MRW1992.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import statsmodels.api as sm\n",
    "from statsmodels.iolib.summary2 import summary_col\n",
    "import matplotlib.pyplot as plt\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.chdir(\"C:\\\\Users\\\\Hariharan\\\\Documents\\\\replications\\\\mrw1992\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv(\"data/mrw.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Reformatting the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "#convenience stuff\n",
    "data = data.drop(data.columns[0], axis = 1)\n",
    "data = data.set_index('country')\n",
    "\n",
    "#creating desired variables\n",
    "data['lngd'] = np.log(data['popgrowth']/100 + 0.05)\n",
    "data['ls'] = np.log(data['i_y']/100)\n",
    "data['const'] = 1\n",
    "data['ls_lngd'] = data['ls'] - data['lngd']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Standard Solow Model\n",
    "\n",
    "The model starts with a Cobb-Douglas production function:\n",
    "\n",
    "\\begin{align*}\n",
    "Y(t) = K(t)^{\\alpha}(A(t)L(t))^{1 - \\alpha}\n",
    "\\end{align*}\n",
    "\n",
    "The evolution of capital is determined by the following equation:\n",
    "\n",
    "\\begin{align*}\n",
    "\\dot{k(t)} = sk(t)^{\\alpha} - (n - g - \\delta)k(t)\n",
    "\\end{align*}\n",
    "\n",
    "From this we can get the steady state level of $k(t)$ by equating $\\dot{k(t)}$ to 0.  \n",
    "We then get a simple formula to determine the steady state $k(t)^*$\n",
    "\n",
    "\\begin{align*}\n",
    "k(t)^* = \\left(\\frac{s}{n + g + \\delta}\\right)^{1/(1 - \\alpha)}\n",
    "\\end{align*}\n",
    "\n",
    "We can take logs of the production function and substitute in $\\dot{k(t)}$ to find the equation we want to estimate, which is the steady state level of output. We also add an initial $A(0)$ which we take as equal to $a + \\epsilon$ where $a$ is a constant and $\\epsilon$ is a country specific shock\n",
    "\n",
    "\\begin{align*}\n",
    "log(Y/L) = a + \\frac{\\alpha}{(1 - \\alpha)}log(s) - \\frac{\\alpha}{(1 - \\alpha)}log(n + g + \\delta) + \\epsilon\n",
    "\\end{align*}\n",
    "\n",
    "where:\n",
    "a is the constant in the assumed $log(A)$ equation  \n",
    "$\\epsilon$ is the country shock for $log(a)$  \n",
    "$\\alpha$ is the share of capital  \n",
    "s is the exogenous savings rate (measured with average share of investment in GDP)  \n",
    "n is the exogenous population growth rate  \n",
    "g is the exogenous tfp growth rate  \n",
    "$\\delta$ is the depriciation rate of capital  \n",
    "\n",
    "This corresponds to equation 7 in the paper"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "data['lschool'] = np.log(data['school'])\n",
    "data['lsch_ngd'] = data['lschool'] - data['lngd']\n",
    "data['ly60'] = np.log(data['rgdpw60'])\n",
    "data['ly85'] = np.log(data['rgdpw85'])\n",
    "data['linv'] = np.log(data['i_y'])\n",
    "\n",
    "#subsetting data\n",
    "data_reg = data.loc[data['n'] == 1, :] #no oil\n",
    "data_d = data_reg[data_reg.i == 1] #pop in 1960 less than 1 mil\n",
    "data_oecd = data_reg[data_reg.o == 1] #oecd"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Unrestricted Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "#no restrictions on coefficients here\n",
    "reg1 = sm.OLS(endog = data_reg['ly85'],\n",
    "              exog = data_reg[['const', 'ls', 'lngd']],\n",
    "             missing = 'drop').fit()\n",
    "\n",
    "reg2 = sm.OLS(endog = data_d['ly85'],\n",
    "             exog = data_d[['const', 'ls', 'lngd']],\n",
    "             missing = 'drop').fit()\n",
    "\n",
    "reg3 = sm.OLS(endog = data_oecd['ly85'],\n",
    "             exog = data_oecd[['const', 'ls', 'lngd']],\n",
    "             missing = 'drop').fit()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Restricted Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "#coeff(log(savings)) = -coeff(log(n+g+d))\n",
    "regr1 = sm.OLS(endog = data_reg['ly85'],\n",
