverage, no = NA) 
  # ) %>% 
  data.frame() %>% 
  psych::describe(.)

              vars     n mean   sd median trimmed  mad min max range skew
book_leverage    1 52294 0.22 0.23   0.16    0.19 0.24   0   1     1 1.03
              kurtosis se
book_leverage     0.37  0

Show the code

data %>% 
  select(gvkey, dyear, book_leverage, debt_ebitda) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  # select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  na.omit() %>% 
  select(book_leverage) %>% 
  filter(book_leverage >= 0 & book_leverage <= 3) %>%
  .$book_leverage %>% 
  hist(., 
       prob = TRUE, # density plot
       col = "#E1DEFC",  
       main = "Density Histogram: raw data from Compustat NA (US only)", 
       xlab = "Book leverage (only show the region [0,3])",
       ylab = "Density", 
       breaks = 50, xlim = c(0, 3) 
       )

Show the code

data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, dlc, dltt, at, book_leverage, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  arrange(-book_leverage) %>% 
  na.omit()

# A tibble: 22,039 × 9
   gvkey dyear contract_intensity_2…¹    dlc    dltt      at book_leverage   sic
   <int> <int>                  <dbl>  <dbl>   <dbl>   <dbl>         <dbl> <int>
 1  8045  2006                 0       0.005 4.5 e-2 5   e-2         1      6500
 2 66660  2007                 0.0119  7.8   1.75e+3 1.76e+3         1.00   1520
 3 61584  2004                 1.48   19.6   6.97e+2 7.17e+2         1.00   8300
 4 13472  2002                 0.0844  0.196 2.89e+0 3.09e+0         0.999  3823
 5  8526  2004                 0.227   2.61  3.67e+1 3.94e+1         0.998  2834
 6 61108  2005                 0.441   0.2   3.3 e-1 5.31e-1         0.998  3826
 7 30950  2005                 0.0266  3.04  7.83e+2 7.88e+2         0.998  4833
 8  6773  2002                 0.466   0.854 3.40e+0 4.26e+0         0.998  7370
 9  8526  2003                 0.227   8.44  3.64e+1 4.50e+1         0.997  2834
10 12252  2005                 0.594   0     7.24e+1 7.26e+1         0.997  2836
# ℹ 22,029 more rows
# ℹ abbreviated name: ¹contract_intensity_2000
# ℹ 1 more variable: ff_30ind_name <chr>

DiD Analysis:

Show the code

etable(
  `(1)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")), 
  `(2)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind_name^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 
)

                                                  (1)               (2)
Dependent Var.:                         book_leverage     book_leverage
                                                                       
contract_intensity_2000 x post_2005   0.0031 (0.0020)   0.0028 (0.0019)
Fixed-Effects:                        ---------------   ---------------
gvkey                                             Yes               Yes
dyear                                             Yes               Yes
ff_30ind_name-dyear                                No               Yes
___________________________________   _______________   _______________
S.E.: Clustered                     by: gvkey & ff_3. by: gvkey & ff_3.
Observations                                   21,599            21,599
R2                                            0.79481           0.79807
Within R2                                     0.00020           0.00017
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

model_a2_a <- data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, indexyear, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2001 & dyear <= 2007) %>%
  mutate(
    gvkey = as.factor(gvkey) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ i(indexyear, contract_intensity_2000, -4) | gvkey + indexyear, data = ., cluster = c("gvkey", "ff_30ind_name"))

model_a2_b <- data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, indexyear, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2001 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ i(indexyear, contract_intensity_2000, -4) | gvkey + indexyear + ff_30ind_name^indexyear, data = ., cluster = c("gvkey", "ff_30ind_name")) # with industry-year FE. 

iplot(list(model_a2_a, model_a2_b), 
        xlab = paste("Time to treatment (0 is Year ", base_year, ")", sep = ""),
        main = "Treatment Effect (TWFE): Book leverage in [0,1]", 
        ref.line = FALSE,
        ref.line.par = list(col = "red", lty = 2),
        pt.join = FALSE
        ); axis(1, at = -7:2); abline(v = 0.5, col = "red", lty = 2, lwd = 3); legend("bottomleft", bty = "n", col = c(1, 2), pch = c(20, 17), legend = c("TWFE", "TWFE + Industry-Year FE"))

Show the code

  # fixest::coefplot()

Compare to results in Section 3.1, the coefficient estimate on the treatment in Section 5.2 not only becomes insignificant, but also changes the direction of the impact.

A3. Sensitivity Analysis of Choosing 2000

In this subsection, we examine how sensitive the coefficient estimate is with respect to changes in the base year for determining the contract intensity of a firm. We still use dyear as the year variable.

Show the code

data_firm_wide <- data %>% 
  select(dyear, gvkey, contract_intensity) %>% 
  group_by(dyear, gvkey) %>% 
  summarise(contract_intensity = mean(contract_intensity, na.rm = TRUE) ) %>% 
  ungroup() %>% 
  filter(dyear < 2002) %>% 
  pivot_wider(names_from = dyear, values_from = contract_intensity) %>% 
  rename_with(.fn = ~ ifelse(grepl(pattern = "\\d{4}", x = .x), paste0("contract_intensity_", .x), no = .x) )

data_did_wide <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm_wide, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classifiaction 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform
  )

We construct a series of new column variables under the name of contract_intensity_[year] with dyear < 2002.

DiD Analysis:

Show the code

reg_contract_intensity_list <- list(); for (contract_base in names(data_firm_wide)[-1] ) {
  ## format the regression
  reg_raw_year <- as.formula(paste("book_leverage ~ ", contract_base, ":post_2005 | gvkey + dyear", sep = "")) 
  ## use different base year(s) 
  reg_year <- data_did_wide %>% 
    select(gvkey, dyear, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, (!!sym(contract_base)) ) %>% 
    filter(dyear >= 2002 & dyear <= 2007) %>% 
    mutate(
      gvkey = as.factor(gvkey), 
      dyear = as.factor(dyear) 
    ) %>% 
    na.omit() %>% 
    fixest::feols(fml = reg_raw_year, data = ., cluster = c("gvkey", "ff_30ind_name")) 
  
  reg_contract_intensity_list[[contract_base]] <- reg_year
  
  assign(x = paste("reg_", contract_base, sep = ""), reg_year) 
} 
## change the name of each did regressions: 
names(reg_contract_intensity_list) <- str_replace_all(string = names(reg_contract_intensity_list), pattern = "contract_intensity_", replacement = "base: dyear=")

etable(reg_contract_intensity_list, vcov = "cluster") %>% 
  knitr::kable(format = "html", caption = "DiD Results with Different Base Year for Contract Intensity") %>% 
  kableExtra::kable_styling(
    font_size = 8,       # Smaller font (default is 16px)
    full_width = FALSE,   # Table width fits content 
    bootstrap_options = c("striped", "condensed")
  )

DiD Results with Different Base Year for Contract Intensity
	base: dye..1997	base: dyear=1998	base: dyear=1999	base: dyear=2000	base: dyear=2001
Dependent Var.:	book_leverage	book_leverage	book_leverage	book_leverage	book_leverage

contract_intensity_1997 x post_2005	0.0676 (0.3000)
contract_intensity_1998 x post_2005		-0.0003 (0.0005)
contract_intensity_1999 x post_2005			-1.797*** (0.2602)
contract_intensity_2000 x post_2005				-0.4184*** (0.1057)
contract_intensity_2001 x post_2005					-0.4592*** (0.0129)
Fixed-Effects:	---------------	----------------	------------------	-------------------	-------------------
gvkey	Yes	Yes	Yes	Yes	Yes
dyear	Yes	Yes	Yes	Yes	Yes
___________________________________	_______________	________________	__________________	___________________	___________________
S.E.: Clustered	by: gvkey	by: gvkey	by: gvkey	by: gvkey	by: gvkey
Observations	17,349	19,003	21,231	22,803	24,018
R2	0.31625	0.51764	0.52312	0.39586	0.31542
Within R2	9.14e-7	9.11e-8	0.00434	0.00336	0.00290

Show the code

coefplot(
  reg_contract_intensity_list, 
  horiz = TRUE
  )

From the regression results, we can see that the treatment effect remains negatively statistically significant at 1% level, if we choose 1999 or 2001 as the base year to measure the contract intensity characteristic of a firm. Choosing years before 1999 will make the treatment effect insignificant. This is understandable as firms characteristics before 1999 may not be a good measure for its characteristics between 2002 and 2007. Our results are robust to most of adjustments in the standard errors. In the base tests, results are invariant when we cluster standard errors at the firm level and also at both the firm and year level.

However, the results are no longer significant when we use the heteroscedasticity-robust (HC1) corrected standard errors, even when using 2000 as the base year for contract_intensity. See the regression results below:

Show the code

etable(reg_contract_intensity_list, vcov = "HC1") %>% 
  knitr::kable(format = "html", caption = "DiD Results with Different Base Years for Contract Intensity") %>% 
  kableExtra::kable_styling(
    font_size = 8,       # Smaller font (default is 16px)
    full_width = FALSE,   # Table width fits content 
    bootstrap_options = c("striped", "condensed")
  )

DiD Results with Different Base Years for Contract Intensity
	base: dye..1997	base: dyear=1998	base: dy..1999	base: dyear=2000	base: dyear=2001
Dependent Var.:	book_leverage	book_leverage	book_leverage	book_leverage	book_leverage

contract_intensity_1997 x post_2005	0.0676 (0.2919)
contract_intensity_1998 x post_2005		-0.0003 (0.0005)
contract_intensity_1999 x post_2005			-1.797 (1.308)
contract_intensity_2000 x post_2005				-0.4184. (0.2329)
contract_intensity_2001 x post_2005					-0.4592 (0.3001)
Fixed-Effects:	---------------	----------------	--------------	-----------------	----------------
gvkey	Yes	Yes	Yes	Yes	Yes
dyear	Yes	Yes	Yes	Yes	Yes
___________________________________	_______________	________________	______________	_________________	________________
S.E. type	Heteroske.-rob.	Heterosked.-rob.	Heterosk.-rob.	Heteroskeda.-rob.	Heterosked.-rob.
Observations	17,349	19,003	21,231	22,803	24,018
R2	0.31625	0.51764	0.52312	0.39586	0.31542
Within R2	9.14e-7	9.11e-8	0.00434	0.00336	0.00290

The numbers in this test, however, are closer to the ones in Samuel’s slides.

