Table 6 in Fafchamps et al. (2014), “Microenterprise growth and the flypaper effect: Evidence from a randomized experiment in Ghana”

To read the created .tex file, the following \(\LaTeX\) packages are required:

• float
• pdflscape
• booktabs
• multicol
• makecell

The means of outcome variables are added to the table, which do not appear in the original table in the paper.

# Install and load packages ---------------
packages <- c(
  "tidyverse",
  "haven",
  "lfe",
  "stargazer",
  "broom",
  "kableExtra",
  "XML",
  "magrittr"
)

pacman::p_load(packages, character.only = TRUE, install = TRUE)

# Load an example dataset ---------------
data <- read_dta("data/ReplicationDataGhanaJDE.dta")

Data cleaning

# variable labels
lhs_var_list <- c(
  "weeklyfood", "clothingshoes", "healthedn", 
  "expend_ceremonies_3months", "expend_total", "expend_logtotal"
  )

# data cleaning
df_table6 <- data %>%
  filter(is.na(trimgroup)) %>%
  group_by(groupnum) %>%
  mutate(
    amounttransferout = ifelse(is.na(madetransfersout), NA, amounttransferout),
    expend_education_3months = ifelse(expend_education_3months > 3000, NA, expend_education_3months),
    expend_total = ifelse(expend_total > 20000, NA, expend_total),
    expend_logtotal = ifelse(expend_logtotal > 10, NA, expend_logtotal),
    expend_fooddrink_outside_week = ifelse(
      is.na(expend_fooddrink_outside_week) & !is.na(expend_fooddrink_home_week), 
      0, expend_fooddrink_outside_week
      ),
    weeklyfood = expend_fooddrink_home_week + expend_fooddrink_outside_week,
    clothingshoes = expend_clothing_3months + expend_footwear_3months,
    expend_ceremonies_3months = ifelse(
      is.na(expend_health_3months) & is.na(expend_education_3months) & is.na(weeklyfood), 
      NA,
      ifelse(is.na(expend_ceremonies_3months), 0, expend_ceremonies_3months)
      ),
    healthedn = expend_health_3months + expend_education_3months,
    across(
      .cols = c("atreatcash", "atreatequip", "wave2", "wave3", "wave4", "wave5", "wave6"),
      .fns = function(x) x * mlowgroup,
      .names = "{col}_mlowgroup"
      ),
    across(
      .cols = c("atreatcash", "atreatequip", "wave2", "wave3", "wave4", "wave5", "wave6"),
      .fns = function(x) x * highgroup,
      .names = "{col}_highgroup"
      )
    ) %>%
  ungroup() %>% 
  mutate(
    across(
      .cols = c(all_of(lhs_var_list), "totalK"),
      .fns = function(x) ifelse(
        x > quantile(x, .995, na.rm = TRUE, type = 2), 
        quantile(x, .995, na.rm = TRUE, type = 2), 
        x
        ),
      .names = "{col}_trunc"
      )
  )

Regression

# list of names of truncated variables
lhs_var_list_trunc <- c(
  "totalK", "totalK_trunc", "madetransfersout", "amounttransferout", 
  sapply(lhs_var_list, function(x) paste0(x, "_trunc"))
  ) %>%
  set_names(NULL)

# function to run regressions with fixed effects 
felm_table6 <- function(outcome, fe, rhs_var, control_var, iv, cluster, data) {
  as.formula(
    paste(
      paste(outcome, paste(c(rhs_var, control_var), collapse = " + "), sep = " ~ "),
      fe, iv, cluster, sep = " | "
    )
  ) %>%
  felm(
    data = data
  )
}