    "             exog = data_reg[['const', 'ls_lngd']],\n",
    "             missing = 'drop').fit()\n",
    "\n",
    "regr2 = sm.OLS(endog = data_d['ly85'],\n",
    "             exog = data_d[['const', 'ls_lngd']],\n",
    "             missing = 'drop').fit()\n",
    "\n",
    "regr3 = sm.OLS(endog = data_oecd['ly85'],\n",
    "             exog = data_oecd[['const', 'ls_lngd']],\n",
    "             missing = 'drop').fit()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Making table 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       Unrestricted Regressions\n",
      "=======================================\n",
      "          Non-Oil  Intermediate   OECD \n",
      "---------------------------------------\n",
      "const     5.43***  5.35***      8.02***\n",
      "          (1.58)   (1.54)       (2.52) \n",
      "ls        1.42***  1.32***      0.50   \n",
      "          (0.14)   (0.17)       (0.43) \n",
      "lngd      -1.99*** -2.02***     -0.74  \n",
      "          (0.56)   (0.53)       (0.85) \n",
      "R^2       0.5925   0.5878       0.0118 \n",
      "N         98.0000  75.0000      22.0000\n",
      "s.e.e.    0.6891   0.6106       0.3774 \n",
      "Implied α 0.59     0.57         0.33   \n",
      "=======================================\n",
      "Standard errors in parentheses.\n",
      "* p<.1, ** p<.05, ***p<.01\n",
      "\n",
      "        Restricted Regressions\n",
      "======================================\n",
      "          Non-Oil Intermediate   OECD \n",
      "--------------------------------------\n",
      "const     6.87*** 7.09***      8.62***\n",
      "          (0.12)  (0.15)       (0.53) \n",
      "ls_lngd   1.49*** 1.43***      0.55   \n",
      "          (0.12)  (0.14)       (0.37) \n",
      "R^2       0.5932  0.5861       0.0582 \n",
      "N         98.0000 75.0000      22.0000\n",
      "s.e.e.    0.6885  0.6119       0.3684 \n",
      "Implied α 0.60    0.59         0.36   \n",
      "======================================\n",
      "Standard errors in parentheses.\n",
      "* p<.1, ** p<.05, ***p<.01\n"
     ]
    }
   ],
   "source": [
    "info_dict = {'R^2': lambda x: x.rsquared_adj,\n",
    "            'N': lambda x: x.nobs,\n",
    "            's.e.e.': lambda x: np.sqrt(x.scale),\n",
    "            'Implied α': lambda x: f\"{x.params[1]/(1 + x.params[1]):.2f}\"}\n",
    "\n",
    "results_unres = summary_col(results = [reg1, reg2, reg3],\n",
    "                           float_format='%0.2f',\n",
    "                           stars = True,\n",
    "                           model_names = ['Non-Oil',\n",
    "                                         'Intermediate',\n",
    "                                         'OECD'],\n",
    "                           info_dict = info_dict,\n",
    "                           regressor_order = ['const',\n",
    "                                             'ls',\n",
    "                                             'lngd'])\n",
    "\n",
    "\n",
    "results_res = summary_col(results = [regr1, regr2, regr3],\n",
    "                           float_format='%0.2f',\n",
    "                           stars = True,\n",
    "                           model_names = ['Non-Oil',\n",
    "                                         'Intermediate',\n",
    "                                         'OECD'],\n",
    "                           info_dict = info_dict,\n",
    "                           regressor_order = ['const',\n",
    "                                             'ls',\n",
    "                                             'ls_lngd'])\n",
    "\n",
    "\n",
    "results_res.add_title('Restricted Regressions')\n",
    "results_unres.add_title('Unrestricted Regressions')\n",
    "print(results_unres)\n",
    "print('')\n",
    "print(results_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Interpretting\n",
    "\n",
    "Lets not get starry eyed with the regression tables too quickly.  \n",