In the last part, we re-run all analyses using fyear as the year variable.

Show the code

## data prep: 
data_firm_wide_fyear <- data %>% 
  select(fyear, gvkey, contract_intensity) %>% 
  filter(fyear < 2002) %>% 
  pivot_wider(names_from = fyear, values_from = contract_intensity) %>% 
  rename_with(.fn = ~ ifelse(grepl(pattern = "\\d{4}", x = .x), paste0("contract_intensity_", .x), no = .x) )

data_did_wide_fyear <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm_wide_fyear, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classifiaction 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (fyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(fyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform
  ) 

## did reg: 
reg_contract_intensity_list_fyear <- list(); for (contract_base in names(data_firm_wide)[-1] ) {
  ## format the regression
  reg_raw_year <- as.formula(paste("book_leverage ~ ", contract_base, ":post_2005 | gvkey + fyear", sep = "")) 
  ## use different base year(s) 
  reg_year <- data_did_wide %>% 
    select(gvkey, fyear, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, (!!sym(contract_base)) ) %>% 
    filter(fyear >= 2002 & fyear <= 2007) %>% 
    mutate(
      gvkey = as.factor(gvkey), 
      fyear = as.factor(fyear) 
    ) %>% 
    na.omit() %>% 
    fixest::feols(fml = reg_raw_year, data = ., cluster = c("gvkey", "ff_30ind_name")) 
  
  reg_contract_intensity_list_fyear[[contract_base]] <- reg_year
  
  assign(x = paste("reg_", contract_base, sep = ""), reg_year) 
} 
## change the name of each did regressions: 
names(reg_contract_intensity_list_fyear) <- str_replace_all(string = names(reg_contract_intensity_list_fyear), pattern = "contract_intensity_", replacement = "base: fyear=")

print("SE clustered at firm level: \n")

[1] "SE clustered at firm level: \n"

Show the code

etable(reg_contract_intensity_list_fyear, cluster = c("gvkey", "ff_30ind_name")) %>%
  knitr::kable(format = "html", caption = "DiD Results with Different Base Year for Contract Intensity") %>% 
  kableExtra::kable_styling(
    font_size = 8,       # Smaller font (default is 16px)
    full_width = FALSE,   # Table width fits content 
    bootstrap_options = c("striped", "condensed")
  )

DiD Results with Different Base Year for Contract Intensity
	base: fye..1997	base: fyear=1998	base: fyear=1999	base: fyear=2000	base: fyear=2001
Dependent Var.:	book_leverage	book_leverage	book_leverage	book_leverage	book_leverage

contract_intensity_1997 x post_2005	0.0139 (0.2941)
contract_intensity_1998 x post_2005		-0.0003 (0.0006)
contract_intensity_1999 x post_2005			-1.781*** (0.2527)
contract_intensity_2000 x post_2005				-0.4244*** (0.1016)
contract_intensity_2001 x post_2005					-0.4708*** (0.0146)
Fixed-Effects:	---------------	----------------	------------------	-------------------	-------------------
gvkey	Yes	Yes	Yes	Yes	Yes
fyear	Yes	Yes	Yes	Yes	Yes
___________________________________	_______________	________________	__________________	___________________	___________________
S.E.: Clustered	by: gvkey & ff_3.	by: gvkey & ff_3.	by: gvkey & ff_3.	by: gvkey & ff_3.	by: gvkey & ff_3.
Observations	17,140	18,782	20,988	22,538	23,739
R2	0.32024	0.52108	0.51044	0.30292	0.26540
Within R2	4.13e-8	1.04e-7	0.00434	0.00112	0.00076

Show the code

print("HC1 robust SE: \n")

[1] "HC1 robust SE: \n"

Show the code

etable(reg_contract_intensity_list_fyear, vcov = "HC1") %>% 
  knitr::kable(format = "html", caption = "DiD Results with Different Base Year for Contract Intensity") %>% 
  kableExtra::kable_styling(
    font_size = 8,       # Smaller font (default is 16px)
    full_width = FALSE,   # Table width fits content 
    bootstrap_options = c("striped", "condensed")
  )

DiD Results with Different Base Year for Contract Intensity
	base: fye..1997	base: fyear=1998	base: fy..1999	base: fyear=2000	base: fyear=2001
Dependent Var.:	book_leverage	book_leverage	book_leverage	book_leverage	book_leverage

contract_intensity_1997 x post_2005	0.0139 (0.2737)
contract_intensity_1998 x post_2005		-0.0003 (0.0005)
contract_intensity_1999 x post_2005			-1.781 (1.296)
contract_intensity_2000 x post_2005				-0.4244. (0.2343)
contract_intensity_2001 x post_2005					-0.4708 (0.2999)
Fixed-Effects:	---------------	----------------	--------------	-----------------	----------------
gvkey	Yes	Yes	Yes	Yes	Yes
fyear	Yes	Yes	Yes	Yes	Yes
___________________________________	_______________	________________	______________	_________________	________________
S.E. type	Heteroske.-rob.	Heterosked.-rob.	Heterosk.-rob.	Heteroskeda.-rob.	Heterosked.-rob.
Observations	17,140	18,782	20,988	22,538	23,739
R2	0.32024	0.52108	0.51044	0.30292	0.26540
Within R2	4.13e-8	1.04e-7	0.00434	0.00112	0.00076

In this case, with HC1 standard errors (same as the default in STATA: robust), only the coefficient using 2000 contract intensity is significant at 10% level, which is very close to the results in Samuel’s slides.

Note 4

The reason for not using two-way clustering is that there are only 6 clusters at the time dimension. Fewer than 10–20 clusters in a dimension can lead to biased standard errors (Cameron & Miller, 2015). With 6 years, two-way clustering would produce unreliable results.

A4. Restrict to Firms w/o Missing Values

Show the code

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classifiaction 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  )

Exclude IPO and Delisting

If we drop IPO and delisting firms from the sample, the results are as follows:

Show the code

data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## drop IPO and delisting firms
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs > n_post & n_post > 0) %>%
  # clean variables
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    ind = as.factor(ff_30ind_name), # industry 
    ind_time = as.factor(paste(ff_30ind_name, dyear, sep = " + ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  ## adding industry*year FE does not change the results: + ind_time 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  fixest::etable()

                                                      .
Dependent Var.:                           book_leverage
                                                       
contract_intensity_2000 x post_2005 -0.4184*** (0.1028)
Fixed-Effects:                      -------------------
gvkey                                               Yes
dyear                                               Yes
___________________________________ ___________________
S.E.: Clustered                       by: gvkey & ff_3.
Observations                                     19,701
R2                                              0.36855
Within R2                                       0.00398
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, indexyear, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2000 & dyear <= 2007) %>% 
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs > n_post & n_post > 0) %>%
  mutate(
    gvkey = as.factor(gvkey) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ i(indexyear, contract_intensity_2000, -1) | gvkey + indexyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  iplot(., 
        xlab = paste("Time to treatment (0 is Year ", base_year, ")", sep = ""),
        main = "Treatment Effect (TWFE): ex. IPO and Delisting", 
        ref.line = FALSE,
        ref.line.par = list(col = "red", lty = 2),
        pt.join = FALSE
        ); axis(1, at = -7:2); abline(v = 0.5, col = "red", lty = 2, lwd = 3)

Show the code

  # fixest::coefplot(sub = "base year: dyear = 2000")

The results are similar to the baseline results in Section 3.1.

Balanced Panel

If we restrict our sample to firms having observations in the full sample between 2002 and 2007, the results are as follows:

Show the code

## baseline: 
cat("\n")
cat("(1) Baseline:") 
cat("\n")
model_a4_a <- data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## drop IPO and delisting firms
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs == 6) %>%
  # clean variables
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    ind = as.factor(ff_30ind_name), # industry 
    ind_time = as.factor(paste(ff_30ind_name, dyear, sep = " + ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  ## adding industry*year FE does not change the results: + ind_time 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 

# model_a4_a %>% fixest::etable()

## baseline + industry*year FE: 
cat("\n")
cat("(2) Baseline + Industry*Year FE:") 
cat("\n")
model_a4_b <- data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## drop IPO and delisting firms
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>% 
  ungroup() %>%
  filter(n_obs == 6) %>%
  # clean variables
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    ind = as.factor(ff_30ind_name), # industry 
    ind_time = as.factor(paste(ff_30ind_name, dyear, sep = " + ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  ## adding industry*year FE does not change the results: + ind_time 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ind^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 

# model_a4_b %>% fixest::etable()

## baseline + constraints on the leverage: 
cat("\n")
cat("(3) Baseline + Leverage Restrictions:") 
cat("\n")
model_a4_c <- data_did %>% 
  # drop values outside the range:
  mutate(
    book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA),
    debt_ebitda = ifelse(test = (debt_ebitda >= 0 & debt_ebitda <= 15), yes = debt_ebitda, no = NA )
  ) %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## drop IPO and delisting firms
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs == 6) %>%
  # clean variables
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    ind = as.factor(ff_30ind_name), # industry 
    ind_time = as.factor(paste(ff_30ind_name, dyear, sep = " + ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  ## adding industry*year FE does not change the results: + ind_time 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 