# function to create a panel of regression results
create_table6_sub <- function(panel) {
  
  fe_1 <- "groupnum"
  fe_2 <- "sheno"
  fe_list <- c(rep(fe_2, 2), rep(fe_1, 2), rep(fe_2, 6))
  iv <- "0"
  cluster <- "sheno"

  if (panel == "A") {
    rhs_var <- c("atreatcashfemale", "atreatequipfemale", "atreatcashmale", "atreatequipmale")
    control_var_1 <- colnames(df_table6)[grepl("^wave\\d(_female)?$", colnames(df_table6))]
    control_var_2 <- colnames(df_table6)[grepl("^wave(5|6)(_female)?$", colnames(df_table6))]
    control_var_3 <- colnames(df_table6)[grepl("^wave(3|4|5|6)?$", colnames(df_table6))]
    
    control_var_list <- vector(mode = "list", length = 10)
    for (i in seq(10)) {
      if (i %in% c(1, 2)) {
        control_var_list[[i]] <- control_var_1
      } else if (i %in% c(3, 4)) {
        control_var_list[[i]] <- control_var_2
      } else {
        control_var_list[[i]] <- control_var_3
      }
    }
  } else if (panel == "B") {
    df_table6 <- df_table6 %>% filter(female == 1)
    rhs_var <- c(
      "atreatcash_mlowgroup", "atreatcash_highgroup", 
      "atreatequip_mlowgroup", "atreatequip_highgroup"
      )
    control_var_1 <- colnames(df_table6)[grepl("^wave\\d(_mlowgroup)?$", colnames(df_table6))]
    control_var_2 <- colnames(df_table6)[grepl("^wave(5|6)(_mlowgroup)?$", colnames(df_table6))]
    
    control_var_list <- vector(mode = "list", length = 10)
    for (i in seq(10)) {
      if (i %in% c(1, 2)) {
        control_var_list[[i]] <- control_var_1
      } else if (i %in% c(3, 4)) {
        control_var_list[[i]] <- control_var_2
      } else {
        control_var_list[[i]] <- control_var_1
      }
    }
  }
  
  # list of functions to conduct F-test in each regression
  test_fun_list <- replicate(2, vector(mode = "list", length = 10), FALSE)
  if (panel == "A") {
    for (i in seq(10)) {
      test_fun_list[[1]][[i]] <- function(x) {
        formatC(
          waldtest(x, ~ atreatcashfemale - atreatequipfemale)["p.F"],
          digits = 3, format = "f"
          )
        }
      test_fun_list[[2]][[i]] <- function(x) {
        formatC(
          waldtest(x, ~ atreatcashmale - atreatequipmale)["p.F"],
          digits = 3, format = "f"
          )
        }
    }
  } else if (panel == "B") {
    for (i in seq(10)) {
      test_fun_list[[1]][[i]] <- function(x) {
        formatC(
          waldtest(x, ~ atreatcash_mlowgroup - atreatcash_highgroup)["p.F"],
          digits = 3, format = "f"
          )
        }
      test_fun_list[[2]][[i]] <- function(x) {
        formatC(
          waldtest(x, ~ atreatequip_mlowgroup - atreatequip_highgroup)["p.F"],
          digits = 3, format = "f"
          )
        }
    }
  }
  
  # run regressions
  reg_res_table6 <- tibble(
    model = pmap(
        list(a = lhs_var_list_trunc, b = fe_list, c = control_var_list), 
        function(a, b, c) felm_table6(a, b, rhs_var, c, iv, cluster, df_table6)
        )
    ) %>%
    mutate(
      test_1 = map2_chr(model, test_fun_list[[1]], function(x, y) y(x)),
      test_2 = map2_chr(model, test_fun_list[[2]], function(x, y) y(x)),
      n_firm = map_int(model, function(x) n_distinct(model.frame(x)$sheno)),
      mean_outcome = map2_chr(
        model, lhs_var_list_trunc, 
        function(x, y) formatC(colMeans(model.frame(x)[y]), 2, format = "f")
        )
      )

  if (panel == "A") {
    covariate_label <- c(
      "Cash treatment * female",
      "In-kind treatment * female",
      "Cash treatment * male",
      "In-kind treatment * male"
    )
    test_1_label <- "Cash = In-kind females"
    test_2_label <- "Cash = In-kind males"
  } else if (panel == "B") {
    covariate_label <- c(
      "Cash treatment * low profits",
      "Cash treatment * high profits",
      "In-kind treatment * low profits",
      "In-kind treatment * high profits"
    )
    test_1_label <- "Cash treatments equal"
    test_2_label <- "In-kind treatments equal"
  }
  