    "\n",
    "MRW take these regression tables with cautious optimism. While we see the direction and significance in the variables we want in the way we expect - i.e. savings positively determines steady state level of output per capita, and $(n + g + \\delta)$ negatively affects output per capita. Also putting a restriction on the coefficient $(n + g + \\delta)$ doesn't really change our results too much. \n",
    "Also, the model seems to explain a lot of the cross country variation in output per capita ($R^2$ is fairly high). But this shouldn't be taken as evidence for the Solow model. \n",
    "\n",
    "The big but here is that the implied $\\alpha$ level isn't consistent with the empirical level of capital share, according to which the capital share should've been around 1/3.  \n",
    "\n",
    "This basically amounts to these regressions being very weak evidence for the Solow model. If we think of all the things the Solow model says, this regression only says 1 thing: savings, population growth and depriciation are highly correlated to GDP, which is consistent with an entire class of growth models. \n",
    "\n",
    "Because of this, MRW augment the solow model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Augmenting the Solow Model\n",
    "\n",
    "MRW Augment the Solow model with human capital. \n",
    "\n",
    "With that, the Cobb-Douglas production function becomes:\n",
    "\\begin{align*}\n",
    "Y(t) = K(t)^{\\alpha}H(t)^{\\beta}(A(t)L(t))^{1 - \\alpha - \\beta}\n",
    "\\end{align*}\n",
    "\n",
    "Evolution of the economy is determined by:\n",
    "\n",
    "\\begin{align*}\n",
    "\\dot{k(t)} = s_{k}y(t) - (n + g +\\delta)k(t)\\\\\n",
    "\\dot{h(t)} = s_{h}y(t) - (n + g +\\delta)h(t)\n",
    "\\end{align*}\n",
    "\n",
    "We can use these to derive the steady state values for k and h by setting $\\dot{k(t)}$ and $\\dot{h(t)}$ to 0.  \n",
    "Using those we can arrive at the equation that we want to estimate.\n",
    "\n",
    "\\begin{align*}\n",
    "log(Y(t)/L(t)) = log(A(0)) + gt + \\frac{\\alpha}{1 - \\alpha - \\beta}log(s_k) - \\frac{\\alpha + \\beta}{1 - \\alpha - \\beta}log(n + g + \\delta) + \\frac{\\beta}{1 - \\alpha -\\beta}log(s_h)\n",
    "\\end{align*}\n",
    "\n",
    "This is very similar to the standard solow model equation we estimated before except for the last term. We can think of the last term being in the error term in our previous specification. This would'nt be a problem by itself but since $s_h$ is determined by $s_k$ and $n$, this would've biased our coefficients in the previous specification"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Unrestricted Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "#no restrictions on coefficients here\n",
    "regh1 = sm.OLS(endog = data_reg['ly85'],\n",
    "             exog = data_reg[['const', 'ls', 'lngd', 'lschool']],\n",
    "             missing = 'drop').fit()\n",
    "\n",
    "regh2 = sm.OLS(endog = data_d['ly85'],\n",
    "             exog = data_d[['const', 'ls', 'lngd', 'lschool']],\n",
    "             missing = 'drop').fit()\n",
    "\n",
    "regh3 = sm.OLS(endog = data_oecd['ly85'],\n",
    "             exog = data_oecd[['const', 'ls', 'lngd', 'lschool']],\n",
    "             missing = 'drop').fit()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Restricted Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "#coeff(log(savings)) = -coeff(log(n+g+d))\n",
    "reghr1 = sm.OLS(endog = data_reg['ly85'],\n",
    "             exog = data_reg[['const', 'ls_lngd', 'lsch_ngd']],\n",
    "             missing = 'drop').fit()\n",
    "\n",
    "reghr2 = sm.OLS(endog = data_d['ly85'],\n",
    "             exog = data_d[['const', 'ls_lngd', 'lsch_ngd']],\n",
    "             missing = 'drop').fit()\n",
    "\n",
    "reghr3 = sm.OLS(endog = data_oecd['ly85'],\n",
    "             exog = data_oecd[['const', 'ls_lngd', 'lsch_ngd']],\n",
    "             missing = 'drop').fit()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       Unrestricted Regressions\n",