# model_a4_c %>% fixest::etable()

## baseline + industry*year FE + constraints on the leverage: 
cat("\n")
cat("(4) Baseline + industry*year FE + Leverage Restrictions:") 
cat("\n")
model_a4_d <- data_did %>% 
  # drop values outside the range:
  mutate(
    book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA),
    debt_ebitda = ifelse(test = (debt_ebitda >= 0 & debt_ebitda <= 15), yes = debt_ebitda, no = NA )
  ) %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## drop IPO and delisting firms
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs == 6) %>%
  # clean variables
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    ind = as.factor(ff_30ind_name), # industry 
    ind_time = as.factor(paste(ff_30ind_name, dyear, sep = " + ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  ## adding industry*year FE does not change the results: + ind_time 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ind^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 


cat("\n \n")
etable(list(`(1) Full Sample` = model_a4_a, 
            `(2) Full Sample` = model_a4_b, 
            `(3) Filtered Sample` = model_a4_c, 
            `(4) Filtered Sample` = model_a4_d  
            ), 
       cluster = c("gvkey", "ff_30ind_name"))


## plot the coefficients pre- and post-treatment: 
data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, indexyear, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2000 & dyear <= 2007) %>%
  group_by(gvkey) %>%
  mutate(
    n_obs = sum(dyear >= 2002 & dyear <= 2007), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs == 6) %>%
  mutate(
    gvkey = as.factor(gvkey) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ i(indexyear, contract_intensity_2000, -1) | gvkey + indexyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  iplot(., 
        xlab = paste("Time to treatment (0 is Year ", base_year, ")", sep = ""),
        main = "Treatment Effect (TWFE): ex. observations with missing values", 
        ref.line = FALSE,
        ref.line.par = list(col = "red", lty = 2),
        pt.join = FALSE
        ); axis(1, at = -7:2); abline(v = 0.5, col = "red", lty = 2, lwd = 3)

Show the code

  # fixest::coefplot(sub = "base year: dyear = 2000")


(1) Baseline:

(2) Baseline + Industry*Year FE:

(3) Baseline + Leverage Restrictions:

(4) Baseline + industry*year FE + Leverage Restrictions:

 
                                        (1) Full Sample     (2) Full Sample
Dependent Var.:                           book_leverage       book_leverage
                                                                           
contract_intensity_2000 x post_2005 -0.4190*** (0.1041) -0.4177*** (0.1032)
Fixed-Effects:                      ------------------- -------------------
gvkey                                               Yes                 Yes
dyear                                               Yes                 Yes
ind-dyear                                            No                 Yes
___________________________________ ___________________ ___________________
S.E.: Clustered                       by: gvkey & ff_3.   by: gvkey & ff_3.
Observations                                     17,935              17,935
R2                                              0.39999             0.40320
Within R2                                       0.00511             0.00510

                                      (3) Filtered ..  (4) Filtered S..
Dependent Var.:                         book_leverage     book_leverage
                                                                       
contract_intensity_2000 x post_2005   0.0021 (0.0061)  -0.0010 (0.0070)
Fixed-Effects:                        ---------------  ----------------
gvkey                                             Yes               Yes
dyear                                             Yes               Yes
ind-dyear                                          No               Yes
___________________________________   _______________  ________________
S.E.: Clustered                     by: gvkey & ff_3. by: gvkey & ff_3.
Observations                                   13,668            13,668
R2                                            0.85252           0.85574
Within R2                                     2.97e-5           5.71e-6
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

The results are robust to the sample selection.

A5. Do different industries behave differently?

Show the code

## create leverage variables 
data <- compustat_na %>% 
  # filter(at > 0) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms 
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year
  ) 

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did_withind <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) %>% 
  ## remove some industries: 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) 
  

## did analysis: 
model_a5_average <- data_did_withind %>% 
      select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, ff_30ind_desc, ff_30ind) %>% 
      filter(dyear >= 2002 & dyear <= 2007) %>% 
      mutate(
        gvkey = as.factor(gvkey), 
        dyear = as.factor(dyear), 
        ff_30ind = as.factor(paste(ff_30ind, ff_30ind_desc, sep = ": ")), 
        group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
      ) %>% 
      na.omit() %>% 
      fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind^dyear, data = ., cluster = c("gvkey", "ff_30ind"))

model_a5_average %>% fixest::etable()

                                                       .
Dependent Var.:                            book_leverage
                                                        
contract_intensity_2000 x post_2005  -0.4159*** (0.1016)
Fixed-Effects:                       -------------------
gvkey                                                Yes
dyear                                                Yes
ff_30ind-dyear                                       Yes
___________________________________  ___________________
S.E.: Clustered                     by: gvkey & ff_30ind
Observations                                      22,803
R2                                               0.40083
Within R2                                        0.00334
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

## did analysis for the treatment effects in each industry: 
did_withind_model <- data_did_withind %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, ff_30ind_desc, ff_30ind) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    ff_30ind = as.factor(paste(ff_30ind, ff_30ind_desc, sep = ": ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005:ff_30ind | gvkey + dyear + ff_30ind^dyear, data = ., cluster = c("gvkey"))

did_withind_model %>% fixest::etable()

                                                                                                                           .
Dependent Var.:                                                                                                book_leverage
                                                                                                                            
contract_intensity_2000 x post_2005 x ff_30ind1 x FoodProducts                                                 2.557 (2.321)
contract_intensity_2000 x post_2005 x ff_30ind10 x Textiles                                                  0.2202 (0.6599)
contract_intensity_2000 x post_2005 x ff_30ind11 x ConstructionandConstructionMaterials                       0.3840 (1.948)
contract_intensity_2000 x post_2005 x ff_30ind12 x SteelWorksEtc                                            0.0945* (0.0372)
contract_intensity_2000 x post_2005 x ff_30ind13 x FabricatedProductsandMachinery                            -1.343 (0.8874)
contract_intensity_2000 x post_2005 x ff_30ind14 x ElectricalEquipment                                         1.513 (1.098)
contract_intensity_2000 x post_2005 x ff_30ind15 x AutomobilesandTrucks                                     24.04*** (1.203)
contract_intensity_2000 x post_2005 x ff_30ind16 x Aircraft,ships,andrailroadequipment                      29.67*** (4.386)
contract_intensity_2000 x post_2005 x ff_30ind17 x PreciousMetals,Non-Metallic,andIndustrialMetalMining        63.12 (67.90)
contract_intensity_2000 x post_2005 x ff_30ind18 x Coal                                                     -0.3487 (0.5070)
contract_intensity_2000 x post_2005 x ff_30ind19 x PetroleumandNaturalGas                                -0.4027*** (0.0041)
contract_intensity_2000 x post_2005 x ff_30ind2 x Beer&Liquor                                             0.7163*** (0.1104)
contract_intensity_2000 x post_2005 x ff_30ind21 x Communication                                             0.3623 (0.5689)
contract_intensity_2000 x post_2005 x ff_30ind22 x PersonalandBusinessServices                           -0.3330*** (0.0082)
contract_intensity_2000 x post_2005 x ff_30ind23 x BusinessEquipment                                      -17.50*** (0.6654)
contract_intensity_2000 x post_2005 x ff_30ind24 x BusinessSuppliesandShippingContainers                      -2.812 (1.908)
contract_intensity_2000 x post_2005 x ff_30ind25 x Transportation                                             -1.039 (1.066)
contract_intensity_2000 x post_2005 x ff_30ind26 x Wholesale                                                  -1.617 (4.053)
contract_intensity_2000 x post_2005 x ff_30ind27 x Retail                                                    0.0586 (0.0405)
contract_intensity_2000 x post_2005 x ff_30ind28 x Restaurants,Hotels,Motels                                   2.457 (2.469)
contract_intensity_2000 x post_2005 x ff_30ind29 x Banking,Insurance,RealEstate,Trading                     -0.0304 (0.3842)
contract_intensity_2000 x post_2005 x ff_30ind3 x TobaccoProducts                                       -2.974*** (7.57e-14)
contract_intensity_2000 x post_2005 x ff_30ind30 x EverythingElse                                              1.480 (1.196)
contract_intensity_2000 x post_2005 x ff_30ind4 x Recreation                                                  -4.806 (4.700)
contract_intensity_2000 x post_2005 x ff_30ind5 x PrintingandPublishing                                        171.2 (204.1)
contract_intensity_2000 x post_2005 x ff_30ind6 x ConsumerGoods                                              0.0226 (0.0278)
contract_intensity_2000 x post_2005 x ff_30ind7 x Apparel                                                  -0.0936* (0.0384)
contract_intensity_2000 x post_2005 x ff_30ind8 x Healthcare,MedicalEquipment,PharmaceuticalProducts      -2.003*** (0.0754)
contract_intensity_2000 x post_2005 x ff_30ind9 x Chemicals                                                   3.559. (1.836)
Fixed-Effects:                                                                                          --------------------
gvkey                                                                                                                    Yes
dyear                                                                                                                    Yes
ff_30ind-dyear                                                                                                           Yes
________________________________________                                                                ____________________
S.E.: Clustered                                                                                                    by: gvkey
Observations                                                                                                          22,803
R2                                                                                                                   0.41325
Within R2                                                                                                            0.02400
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

fixest::coefplot(did_withind_model,
                 order = str_extract(string = names(sort(coef(did_withind_model), decreasing = FALSE)), pattern = "(?<=ind).*"), 
                 horiz = TRUE, 
                 drop = "Print", # remove the number 5: Printing & Publishing
                 main = "Treatment Effect Estimates for Fama-French 30 Industries", 
                 sub = "*Printing and Publishing (Industry 5) is excluded. Red dashed line indicates the average treatment effect."
                 ); abline(v = model_a5_average$coefficients, col = 2, lty = 2, lwd = 3)

In this figure, we present the heterogeneous treatment effect of the 2005 policy on the firm leverage across different industries. We include the firm, year and the industry-year fixed effects in the difference-in-difference analysis. We can see that the treatment effect is negative and statistically significant in Business equipment (23), Healthcare (8), Petroleum and Nature Gas (19), Personal and Business Services (22). While the effect is positive and statistically significant in Automobiles and Trucks (15) and Aircraft, ships, and rail road equipment (16).