  # create a panel of regression results
  # and save as .html file
  reg_res_table6 %>%
    pull(model) %>%
    stargazer(
      dep.var.labels.include = FALSE,
      keep = rhs_var,
      covariate.labels = covariate_label,
      title = "Table",
      add.lines = list(
        c("Number of firms", reg_res_table6$n_firm),
        c("p-values for testing:", rep("", 10)),
        c(test_1_label, reg_res_table6$test_1),
        c(test_2_label, reg_res_table6$test_2),
        c("Mean of outcomes", reg_res_table6$mean_outcome)
      ),
      type = "html",
      out = paste0("html/FMQW2014_table6_replicate_", panel, ".html"),
      omit.stat = c("rsq", "adj.rsq", "ser"),
      table.layout = "=#c-t-sa-n",
      digits = 2
    )
  
}

create_table6_sub("A")
create_table6_sub("B")

# read .html file as a table and remove irrelevant rows
table6_A <- readHTMLTable("html/FMQW2014_table6_replicate_A.html") %>%
  .$Table %>%
  drop_na() %>%
  filter(!apply(., 1, function(x) all(x == ""))) %>%
  slice(-1) %>%
  as.matrix() %>%
  set_colnames(NULL)
# add indentations
table6_A[(nrow(table6_A) - 2):(nrow(table6_A) - 1), 1] <- paste0(
  "\\hspace{1em} ", 
  table6_A[(nrow(table6_A) - 2):(nrow(table6_A) - 1), 1]
  )
table6_A[,1] <- str_replace(table6_A[,1], "p-values", "$p$-values")

# read .html file as a table and remove irrelevant rows
table6_B <- readHTMLTable("html/FMQW2014_table6_replicate_B.html") %>%
  .$Table %>%
  drop_na() %>%
  filter(!apply(., 1, function(x) all(x == ""))) %>%
  slice(-1) %>%
  as.matrix() %>%
  set_colnames(NULL)
# add indentations
table6_B[(nrow(table6_B) - 2):(nrow(table6_B) - 1), 1] <- paste0(
  "\\hspace{1em} ", 
  table6_B[(nrow(table6_B) - 2):(nrow(table6_B) - 1), 1]
  )
table6_B[,1] <- str_replace(table6_B[,1], "p-values", "$p$-values")

# combine two panels and make .tex file
rbind(table6_A, table6_B) %>%
  kable("latex", booktabs = TRUE, escape = FALSE) %>%
  kable_styling(latex_options = "scale_down") %>%
  add_header_above(c("", sapply(seq(10), function(x) paste0("(", x, ")")))) %>%
  add_header_above(c(" ", rep("FE", 2), rep("OLS", 2), rep("FE", 6))) %>%
  add_header_above(c(
    "",
    "Capital\nstock",
    "Truncated\ncapital\nstock",
    "Made a\ntransfer\nout",
    "Amount\ntransferred\nout",
    "Weekly\nfood\nspending",
    "Quarterly\nclothing\nspending",
    "Health and\neducation\nspending",
    "Quarterly\nceremonies\nspending",
    "Total\nannual\nspending",
    "Log\nannual\nspending"
    )) %>%
  pack_rows(index = c(
    "Panel A: Males and females" = nrow(table6_A), 
    "Panel B: Female sub-sample" = nrow(table6_B)
    )) %>%
  save_kable("tex/FMQW2014_table6_replicate.tex")