      "=======================================\n",
      "          Non-Oil  Intermediate   OECD \n",
      "---------------------------------------\n",
      "const     3.83***  4.43***      5.10*  \n",
      "          (1.18)   (1.16)       (2.47) \n",
      "ls        0.70***  0.70***      0.28   \n",
      "          (0.13)   (0.15)       (0.39) \n",
      "lngd      -1.75*** -1.50***     -1.08  \n",
      "          (0.42)   (0.40)       (0.76) \n",
      "lschool   0.65***  0.73***      0.77** \n",
      "          (0.07)   (0.10)       (0.29) \n",
      "R^2       0.7788   0.7714       0.2444 \n",
      "N         98.0000  75.0000      22.0000\n",
      "s.e.e.    0.5077   0.4547       0.3300 \n",
      "Implied α 0.30     0.29         0.14   \n",
      "Implied β 0.28     0.30         0.38   \n",
      "=======================================\n",
      "Standard errors in parentheses.\n",
      "* p<.1, ** p<.05, ***p<.01\n",
      "\n",
      "        Restricted Regressions\n",
      "======================================\n",
      "          Non-Oil Intermediate   OECD \n",
      "--------------------------------------\n",
      "const     4.83*** 4.59***      5.18***\n",
      "          (0.24)  (0.34)       (1.36) \n",
      "ls_lngd   0.74*** 0.71***      0.28   \n",
      "          (0.12)  (0.14)       (0.33) \n",
      "lsch_ngd  0.66*** 0.73***      0.77** \n",
      "          (0.07)  (0.09)       (0.28) \n",
      "R^2       0.7794  0.7745       0.2841 \n",
      "N         98.0000 75.0000      22.0000\n",
      "s.e.e.    0.5070  0.4516       0.3212 \n",
      "Implied α 0.31    0.29         0.14   \n",
      "Implied β 0.27    0.30         0.37   \n",
      "======================================\n",
      "Standard errors in parentheses.\n",
      "* p<.1, ** p<.05, ***p<.01\n"
     ]
    }
   ],
   "source": [
    "info_dictu = {'R^2': lambda x: x.rsquared_adj,\n",
    "            'N': lambda x: x.nobs,\n",
    "            's.e.e.': lambda x: np.sqrt(x.scale),\n",
    "            'Implied α': lambda x: f\"{x.params[1]/(1 + x.params[1] + x.params[3]):.2f}\",\n",
    "            'Implied β': lambda x: f\"{x.params[3]/(1 + x.params[1] + x.params[3]):.2f}\"}\n",
    "\n",
    "info_dictr = {'R^2': lambda x: x.rsquared_adj,\n",
    "            'N': lambda x: x.nobs,\n",
    "            's.e.e.': lambda x: np.sqrt(x.scale),\n",
    "            'Implied α': lambda x: f\"{x.params[1]/(1 + x.params[1] + x.params[2]):.2f}\",\n",
    "            'Implied β': lambda x: f\"{x.params[2]/(1 + x.params[1] + x.params[2]):.2f}\"}\n",
    "\n",
    "results_unres = summary_col(results = [regh1, regh2, regh3],\n",
    "                           float_format='%0.2f',\n",
    "                           stars = True,\n",
    "                           model_names = ['Non-Oil',\n",
    "                                         'Intermediate',\n",
    "                                         'OECD'],\n",
    "                           info_dict = info_dictu,\n",
    "                           regressor_order = ['const',\n",
    "                                             'ls',\n",
    "                                             'lngd',\n",
    "                                             'lschool'])\n",
    "\n",
    "\n",
    "results_res = summary_col(results = [reghr1, reghr2, reghr3],\n",
    "                           float_format='%0.2f',\n",
    "                           stars = True,\n",
    "                           model_names = ['Non-Oil',\n",
    "                                         'Intermediate',\n",
    "                                         'OECD'],\n",
    "                           info_dict = info_dictr,\n",
    "                           regressor_order = ['const',\n",
    "                                             'ls',\n",
    "                                             'ls_lngd',\n",
    "                                             'lsch_ngd'])\n",