Note 5

Number of observations in each industry

Show the code

data_did_withind %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>%  
  na.omit() %>%
  group_by(ff_30ind) %>% 
  summarise(n_obs = n(), ff_30ind_name = unique(ff_30ind_desc)) %>%
  ungroup() %>% 
  mutate(`n_per %` = round(n_obs / sum(n_obs) * 100, digits = 2)) %>% 
  arrange(-n_obs) %>% 
  gt() %>% 
  tab_style(
    style = cell_text(size = "x-small"),  # Reduce font size
    locations = cells_column_labels(columns = everything()) # reduce the size of tab headers 
  ) %>% 
  tab_style(
    style = cell_text(size = "x-small"),  # Reduce font size
    locations = cells_body()
  )

ff_30ind	n_obs	ff_30ind_name	n_per %
22	3277	Personal and Business Services	16.86
23	2717	Business Equipment	13.98
8	2570	Healthcare, Medical Equipment, Pharmaceutical Products	13.22
27	1277	Retail	6.57
29	1220	Banking, Insurance, Real Estate, Trading	6.28
30	767	Everything Else	3.95
26	755	Wholesale	3.88
13	706	Fabricated Products and Machinery	3.63
21	624	Communication	3.21
11	581	Construction and Construction Materials	2.99
19	576	Petroleum and Natural Gas	2.96
25	501	Transportation	2.58
28	488	Restaurants, Hotels, Motels	2.51
4	443	Recreation	2.28
9	400	Chemicals	2.06
1	391	Food Products	2.01
7	327	Apparel	1.68
6	271	Consumer Goods	1.39
14	267	Electrical Equipment	1.37
15	251	Automobiles and Trucks	1.29
24	242	Business Supplies and Shipping Containers	1.25
5	214	Printing and Publishing	1.10
16	157	Aircraft, ships, and railroad equipment	0.81
12	155	Steel Works Etc	0.80
10	83	Textiles	0.43
2	71	Beer & Liquor	0.37
17	67	Precious Metals, Non-Metallic, and Industrial Metal Mining	0.34
18	26	Coal	0.13
3	10	Tobacco Products	0.05

This heterogeneous treatment effect is consistent with our hypothesis. The changes in Section 365(d)(4) are mainly on the leases of non-residential real property and thus should have a significant negative impact on the industries with more non-residential real property leases, e.g. industry 8, 23, 22. Its impact on industries with a lot of executory contract unrelated to the real property (e.g. industry 15, 16) should be non-negative.

In the figure, we remove the Industry 5 (Printing & Publishing). Comparing to industries with positive treatment effects, industries with negative treatment effects compose the majority of observations in the sample. This is consistent with our main results in Section 3.1 and in Section 5.4.

A6. Re-run Main Tests with More Filters

In Section 5.2, we find that restricting the dependent variable \in [0,1] significantly impact the treatment effect estimate from the DiD analysis. We therefore re-do main analysis from previous sections.

Dataset without filters (i.e. baseline)

In this setting, we do not put any filters on any variables in the regression. The result is exactly the same as in Section 3.1.

Show the code

rm(list = ls()[! (ls() %in% c("base_year", "compustat_na"))])

## create leverage variables 
data <- compustat_na %>% 
  # filter(at > 0) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    # contract_intensity = (totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year 
  ) 

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) # %>% 
  # # drop values outside the range:
  # mutate(
  #   book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA),
  #   debt_ebitda = ifelse(test = (debt_ebitda >= 0 & debt_ebitda <= 15), yes = debt_ebitda, no = NA )
  # )

etable(
  `(1)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")), 
  `(2)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind_name^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 
)

                                                    (1)                 (2)
Dependent Var.:                           book_leverage       book_leverage
                                                                           
contract_intensity_2000 x post_2005 -0.4184*** (0.1028) -0.4159*** (0.1016)
Fixed-Effects:                      ------------------- -------------------
gvkey                                               Yes                 Yes
dyear                                               Yes                 Yes
ff_30ind_name-dyear                                  No                 Yes
___________________________________ ___________________ ___________________
S.E.: Clustered                       by: gvkey & ff_3.   by: gvkey & ff_3.
Observations                                     22,803              22,803
R2                                              0.39586             0.40083
Within R2                                       0.00336             0.00334
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Dataset with filters on `at`

For at > 0, the results are invariant.

Show the code

rm(list = ls()[! (ls() %in% c("base_year", "compustat_na"))])

## create leverage variables 
data <- compustat_na %>% 
  filter(at > 0) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    # contract_intensity = (totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year 
  ) %>% 
  filter(book_leverage >= 0)

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) # %>% 
  # # drop values outside the range:
  # mutate(
  #   book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA),
  #   debt_ebitda = ifelse(test = (debt_ebitda >= 0 & debt_ebitda <= 15), yes = debt_ebitda, no = NA )
  # )

etable(
  `(1)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")), 
  `(2)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind_name^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 
)

                                                    (1)                 (2)
Dependent Var.:                           book_leverage       book_leverage
                                                                           
contract_intensity_2000 x post_2005 -0.4184*** (0.1028) -0.4159*** (0.1016)
Fixed-Effects:                      ------------------- -------------------
gvkey                                               Yes                 Yes
dyear                                               Yes                 Yes
ff_30ind_name-dyear                                  No                 Yes
___________________________________ ___________________ ___________________
S.E.: Clustered                       by: gvkey & ff_3.   by: gvkey & ff_3.
Observations                                     22,779              22,779
R2                                              0.39586             0.40083
Within R2                                       0.00336             0.00334
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

For at > 0.05, the treatment effects are still negative, but no longer significant.

Show the code

rm(list = ls()[! (ls() %in% c("base_year", "compustat_na"))])

## create leverage variables 
data <- compustat_na %>% 
  filter(at > 0.05) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    # contract_intensity = (totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year 
  ) %>% 
  filter(book_leverage < Inf)

psych::describe(data.frame(book_leverage = data$book_leverage))

              vars      n mean   sd median trimmed  mad   min    max  range
book_leverage    1 103021 0.43 4.95   0.19    0.22 0.27 -0.05 843.86 843.91
                skew kurtosis   se
book_leverage 113.91 15769.32 0.02

Show the code

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) # %>% 
  # # drop values outside the range:
  # mutate(
  #   book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA),
  #   debt_ebitda = ifelse(test = (debt_ebitda >= 0 & debt_ebitda <= 15), yes = debt_ebitda, no = NA )
  # )

etable(
  `(1)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")), 
  `(2)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind_name^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 
)

                                                  (1)               (2)
Dependent Var.:                         book_leverage     book_leverage
                                                                       
contract_intensity_2000 x post_2005  -0.0227 (0.0772)  -0.0134 (0.0913)
Fixed-Effects:                       ----------------  ----------------
gvkey                                             Yes               Yes
dyear                                             Yes               Yes
ff_30ind_name-dyear                                No               Yes
___________________________________  ________________  ________________
S.E.: Clustered                     by: gvkey & ff_3. by: gvkey & ff_3.
Observations                                   22,592            22,592
R2                                            0.51966           0.52440
Within R2                                     2.04e-5           6.44e-6
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Dataset with filters on `book_leverage`

For book_leverage ∈ [0,5], the treatment effect estimates even turn into significantly positive values.

Show the code

rm(list = ls()[! (ls() %in% c("base_year", "compustat_na"))])

## create leverage variables 
data <- compustat_na %>% 
  filter(at > 0.05) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    # contract_intensity = (totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year 
  ) %>% 
  filter(book_leverage < Inf)

psych::describe(data.frame(book_leverage = data$book_leverage))

              vars      n mean   sd median trimmed  mad   min    max  range
book_leverage    1 103021 0.43 4.95   0.19    0.22 0.27 -0.05 843.86 843.91
                skew kurtosis   se
book_leverage 113.91 15769.32 0.02

Show the code

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) %>% 
  # drop values outside the range:
  mutate(
    book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 5), yes = book_leverage, no = NA) 
  )

etable(
  `(1)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")), 
  `(2)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind_name^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 
)

                                                  (1)               (2)
Dependent Var.:                         book_leverage     book_leverage
                                                                       
contract_intensity_2000 x post_2005  0.0358* (0.0141)  0.0382* (0.0158)
Fixed-Effects:                       ----------------  ----------------
gvkey                                             Yes               Yes
dyear                                             Yes               Yes
ff_30ind_name-dyear                                No               Yes
___________________________________  ________________  ________________
S.E.: Clustered                     by: gvkey & ff_3. by: gvkey & ff_3.
Observations                                   22,434            22,434
R2                                            0.70006           0.70265
Within R2                                     0.00093           0.00095
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

For book_leverage ∈ [0,3], the treatment effect estimates are still positive, but not statistically significant anymore.

Show the code

rm(list = ls()[! (ls() %in% c("base_year", "compustat_na"))])

## create leverage variables 
data <- compustat_na %>% 
  filter(at > 0.05) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    # contract_intensity = (totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year 
  ) %>% 
  filter(book_leverage < Inf)

psych::describe(data.frame(book_leverage = data$book_leverage))

              vars      n mean   sd median trimmed  mad   min    max  range
book_leverage    1 103021 0.43 4.95   0.19    0.22 0.27 -0.05 843.86 843.91
                skew kurtosis   se
book_leverage 113.91 15769.32 0.02

Show the code

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) %>% 
  # drop values outside the range:
  mutate(
    book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 3), yes = book_leverage, no = NA) 
  )

etable(
  `(1)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")), 
  `(2)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind_name^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 
)

                                                  (1)               (2)
Dependent Var.:                         book_leverage     book_leverage
                                                                       
contract_intensity_2000 x post_2005   0.0191 (0.0115)   0.0183 (0.0129)
Fixed-Effects:                        ---------------   ---------------
gvkey                                             Yes               Yes
dyear                                             Yes               Yes
ff_30ind_name-dyear                                No               Yes
___________________________________   _______________   _______________
S.E.: Clustered                     by: gvkey & ff_3. by: gvkey & ff_3.
Observations                                   22,296            22,296
R2                                            0.74885           0.75112
Within R2                                     0.00051           0.00042
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

For book_leverage ∈ [0,1], the treatment effect estimates are indifferent from zero.

Show the code

rm(list = ls()[! (ls() %in% c("base_year", "compustat_na"))])

## create leverage variables 
data <- compustat_na %>% 
  filter(at > 0.05) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    # contract_intensity = (totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year 
  ) %>% 
  filter(book_leverage < Inf)

psych::describe(data.frame(book_leverage = data$book_leverage))

              vars      n mean   sd median trimmed  mad   min    max  range
book_leverage    1 103021 0.43 4.95   0.19    0.22 0.27 -0.05 843.86 843.91
                skew kurtosis   se
book_leverage 113.91 15769.32 0.02

Show the code

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## exclude certain industries 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) %>% 
  # drop values outside the range:
  mutate(
    book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA) 
  )

etable(
  `(1)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear, data = ., cluster = c("gvkey", "ff_30ind_name")), 
  `(2)` = data_did %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind_name^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) 
)

                                                  (1)               (2)
Dependent Var.:                         book_leverage     book_leverage
                                                                       
contract_intensity_2000 x post_2005   0.0034 (0.0098)   0.0022 (0.0110)
Fixed-Effects:                        ---------------   ---------------
gvkey                                             Yes               Yes
dyear                                             Yes               Yes
ff_30ind_name-dyear                                No               Yes
___________________________________   _______________   _______________
S.E.: Clustered                     by: gvkey & ff_3. by: gvkey & ff_3.
Observations                                   21,480            21,480
R2                                            0.79683           0.80012
Within R2                                     4.19e-5           1.57e-5
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Treatment Effects across Industries

If we use all observations in the dataset, the results are the same as those in Section 5.5.

Show the code

## create leverage variables 
data <- compustat_na %>% 
  # filter(at > 0) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms 
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year
  ) 

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did_withind <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) %>% 
  ## remove some industries: 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) 
  

## did analysis: 
model_a5_average <- data_did_withind %>% 
      select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, ff_30ind_desc, ff_30ind) %>% 
      filter(dyear >= 2002 & dyear <= 2007) %>% 
      mutate(
        gvkey = as.factor(gvkey), 
        dyear = as.factor(dyear), 
        ff_30ind = as.factor(paste(ff_30ind, ff_30ind_desc, sep = ": ")), 
        group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
      ) %>% 
      na.omit() %>% 
      fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + sic^dyear, data = ., cluster = c("gvkey", "ff_30ind_name"))

model_a5_average %>% fixest::etable()

                                                      .
Dependent Var.:                           book_leverage
                                                       
contract_intensity_2000 x post_2005 -0.4161*** (0.1010)
Fixed-Effects:                      -------------------
gvkey                                               Yes
dyear                                               Yes
sic-dyear                                           Yes
___________________________________ ___________________
S.E.: Clustered                       by: gvkey & ff_3.
Observations                                     22,803
R2                                              0.43833
Within R2                                       0.00346
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

## did analysis for the treatment effects in each industry: 
did_withind_model <- data_did_withind %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, ff_30ind_desc, ff_30ind) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    ff_30ind = as.factor(paste(ff_30ind, ff_30ind_desc, sep = ": ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005:ff_30ind | gvkey + dyear + ff_30ind^dyear, data = ., cluster = c("gvkey"))

did_withind_model %>% fixest::etable()

                                                                                                                           .
Dependent Var.:                                                                                                book_leverage
                                                                                                                            
contract_intensity_2000 x post_2005 x ff_30ind1 x FoodProducts                                                 2.557 (2.321)
contract_intensity_2000 x post_2005 x ff_30ind10 x Textiles                                                  0.2202 (0.6599)
contract_intensity_2000 x post_2005 x ff_30ind11 x ConstructionandConstructionMaterials                       0.3840 (1.948)
contract_intensity_2000 x post_2005 x ff_30ind12 x SteelWorksEtc                                            0.0945* (0.0372)
contract_intensity_2000 x post_2005 x ff_30ind13 x FabricatedProductsandMachinery                            -1.343 (0.8874)
contract_intensity_2000 x post_2005 x ff_30ind14 x ElectricalEquipment                                         1.513 (1.098)
contract_intensity_2000 x post_2005 x ff_30ind15 x AutomobilesandTrucks                                     24.04*** (1.203)
contract_intensity_2000 x post_2005 x ff_30ind16 x Aircraft,ships,andrailroadequipment                      29.67*** (4.386)
contract_intensity_2000 x post_2005 x ff_30ind17 x PreciousMetals,Non-Metallic,andIndustrialMetalMining        63.12 (67.90)
contract_intensity_2000 x post_2005 x ff_30ind18 x Coal                                                     -0.3487 (0.5070)
contract_intensity_2000 x post_2005 x ff_30ind19 x PetroleumandNaturalGas                                -0.4027*** (0.0041)
contract_intensity_2000 x post_2005 x ff_30ind2 x Beer&Liquor                                             0.7163*** (0.1104)
contract_intensity_2000 x post_2005 x ff_30ind21 x Communication                                             0.3623 (0.5689)
contract_intensity_2000 x post_2005 x ff_30ind22 x PersonalandBusinessServices                           -0.3330*** (0.0082)
contract_intensity_2000 x post_2005 x ff_30ind23 x BusinessEquipment                                      -17.50*** (0.6654)
contract_intensity_2000 x post_2005 x ff_30ind24 x BusinessSuppliesandShippingContainers                      -2.812 (1.908)
contract_intensity_2000 x post_2005 x ff_30ind25 x Transportation                                             -1.039 (1.066)
contract_intensity_2000 x post_2005 x ff_30ind26 x Wholesale                                                  -1.617 (4.053)
contract_intensity_2000 x post_2005 x ff_30ind27 x Retail                                                    0.0586 (0.0405)
contract_intensity_2000 x post_2005 x ff_30ind28 x Restaurants,Hotels,Motels                                   2.457 (2.469)
contract_intensity_2000 x post_2005 x ff_30ind29 x Banking,Insurance,RealEstate,Trading                     -0.0304 (0.3842)
contract_intensity_2000 x post_2005 x ff_30ind3 x TobaccoProducts                                       -2.974*** (7.57e-14)
contract_intensity_2000 x post_2005 x ff_30ind30 x EverythingElse                                              1.480 (1.196)
contract_intensity_2000 x post_2005 x ff_30ind4 x Recreation                                                  -4.806 (4.700)
contract_intensity_2000 x post_2005 x ff_30ind5 x PrintingandPublishing                                        171.2 (204.1)
contract_intensity_2000 x post_2005 x ff_30ind6 x ConsumerGoods                                              0.0226 (0.0278)
contract_intensity_2000 x post_2005 x ff_30ind7 x Apparel                                                  -0.0936* (0.0384)
contract_intensity_2000 x post_2005 x ff_30ind8 x Healthcare,MedicalEquipment,PharmaceuticalProducts      -2.003*** (0.0754)
contract_intensity_2000 x post_2005 x ff_30ind9 x Chemicals                                                   3.559. (1.836)
Fixed-Effects:                                                                                          --------------------
gvkey                                                                                                                    Yes
dyear                                                                                                                    Yes
ff_30ind-dyear                                                                                                           Yes
________________________________________                                                                ____________________
S.E.: Clustered                                                                                                    by: gvkey
Observations                                                                                                          22,803
R2                                                                                                                   0.41325
Within R2                                                                                                            0.02400
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

fixest::coefplot(did_withind_model,
                 order = str_extract(string = names(sort(coef(did_withind_model), decreasing = FALSE)), pattern = "(?<=ind).*"), 
                 horiz = TRUE, 
                 drop = "Print", # remove the number 5: Printing & Publishing
                 main = "", 
                 sub = ""
                 ); 
title(main = expression(bold("Treatment Effect Estimates for Fama-French 30 Industries")), adj = 0.2); 
mtext("*Printing and Publishing (Industry 5) is excluded. Red dashed line indicates the average treatment effect.", side = 1, line = 4, adj = 0); 
abline(v = model_a5_average$coefficients, col = 2, lty = 2, lwd = 3)

For the dataset with filters on book_leverage ∈ [0,1], we re-run all the tests.