    "\n",
    "\n",
    "results_res.add_title('Restricted Regressions')\n",
    "results_unres.add_title('Unrestricted Regressions')\n",
    "print(results_unres)\n",
    "print('')\n",
    "print(results_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Interpretation and Evaluation\n",
    "\n",
    "The $\\alpha$ and $\\beta$ implied from the regressions are much closer to what we would expect. So adding human capital seems to have improved the Solow model's performance. $R^2$ also seems to be much higher.\n",
    "\n",
    "There's a big elephant in the room which we should address at this point. Endogeneity.\n",
    "\n",
    "MRW really take no worry of endogeneity issues in these regressions. Endogeneity comes in 3 flavours:\n",
    "1. Ommitted variable bias: Is the error term in the equations we estimated orthogonal to the exogenous regressors and the outcome? Probably not. For example, level of institutions in a country would be correlated to the savings rate and the GDP (possible correlated positively with both). This would mean that we have estimates with an upward bias\n",
    "2. Reverse Causality: Is it likely that gdp affects savings? Yes. We can imagine that higher gdp would increase or decrease the savings rate. Either way, this again implies a bias in our estimates.\n",
    "3. Measurement Error: Is it likely that one or more of our RHS variables are mismeasuring what we actually want to measure. Take for instance human-capital. We assumed $s_h$ to be our proxy for it. But if there is measurement error there, that error will be present in our residual in the regression, which will then be correlated with our RHS variable thus biasing the estimates.\n",
    "\n",
    "It almost seems circular to follow these regressions to conclude that the Solow model is right. If we take these regressions as correctly specified, we are basically assuming all our righthand-side variables to be exogenous, which is already half way there to the assumptions of Solow, where we take savings and growth rates as exogenous.\n",
    "\n",
    "It will also be interesting to see how sensitive these results are to what we use for human capital. For example, it isn't clear why we should restrict human-capital to be restricted to the share of the working-age population in secondary school.\n",
    "\n",
    "That is not to say that these regressions are useless. Even if the argument is a bit circular, the fact that the magnitudes of $\\alpha$ and $\\beta$ turn out as predicted is still impressive. But it seems more predictive than causal"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Solow Rate of Convergence\n",
    "\n",
    "Now we move on to estimating the rate of convergence in a Solow model.\n",
    "\n",
    "If we start out with our base cobb douglas equation for output per worker, try to solve for the relative growth rate of y, on the way applying some taylor approxiamtions and stuff we know from differential equations, we get a linear equation that we can estimate to give us an idea of the rate of convergence:\n",
    "\n",
    "\\begin{align}\n",
    "\\ln(y(t)) - \\ln(y(0)) = (1 - e^{-\\lambda t})\\left(\\frac{\\alpha}{1 - \\alpha - \\beta}\\right)\\ln(s_k) + (1 - e^{-\\lambda t})\\left(\\frac{\\beta}{1 - \\alpha - \\beta}\\right)\\ln(s_h) - (1 - e^{-\\lambda t})\\left(\\frac{\\alpha + \\beta}{1 - \\alpha + \\beta}\\right)\\ln(n + g+ \\delta) - (1 - e^{-\\lambda t})\\ln(y(0))\n",
    "\\end{align}\n",
    "\n",
    "First we will go through a much more basic regression: \n",
    "\\begin{align}\n",
    "\\ln(y(t)) - \\ln(y(0)) = \\beta_0 + \\beta_1 \\ln(y(0)) + \\epsilon\n",
    "\\end{align}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "regcon1 = sm.OLS(endog = (data_reg['ly85'] - data_reg['ly60']),\n",
    "               exog = data_reg[['const', 'ly60']],\n",
    "               missing = 'drop').fit()\n",
    "\n",
    "regcon2 = sm.OLS(endog = (data_d['ly85'] - data_d['ly60']),\n",