Show the code

## create leverage variables 
data <- compustat_na %>% 
  # filter(at > 0) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms 
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year
  ) 

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did_withind2 <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) %>% 
  ## remove some industries: 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  # drop values outside the range:
  mutate(
    book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA) 
  ) # %>% 
  # ## keep only firms with all observations over the full period. 
  # group_by(gvkey) %>% 
  # mutate(
  #   n_obs = n(), # number of observations for each firm
  #   n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  # ) %>%
  # ungroup() %>%
  # filter(n_obs > n_post & n_post > 0)

## did analysis: 
model_a5_average2 <- data_did_withind2 %>% 
      select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, ff_30ind_desc, ff_30ind) %>% 
      filter(dyear >= 2002 & dyear <= 2007) %>% 
      mutate(
        gvkey = as.factor(gvkey), 
        dyear = as.factor(dyear), 
        ff_30ind = as.factor(paste(ff_30ind, ff_30ind_desc, sep = ": ")), 
        group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
      ) %>% 
      na.omit() %>% 
      fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind^dyear, data = ., cluster = c("gvkey", "ff_30ind_name"))

model_a5_average2 %>% fixest::etable()

                                                    .
Dependent Var.:                         book_leverage
                                                     
contract_intensity_2000 x post_2005   0.0028 (0.0019)
Fixed-Effects:                        ---------------
gvkey                                             Yes
dyear                                             Yes
ff_30ind-dyear                                    Yes
___________________________________   _______________
S.E.: Clustered                     by: gvkey & ff_3.
Observations                                   21,599
R2                                            0.79807
Within R2                                     0.00017
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

## did analysis for the treatment effects in each industry: 
did_withind_model2 <- data_did_withind2 %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, ff_30ind_desc, ff_30ind) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  mutate(
    gvkey = as.factor(gvkey), 
    dyear = as.factor(dyear), 
    ff_30ind = as.factor(paste(ff_30ind, ff_30ind_desc, sep = ": ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005:ff_30ind | gvkey + dyear + ff_30ind^dyear, data = ., cluster = c("gvkey"))

did_withind_model2 %>% fixest::etable()

                                                                                                                           .
Dependent Var.:                                                                                                book_leverage
                                                                                                                            
contract_intensity_2000 x post_2005 x ff_30ind1 x FoodProducts                                              -0.0089 (0.0829)
contract_intensity_2000 x post_2005 x ff_30ind10 x Textiles                                                  0.1242 (0.7233)
contract_intensity_2000 x post_2005 x ff_30ind11 x ConstructionandConstructionMaterials                     -0.0240 (0.0261)
contract_intensity_2000 x post_2005 x ff_30ind12 x SteelWorksEtc                                            0.0686* (0.0278)
contract_intensity_2000 x post_2005 x ff_30ind13 x FabricatedProductsandMachinery                           -0.1308 (0.1273)
contract_intensity_2000 x post_2005 x ff_30ind14 x ElectricalEquipment                                     0.3717** (0.1228)
contract_intensity_2000 x post_2005 x ff_30ind15 x AutomobilesandTrucks                                     -0.3024 (0.2737)
contract_intensity_2000 x post_2005 x ff_30ind16 x Aircraft,ships,andrailroadequipment                      -0.4004 (0.2792)
contract_intensity_2000 x post_2005 x ff_30ind17 x PreciousMetals,Non-Metallic,andIndustrialMetalMining     -0.2696 (0.3718)
contract_intensity_2000 x post_2005 x ff_30ind18 x Coal                                                     -0.3487 (0.5071)
contract_intensity_2000 x post_2005 x ff_30ind19 x PetroleumandNaturalGas                                 0.0263*** (0.0027)
contract_intensity_2000 x post_2005 x ff_30ind2 x Beer&Liquor                                             0.7163*** (0.1104)
contract_intensity_2000 x post_2005 x ff_30ind21 x Communication                                            -0.0166 (0.0234)
contract_intensity_2000 x post_2005 x ff_30ind22 x PersonalandBusinessServices                              -0.0280 (0.0272)
contract_intensity_2000 x post_2005 x ff_30ind23 x BusinessEquipment                                      0.0060*** (0.0015)
contract_intensity_2000 x post_2005 x ff_30ind24 x BusinessSuppliesandShippingContainers                    -0.0471 (0.1226)
contract_intensity_2000 x post_2005 x ff_30ind25 x Transportation                                           -0.0082 (0.0177)
contract_intensity_2000 x post_2005 x ff_30ind26 x Wholesale                                                -0.0394 (0.0467)
contract_intensity_2000 x post_2005 x ff_30ind27 x Retail                                                    0.0106 (0.0129)
contract_intensity_2000 x post_2005 x ff_30ind28 x Restaurants,Hotels,Motels                                -0.0222 (0.0303)
contract_intensity_2000 x post_2005 x ff_30ind29 x Banking,Insurance,RealEstate,Trading                      0.0416 (0.0462)
contract_intensity_2000 x post_2005 x ff_30ind3 x TobaccoProducts                                       -2.974*** (2.05e-16)
contract_intensity_2000 x post_2005 x ff_30ind30 x EverythingElse                                            0.0072 (0.0307)
contract_intensity_2000 x post_2005 x ff_30ind4 x Recreation                                                0.1603* (0.0749)
contract_intensity_2000 x post_2005 x ff_30ind5 x PrintingandPublishing                                      0.1978 (0.1655)
contract_intensity_2000 x post_2005 x ff_30ind6 x ConsumerGoods                                              0.0127 (0.0295)
contract_intensity_2000 x post_2005 x ff_30ind7 x Apparel                                                  -0.0755* (0.0317)
contract_intensity_2000 x post_2005 x ff_30ind8 x Healthcare,MedicalEquipment,PharmaceuticalProducts         0.0012 (0.0012)
contract_intensity_2000 x post_2005 x ff_30ind9 x Chemicals                                                -0.3381* (0.1453)
Fixed-Effects:                                                                                          --------------------
gvkey                                                                                                                    Yes
dyear                                                                                                                    Yes
ff_30ind-dyear                                                                                                           Yes
________________________________________                                                                ____________________
S.E.: Clustered                                                                                                    by: gvkey
Observations                                                                                                          21,599
R2                                                                                                                   0.79900
Within R2                                                                                                            0.00481
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

fixest::coefplot(did_withind_model2,
                 order = str_extract(string = names(sort(coef(did_withind_model2), decreasing = FALSE)), pattern = "(?<=ind).*"), 
                 horiz = TRUE, 
                 drop = "Print", # remove the number 5: Printing & Publishing
                 main = "", 
                 sub = ""
                 ); 
title(main = expression(bold("Treatment Effect Estimates for Fama-French 30 Industries (Book leverage" %in% "[0,1])")), adj = 0.2); 
mtext("*Printing and Publishing (Industry 5) is excluded. Red dashed line indicates the average treatment effect.", side = 1, line = 4, adj = 0); 
abline(v = model_a5_average2$coefficients, col = 2, lty = 2, lwd = 3)

The industry-by-industry estimates show that observations for ff_30ind == 8, 22, 23 (Healthcare, Medical Equipment, Pharmaceutical Products; Personal and Business Services; Business Equipment) are significantly reduced and this contributes to the change in the average treatment effect estimate in Section 5.2.

We also compare the treatment effects from the two samples:

as the coefficient estimates for industry 5 are too large, they are excluded in the following graph.

Show the code

fixest::coefplot(list(did_withind_model, did_withind_model2), 
                 order = paste("ind", sort(as.numeric(str_extract(names(coef(did_withind_model2)), pattern = "(?<=ind)[^:]+"))), ":", sep = ""), 
                 horiz = TRUE, 
                 drop = "Print", # remove the number 5: Printing & Publishing
                 main = "", 
                 sub = "", sep = 0.3
                 ); 
title(main = expression(bold("Treatment Effect Estimates for Fama-French 30 Industries")), adj = 0.2); 
mtext("*Printing and Publishing (Industry 5) is excluded. Red dashed line indicates the average treatment effect.", side = 1, line = 4, adj = 0); 
legend("bottomleft", bty = "n", col = c(1, 2), pch = c(20, 17), 
       legend = c("Full Sample", expression("Book leverage" %in% "[0,1]")))

We also check the impact of this filter, book_leverage ∈ [0,1], on the sample. The left table below gives the number of firm-year observations for each industry in the full sample and the right one gives the number of observations for each industry in the filter sample. We can see a decrease in the number of observations in the three industries, 8, 22 and 23, after applying the filter. These three industries happen to be the ones contribute most to the average treatment effect.

Show the code

tbl_count1 <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## remove some industries: 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>%  
  na.omit() %>%
  group_by(ff_30ind) %>% 
  summarise(n_obs = n(), ff_30ind_name = unique(ff_30ind_desc)) %>%
  ungroup() %>% 
  mutate(`n_per %` = round(n_obs / sum(n_obs) * 100, digits = 2)) %>% 
  arrange(-n_obs) %>% 
  gt() %>% 
  tab_spanner(
    label = "(1) Full Sample",
    columns = c(n_obs, `n_per %`)
  ) %>% 
  tab_style(
    style = cell_text(size = "x-small"),  # Reduce font size
    locations = cells_column_labels(columns = everything()) # reduce the size of tab headers 
  ) %>% 
  tab_style(
    style = cell_text(size = "x-small"),  # Reduce font size
    locations = cells_body()
  ) 
tbl_count1
tbl_count2 <- data_did_withind2 %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>%  
  na.omit() %>%
  group_by(ff_30ind) %>% 
  summarise(n_obs = n(), ff_30ind_name = unique(ff_30ind_desc)) %>%
  ungroup() %>% 
  mutate(`n_per %` = round(n_obs / sum(n_obs) * 100, digits = 2)) %>% 
  arrange(-n_obs) %>% 
  gt() %>% 
  tab_spanner(
    label = html("(2) book leverage &isin; [0,1]"), 
    columns = c(n_obs, `n_per %`)
  ) %>% 
  tab_style(
    style = cell_text(size = "x-small"),  # Reduce font size
    locations = cells_column_labels(columns = everything()) # reduce the size of tab headers 
  ) %>% 
  tab_style(
    style = cell_text(size = "x-small"),  # Reduce font size
    locations = cells_body()
  ) 
tbl_count2

ff_30ind	(1) Full Sample		ff_30ind_name
ff_30ind	n_obs	n_per %	ff_30ind_name
22	3277	16.86	Personal and Business Services
23	2717	13.98	Business Equipment
8	2570	13.22	Healthcare, Medical Equipment, Pharmaceutical Products
27	1277	6.57	Retail
29	1220	6.28	Banking, Insurance, Real Estate, Trading
30	767	3.95	Everything Else
26	755	3.88	Wholesale
13	706	3.63	Fabricated Products and Machinery
21	624	3.21	Communication
11	581	2.99	Construction and Construction Materials
19	576	2.96	Petroleum and Natural Gas
25	501	2.58	Transportation
28	488	2.51	Restaurants, Hotels, Motels
4	443	2.28	Recreation
9	400	2.06	Chemicals
1	391	2.01	Food Products
7	327	1.68	Apparel
6	271	1.39	Consumer Goods
14	267	1.37	Electrical Equipment
15	251	1.29	Automobiles and Trucks
24	242	1.25	Business Supplies and Shipping Containers
5	214	1.10	Printing and Publishing
16	157	0.81	Aircraft, ships, and railroad equipment
12	155	0.80	Steel Works Etc
10	83	0.43	Textiles
2	71	0.37	Beer & Liquor
17	67	0.34	Precious Metals, Non-Metallic, and Industrial Metal Mining
18	26	0.13	Coal
3	10	0.05	Tobacco Products

ff_30ind	(2) book leverage ∈ [0,1]		ff_30ind_name
ff_30ind	n_obs	n_per %	ff_30ind_name
22	3131	16.86	Personal and Business Services
23	2615	14.08	Business Equipment
8	2439	13.13	Healthcare, Medical Equipment, Pharmaceutical Products
27	1251	6.74	Retail
29	1183	6.37	Banking, Insurance, Real Estate, Trading
26	737	3.97	Wholesale
30	682	3.67	Everything Else
13	669	3.60	Fabricated Products and Machinery
11	563	3.03	Construction and Construction Materials
19	560	3.02	Petroleum and Natural Gas
21	547	2.95	Communication
25	495	2.67	Transportation
28	469	2.53	Restaurants, Hotels, Motels
4	419	2.26	Recreation
1	384	2.07	Food Products
9	371	2.00	Chemicals
7	322	1.73	Apparel
14	260	1.40	Electrical Equipment
6	249	1.34	Consumer Goods
15	241	1.30	Automobiles and Trucks
24	232	1.25	Business Supplies and Shipping Containers
5	201	1.08	Printing and Publishing
12	153	0.82	Steel Works Etc
16	146	0.79	Aircraft, ships, and railroad equipment
10	79	0.43	Textiles
2	71	0.38	Beer & Liquor
17	66	0.36	Precious Metals, Non-Metallic, and Industrial Metal Mining
18	26	0.14	Coal
3	10	0.05	Tobacco Products

To further validate our conjecture that the filter affecting different industries unequally, we remove all firms which contain book_leverage outside the [0,1] range. We re-do the main analysis and the industry level analysis on a balanced panel:

Show the code

## create leverage variables 
data <- compustat_na %>% 
  # filter(at > 0) %>% 
  filter(indfmt == "INDL") %>% ## remove all financial firms 
  mutate(
    ## leverage variables: 
    book_leverage = (dlc + dltt) / at, 
    debt_ebitda = (dlc + dltt) / ebitda, 
    ## treatment intensity: 
    totalrent = mrc1 + mrc2 + mrc3 + mrc4 + mrc5 + mrct, # total rent w/o discount
    # contract_intensity = (dclo + xrent * 3) / at, # executory contract intensity
    contract_intensity = (dclo + totalrent) / at, # executory contract intensity
    ## other variables: 
    dyear = as.integer(substr(datadate, start = 1, stop = 4)) # data year
  ) 

data_firm <- data %>% 
  filter(dyear == 2000) %>% # use `dyear` may create multiple observations for the same firm. E.g. filter(gvkey == 23535, dyear == 2005) 
  select(gvkey, contract_intensity_2000 = contract_intensity) %>% 
  group_by(gvkey) %>% 
  summarise(contract_intensity_2000 = mean(contract_intensity_2000, na.rm = TRUE) ) %>% 
  ungroup() 

FF_30ind <- as_tibble(read.csv(file = "C:/Users/13613/OneDrive - Handelshögskolan i Stockholm/Projects_2024/BeckerJosephsonXu_2025/AFA_meeting_response/Data/ff_30ind.csv"))

data_did_withind3 <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) %>% 
  ## remove some industries: 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    ff_30ind != 5, # exclude (number 5) 
    fic == "USA", 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  # drop values outside the range:
  mutate(
    book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA) 
  ) 
  
## did analysis: 
data_did_withind3 %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## keep only firms with all observations over the full period.
  na.omit() %>% 
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs >= 6) %>% 
  ## re-clean  
  mutate(
    gvkey = as.factor(gvkey) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  fixest::etable()

                                                    .
Dependent Var.:                         book_leverage
                                                     
contract_intensity_2000 x post_2005  -0.0037 (0.0107)
Fixed-Effects:                       ----------------
gvkey                                             Yes
dyear                                             Yes
ff_30ind-dyear                                    Yes
___________________________________  ________________
S.E.: Clustered                     by: gvkey & ff_3.
Observations                                   15,449
R2                                            0.78062
Within R2                                      3.8e-5
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

## did analysis for the treatment effects in each industry: 
did_withind_model3 <- data_did_withind3 %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, ff_30ind_desc, ff_30ind) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## keep only firms with all observations over the full period.
  na.omit() %>% 
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs >= 6) %>% 
  ## 
  mutate(
    gvkey = as.factor(gvkey), 
    ff_30ind = as.factor(paste(ff_30ind, ff_30ind_desc, sep = ": ")), 
    group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005:ff_30ind | gvkey + dyear + ff_30ind^dyear, data = ., cluster = c("gvkey"))

model_a5_average3 <- data_did_withind3 %>% 
      select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind_name, ff_30ind_desc, ff_30ind) %>% 
      filter(dyear >= 2002 & dyear <= 2007) %>% 
      ## keep only firms with all observations over the full period.
      na.omit() %>% 
      group_by(gvkey) %>%
      mutate(
        n_obs = n(), # number of observations for each firm
        n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
      ) %>%
      ungroup() %>%
      filter(n_obs >= 6) %>% 
      ## 
      mutate(
        gvkey = as.factor(gvkey), 
        ff_30ind = as.factor(paste(ff_30ind, ff_30ind_desc, sep = ": ")), 
        group = contract_intensity_2000 > median(contract_intensity_2000, na.rm = TRUE) 
      ) %>% 
      na.omit() %>% 
      fixest::feols(fml = book_leverage ~ contract_intensity_2000:post_2005 | gvkey + dyear + sic^dyear, data = ., cluster = c("gvkey", "ff_30ind_name"))

did_withind_model3 %>% fixest::etable()

                                                                                                                           .
Dependent Var.:                                                                                                book_leverage
                                                                                                                            
contract_intensity_2000 x post_2005 x ff_30ind1 x FoodProducts                                               0.0604 (0.0738)
contract_intensity_2000 x post_2005 x ff_30ind10 x Textiles                                                  0.2196 (0.7334)
contract_intensity_2000 x post_2005 x ff_30ind11 x ConstructionandConstructionMaterials                     -0.0205 (0.0245)
contract_intensity_2000 x post_2005 x ff_30ind12 x SteelWorksEtc                                             0.0470 (0.0302)
contract_intensity_2000 x post_2005 x ff_30ind13 x FabricatedProductsandMachinery                           -0.0556 (0.1267)
contract_intensity_2000 x post_2005 x ff_30ind14 x ElectricalEquipment                                    0.4162*** (0.1213)
contract_intensity_2000 x post_2005 x ff_30ind15 x AutomobilesandTrucks                                     -0.2071 (0.2629)
contract_intensity_2000 x post_2005 x ff_30ind16 x Aircraft,ships,andrailroadequipment                      -0.3590 (0.2775)
contract_intensity_2000 x post_2005 x ff_30ind17 x PreciousMetals,Non-Metallic,andIndustrialMetalMining      -0.3417 (1.645)
contract_intensity_2000 x post_2005 x ff_30ind18 x Coal                                                     -0.3487 (0.5080)
contract_intensity_2000 x post_2005 x ff_30ind19 x PetroleumandNaturalGas                                 0.2578*** (0.0543)
contract_intensity_2000 x post_2005 x ff_30ind2 x Beer&Liquor                                             0.7163*** (0.1106)
contract_intensity_2000 x post_2005 x ff_30ind21 x Communication                                            -0.0220 (0.0208)
contract_intensity_2000 x post_2005 x ff_30ind22 x PersonalandBusinessServices                              -0.0351 (0.0316)
contract_intensity_2000 x post_2005 x ff_30ind23 x BusinessEquipment                                         0.0206 (0.0317)
contract_intensity_2000 x post_2005 x ff_30ind24 x BusinessSuppliesandShippingContainers                    -0.1955 (0.2194)
contract_intensity_2000 x post_2005 x ff_30ind25 x Transportation                                            0.0172 (0.0123)
contract_intensity_2000 x post_2005 x ff_30ind26 x Wholesale                                                -0.0344 (0.0411)
contract_intensity_2000 x post_2005 x ff_30ind27 x Retail                                                    0.0060 (0.0130)
contract_intensity_2000 x post_2005 x ff_30ind28 x Restaurants,Hotels,Motels                                -0.0194 (0.0367)
contract_intensity_2000 x post_2005 x ff_30ind29 x Banking,Insurance,RealEstate,Trading                     -0.0034 (0.0351)
contract_intensity_2000 x post_2005 x ff_30ind3 x TobaccoProducts                                       -2.974*** (4.44e-16)
contract_intensity_2000 x post_2005 x ff_30ind30 x EverythingElse                                            0.0336 (0.0290)
contract_intensity_2000 x post_2005 x ff_30ind4 x Recreation                                              0.2530*** (0.0512)
contract_intensity_2000 x post_2005 x ff_30ind6 x ConsumerGoods                                              0.0175 (0.0298)
contract_intensity_2000 x post_2005 x ff_30ind7 x Apparel                                                  -0.0687* (0.0318)
contract_intensity_2000 x post_2005 x ff_30ind8 x Healthcare,MedicalEquipment,PharmaceuticalProducts      -0.0624** (0.0217)
contract_intensity_2000 x post_2005 x ff_30ind9 x Chemicals                                               -0.3987** (0.1323)
Fixed-Effects:                                                                                          --------------------
gvkey                                                                                                                    Yes
dyear                                                                                                                    Yes
ff_30ind-dyear                                                                                                           Yes
________________________________________                                                                ____________________
S.E.: Clustered                                                                                                    by: gvkey
Observations                                                                                                          15,449
R2                                                                                                                   0.78236
Within R2                                                                                                            0.00795
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Show the code