    "               exog = data_d[['const', 'ly60']],\n",
    "               missing = 'drop').fit()\n",
    "\n",
    "regcon3 = sm.OLS(endog = (data_oecd['ly85'] - data_oecd['ly60']),\n",
    "               exog = data_oecd[['const', 'ly60']],\n",
    "               missing = 'drop').fit()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    Tests for Unconditional Convergence\n",
      "============================================\n",
      "           Non-Oil  Intermediate     OECD   \n",
      "--------------------------------------------\n",
      "const     -0.26658  0.58752      3.68629*** \n",
      "          (0.37960) (0.43293)    (0.68490)  \n",
      "ly60      0.09431*  -0.00424     -0.34110***\n",
      "          (0.04962) (0.05484)    (0.07852)  \n",
      "R^2       0.0262    -0.0136      0.4597     \n",
      "N         98.0000   75.0000      22.0000    \n",
      "s.e.e.    0.4405    0.4077       0.1830     \n",
      "Implied λ -0.00360  0.00017      0.01669    \n",
      "============================================\n",
      "Standard errors in parentheses.\n",
      "* p<.1, ** p<.05, ***p<.01\n"
     ]
    }
   ],
   "source": [
    "info_dictroc = {'R^2': lambda x: x.rsquared_adj,\n",
    "            'N': lambda x: x.nobs,\n",
    "            's.e.e.': lambda x: np.sqrt(x.scale),\n",
    "            'Implied λ': lambda x: f\"{-np.log(x.params[1] + 1)/25:.5f}\"}\n",
    "\n",
    "table_roc = summary_col(results = [regcon1, regcon2, regcon3],\n",
    "                           float_format='%0.5f',\n",
    "                           stars = True,\n",
    "                           model_names = ['Non-Oil',\n",
    "                                         'Intermediate',\n",
    "                                         'OECD'],\n",
    "                           info_dict = info_dictroc,\n",
    "                           regressor_order = ['const',\n",
    "                                             'ly60'])\n",
    "\n",
    "table_roc.add_title('Tests for Unconditional Convergence')\n",
    "print(table_roc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "regcon4 = sm.OLS(endog = (data_reg['ly85'] - data_reg['ly60']),\n",
    "               exog = data_reg[['const', 'ly60', 'ls', 'lngd']],\n",
    "               missing = 'drop').fit()\n",
    "\n",
    "regcon5 = sm.OLS(endog = (data_d['ly85'] - data_d['ly60']),\n",
    "               exog = data_d[['const', 'ly60', 'ls', 'lngd']],\n",
    "               missing = 'drop').fit()\n",
    "\n",
    "regcon6 = sm.OLS(endog = (data_oecd['ly85'] - data_oecd['ly60']),\n",
    "               exog = data_oecd[['const', 'ly60', 'ls', 'lngd']],\n",
    "               missing = 'drop').fit()\n",
    "\n",
    "\n",
    "\n",
    "regcon7 = sm.OLS(endog = (data_reg['ly85'] - data_reg['ly60']),\n",
    "               exog = data_reg[['const', 'ly60', 'ls', 'lngd', 'lschool']],\n",
    "               missing = 'drop').fit()\n",
    "\n",
    "regcon8 = sm.OLS(endog = (data_d['ly85'] - data_d['ly60']),\n",
    "               exog = data_d[['const', 'ly60', 'ls', 'lngd', 'lschool']],\n",
    "               missing = 'drop').fit()\n",
    "\n",
    "regcon9 = sm.OLS(endog = (data_oecd['ly85'] - data_oecd['ly60']),\n",
    "               exog = data_oecd[['const', 'ly60', 'ls', 'lngd', 'lschool']],\n",
    "               missing = 'drop').fit()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "      Tests for Conditional Convergence\n",
      "==============================================\n",
      "            Non-Oil   Intermediate     OECD   \n",
      "----------------------------------------------\n",
      "const     1.91938**   2.24968**    2.14036*   \n",
      "          (0.83367)   (0.85472)    (1.18068)  \n",
      "ly60      -0.14090*** -0.22783***  -0.34991***\n",
      "          (0.05202)   (0.05725)    (0.06574)  \n",
      "ls        0.64724***  0.64587***   0.39010**  \n",
      "          (0.08670)   (0.10392)    (0.17612)  \n",