fixest::coefplot(did_withind_model3,
                 order = str_extract(string = names(sort(coef(did_withind_model3), decreasing = FALSE)), pattern = "(?<=ind).*"), 
                 horiz = TRUE, 
                 drop = "Print", # remove the number 5: Printing & Publishing
                 main = "", 
                 sub = ""
                 ); 
title(main = expression(bold("Treatment Effect Estimates for Fama-French 30 Industries (Book leverage" %in% "[0,1] & Balanced Panel)")), adj = 0.2); 
mtext("*Printing and Publishing (Industry 5) is excluded. Red dashed line indicates the average treatment effect.", side = 1, line = 4, adj = 0); 
abline(v = model_a5_average3$coefficients, col = 2, lty = 2, lwd = 3)

The average treatment effect estimate is negative but insignificant in the filtered sample. In particular, we have seen a large change in the treatment effect estimate in industries 8 and 23. As these industries consist of most of the observations in our sample (over 25%), this significantly reduce the magnitude of the average treatment effect estimate in the sample and make it insignificant.

In this subsection, let’s look into the dynamics of those excluded observations:

Show the code

## data prep: 
data_did_withind3 <- data %>% 
  ## merge the grouping variable 
  left_join(y = data_firm, by = join_by(gvkey) ) %>% 
  ## merge the FF 30 industry classification 
  left_join(y = FF_30ind, by = join_by(sic)) %>% 
  ## pre- & post- 
  mutate(
    indexyear = (dyear - base_year), # year index with 2005 as base 0.
    post_2005 = ifelse(dyear > base_year, yes = 1, no = 0)  # whether it is after the 2005 reform 
  ) %>% 
  ## remove some industries: 
  filter(
    ff_30ind != 20, # exclude Utiltiy (number 20) 
    ff_30ind != 5, # exclude (number 5) 
    !(sic >= 6000 & sic <= 6199) # exclude banks under Banking, Insurance, Real Estate, Trading (number 29)
  ) %>% 
  # drop values outside the range:
  mutate(
    book_leverage_filter = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA) 
  ) 

data_did_withind3 %>% 
  filter(is.na(book_leverage_filter)) %>% 
  filter(dyear >= 2000 & dyear <= 2007 & !is.na(book_leverage) & !is.na(contract_intensity_2000)) %>% 
  select(conm, tic, 
         # tic, cik, fic, 
         dyear, 
         # fyear, datadate,
         ci_2000 = contract_intensity_2000, book_leverage, 
         at, dlc, dltt, ff_30ind_name)

# A tibble: 1,799 × 9
   conm       tic   dyear ci_2000 book_leverage     at   dlc  dltt ff_30ind_name
   <chr>      <chr> <int>   <dbl>         <dbl>  <dbl> <dbl> <dbl> <chr>        
 1 WORLDS INC WDDD   2000  0.254           1.06  1.96  0.66   1.42 Servs        
 2 WORLDS INC WDDD   2001  0.254         364.    0.006 2.18   0    Servs        
 3 WORLDS INC WDDD   2002  0.254         170.    0.013 2.22   0    Servs        
 4 WORLDS INC WDDD   2003  0.254        1116.    0.002 2.23   0    Servs        
 5 WORLDS INC WDDD   2004  0.254        1037     0.002 2.07   0    Servs        
 6 WORLDS INC WDDD   2005  0.254         613.    0.003 1.84   0    Servs        
 7 WORLDS INC WDDD   2006  0.254         920.    0.002 1.84   0    Servs        
 8 WORLDS INC WDDD   2007  0.254           2.24  0.346 0.774  0    Servs        
 9 B-FAST CO… BFTC   2000  0.0149          1.38 11.6   0.376 15.7  Other        
10 B-FAST CO… BFTC   2001  0.0149          1.53 11.3   0.935 16.4  Other        
# ℹ 1,789 more rows

Show the code

## boxplot 
data_did_withind3 %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind, ff_30ind_name, book_leverage_filter) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## keep only firms with all observations over the full period.
  # na.omit() %>% 
  group_by(gvkey) %>%
  mutate(
    n_obs = sum(!is.na(book_leverage_filter)), # number of non-NA observations for each firm 
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs < 6) %>% 
  mutate(
    dyear = as.factor(dyear)
  ) %>% 
  ggplot(data = ., aes(x = dyear, y = book_leverage)) +
  ## box plot
  geom_boxplot(notch = FALSE, outlier.shape = 4) +
  ## add dot plot
  geom_dotplot(binaxis = "y",
               binwidth = 1/50,
               position = position_jitter(width = 0, height = 0.1, seed = 123),
               stackdir = "center",
               dotsize = 0.3)

Warning: Removed 6911 rows containing non-finite outside the scale range
(`stat_boxplot()`).

Warning: Removed 6762 rows containing missing values or values outside the scale range
(`stat_bindot()`).

Show the code

## violin 
data_did_withind3 %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## keep only firms with all observations over the full period.
  na.omit() %>% 
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs < 6) %>% 
  mutate(
    dyear = as.factor(dyear)
  ) %>% 
  ggplot(data = ., aes(x = dyear, y = book_leverage)) + 
  geom_violin()

Warning: Removed 28 rows containing non-finite outside the scale range
(`stat_ydensity()`).

Show the code

## density 
data_did_withind3 %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind, ff_30ind_name) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## keep only firms with all observations over the full period.
  na.omit() %>% 
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  filter(n_obs < 6) %>% 
  mutate(
    dyear = as.factor(dyear)
  ) %>% 
  ggplot(., aes(x = book_leverage, color = factor(dyear)) ) + 
  geom_density(alpha = 0.3) + 
  labs(x = "Book Leverage", y = "Density", title = "Density Plot of Book Leverage by Year") +
  theme_minimal() +
  theme(legend.title = element_blank())

Warning: Removed 28 rows containing non-finite outside the scale range
(`stat_density()`).

Show the code

cat("More tests: use `debt_ebitda` ratio as the dependent variable ")

More tests: use `debt_ebitda` ratio as the dependent variable

Show the code

## did analysis: 
data_did_withind %>% 
  select(gvkey, dyear, contract_intensity_2000, post_2005, book_leverage, debt_ebitda, sic, ff_30ind, ff_30ind_name) %>% 
  mutate(
    book_leverage = ifelse(test = (book_leverage >= 0 & book_leverage <= 1), yes = book_leverage, no = NA),
    debt_ebitda = ifelse(test = (debt_ebitda >= 0 & debt_ebitda <= 15), yes = debt_ebitda, no = NA )
  ) %>% 
  filter(dyear >= 2002 & dyear <= 2007) %>% 
  ## keep only firms with all observations over the full period.
  na.omit() %>% 
  group_by(gvkey) %>%
  mutate(
    n_obs = n(), # number of observations for each firm
    n_post = sum(post_2005, na.rm = TRUE) # number of obs after 2005
  ) %>%
  ungroup() %>%
  # filter(n_obs == 6) %>% 
  ## re-clean  
  mutate(
    gvkey = as.factor(gvkey) 
  ) %>% 
  na.omit() %>% 
  fixest::feols(fml = debt_ebitda ~ contract_intensity_2000:post_2005 | gvkey + dyear + ff_30ind^dyear, data = ., cluster = c("gvkey", "ff_30ind_name")) %>% 
  fixest::etable()

                                                    .
Dependent Var.:                           debt_ebitda
                                                     
contract_intensity_2000 x post_2005   0.0086 (0.0794)
Fixed-Effects:                        ---------------
gvkey                                             Yes
dyear                                             Yes
ff_30ind-dyear                                    Yes
___________________________________   _______________
S.E.: Clustered                     by: gvkey & ff_3.
Observations                                   16,865
R2                                            0.74609
Within R2                                      1.3e-6
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

There seems to be a shift.

Important 2

Use TNIC or other word classification techniques to identify firms more likely to have a lot of non-residential lease contracts.

S1. Download and Load the Raw Dataset

S2. Data Cleaning and Prep

S3. Difference-in-Difference Examination

s3.1. try-and-error with calender year dyear

s3.2. try-and-error with calender year fyear

S4. [Deprecated] Industry Level DiD

Appendix

A1. Results using different measures of “Net Rental Expenses”

A1.1. try-and-error with calender year dyear

A1.2. try-and-error with calender year fyear

A2. Results using data filters in the main paper

A3. Sensitivity Analysis of Choosing 2000

A4. Restrict to Firms w/o Missing Values

Exclude IPO and Delisting

Balanced Panel

A5. Do different industries behave differently?

A6. Re-run Main Tests with More Filters

Dataset without filters (i.e. baseline)

Dataset with filters on at

Dataset with filters on book_leverage

Treatment Effects across Industries

EndNotes

s3.1. try-and-error with calender year `dyear`

s3.2. try-and-error with calender year `fyear`

A1.1. try-and-error with calender year `dyear`

A1.2. try-and-error with calender year `fyear`

Dataset with filters on `at`

Dataset with filters on `book_leverage`