      "lngd      -0.30235    -0.45746     -0.76624** \n",
      "          (0.30438)   (0.30743)    (0.34523)  \n",
      "R^2       0.3828      0.3526       0.6228     \n",
      "N         98.0000     75.0000      22.0000    \n",
      "s.e.e.    0.3507      0.3258       0.1529     \n",
      "Implied λ 0.00607     0.01034      0.01723    \n",
      "==============================================\n",
      "Standard errors in parentheses.\n",
      "* p<.1, ** p<.05, ***p<.01\n",
      " \n",
      "      Tests for Conditional Convergence\n",
      "==============================================\n",
      "            Non-Oil   Intermediate     OECD   \n",
      "----------------------------------------------\n",
      "const     1.95719**   2.46351***   1.70676    \n",
      "          (0.77744)   (0.80116)    (1.16846)  \n",
      "ly60      -0.28837*** -0.36599***  -0.39769***\n",
      "          (0.06158)   (0.06743)    (0.07016)  \n",
      "ls        0.52374***  0.53756***   0.33179*   \n",
      "          (0.08687)   (0.10229)    (0.17338)  \n",
      "lngd      -0.50566*   -0.54499*    -0.86341** \n",
      "          (0.28861)   (0.28843)    (0.33770)  \n",
      "lschool   0.23112***  0.27045***   0.22770    \n",
      "          (0.05946)   (0.08037)    (0.14501)  \n",
      "R^2       0.4633      0.4348       0.6512     \n",
      "N         98.0000     75.0000      22.0000    \n",
      "s.e.e.    0.3270      0.3044       0.1470     \n",
      "Implied λ 0.01361     0.01823      0.02028    \n",
      "==============================================\n",
      "Standard errors in parentheses.\n",
      "* p<.1, ** p<.05, ***p<.01\n"
     ]
    }
   ],
   "source": [
    "info_dictroc2 = {'R^2': lambda x: x.rsquared_adj,\n",
    "            'N': lambda x: x.nobs,\n",
    "            's.e.e.': lambda x: np.sqrt(x.scale),\n",
    "            'Implied λ': lambda x: f\"{-np.log(x.params[1] + 1)/25:.5f}\"}\n",
    "\n",
    "table_roc2 = summary_col(results = [regcon4, regcon5, regcon6],\n",
    "                           float_format='%0.5f',\n",
    "                           stars = True,\n",
    "                           model_names = ['Non-Oil',\n",
    "                                         'Intermediate',\n",
    "                                         'OECD'],\n",
    "                           info_dict = info_dictroc,\n",
    "                           regressor_order = ['const',\n",
    "                                             'ly60',\n",
    "                                             'ls',\n",
    "                                             'lngd'])\n",
    "\n",
    "\n",
    "table_roc3 = summary_col(results = [regcon7, regcon8, regcon9],\n",
    "                           float_format='%0.5f',\n",
    "                           stars = True,\n",
    "                           model_names = ['Non-Oil',\n",
    "                                         'Intermediate',\n",
    "                                         'OECD'],\n",
    "                           info_dict = info_dictroc,\n",
    "                           regressor_order = ['const',\n",
    "                                             'ly60',\n",
    "                                             'ls',\n",
    "                                             'lngd',\n",
    "                                             'lschool'])\n",
    "\n",
    "table_roc2.add_title('Tests for Conditional Convergence')\n",
    "table_roc3.add_title('Tests for Conditional Convergence')\n",
    "print(table_roc2)\n",
    "print(\" \")\n",
    "print(table_roc3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "regconr1 = sm.OLS(endog = (data_reg['ly85'] - data_reg['ly60']),\n",
    "               exog = data_reg[['const', 'ly60', 'ls_lngd', 'lsch_ngd']],\n",
    "               missing = 'drop').fit()\n",
    "\n",
    "regconr2 = sm.OLS(endog = (data_d['ly85'] - data_d['ly60']),\n",
    "               exog = data_d[['const', 'ly60', 'ls_lngd', 'lsch_ngd']],\n",
    "               missing = 'drop').fit()\n",
    "\n",
    "regconr3 = sm.OLS(endog = (data_oecd['ly85'] - data_oecd['ly60']),\n",
    "               exog = data_oecd[['const', 'ly60', 'ls_lngd', 'lsch_ngd']],\n",
    "               missing = 'drop').fit()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Replicating Figure 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       Rate of Convergence - Restricted\n",
      "==============================================\n",
      "            Non-Oil   Intermediate     OECD   \n",
      "----------------------------------------------\n",
      "const     1.37384***  1.86679***   2.44270*** \n",
      "          (0.33571)   (0.36760)    (0.69118)  \n",
      "ly60      -0.29790*** -0.37236***  -0.40221***\n",
      "          (0.06041)   (0.06687)    (0.06918)  \n",
      "ls_lngd   0.50067***  0.50635***   0.39532**  \n",
      "          (0.08219)   (0.09508)    (0.15174)  \n",
      "lsch_ngd  0.23519***  0.26569***   0.24125    \n",
      "          (0.05916)   (0.08000)    (0.14244)  \n",
      "R^2       0.4651      0.4371       0.6586     \n",
      "N         98.0000     75.0000      22.0000    \n",
      "s.e.e.    0.3265      0.3038       0.1454     \n",
      "Implied λ 0.01415     0.01863      0.02058    \n",
      "==============================================\n",
      "Standard errors in parentheses.\n",
      "* p<.1, ** p<.05, ***p<.01\n"
     ]
    }
   ],
   "source": [
    "info_dictrocr = {'R^2': lambda x: x.rsquared_adj,\n",
    "            'N': lambda x: x.nobs,\n",
    "            's.e.e.': lambda x: np.sqrt(x.scale),\n",
    "            'Implied λ': lambda x: f\"{-np.log(x.params[1] + 1)/25:.5f}\"}\n",
    "table_rocr = summary_col(results = [regconr1, regconr2, regconr3],\n",
    "                           float_format='%0.5f',\n",
    "                           stars = True,\n",
    "                           model_names = ['Non-Oil',\n",
    "                                         'Intermediate',\n",
    "                                         'OECD'],\n",
    "                           info_dict = info_dictroc,\n",
    "                           regressor_order = ['const',\n",
    "                                             'ly60',\n",
    "                                             'ls_lngd', \n",
    "                                              'lsch_ngd'])\n",
    "\n",
    "table_rocr.add_title('Rate of Convergence - Restricted')\n",
    "print(table_rocr)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Replicating plots\n",
    "\n",
    "First we carry out some regressions for panel B and C to get out the residuals we need"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "reg_plt2 = sm.OLS(data_d['ly60'],\n",
    "                 data_d[['lngd', 'linv']],\n",
    "                 missing = 'drop').fit()\n",
    "residly601 = reg_plt2.resid\n",
    "\n",
    "reg_plt3 = sm.OLS(data_d['ly85'] - data_d['ly60'],\n",
    "                 data_d[['lngd', 'linv']],\n",
    "                 missing = 'drop').fit()\n",
    "residgr1 = reg_plt3.resid\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "reg_plt4 = sm.OLS(data_d['ly60'],\n",
    "                 data_d[['lngd', 'linv', 'lschool']],\n",
    "                 missing = 'drop').fit()\n",
    "residly602 = reg_plt4.resid\n",
    "\n",
    "reg_plt5 = sm.OLS(data_d['ly85'] - data_d['ly60'],\n",
    "                 data_d[['lngd', 'linv', 'lschool']],\n",
    "                 missing = 'drop').fit()\n",
    "residgr2 = reg_plt5.resid\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x864 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(3, 1, sharex = 'col', figsize = (10, 12))\n",
    "ax[0].scatter(data_d['ly60'], (data_d['ly85'] - data_d['ly60'])*100/25)\n",
    "ax[0].set_xlabel('Log GDP Per Worker in 1960')\n",
    "ax[0].set_ylabel('Growth Rate: 1960 - 1985')\n",
    "ax[0].set_title('Unconditional')\n",
    "\n",
    "\n",
    "ax[1].scatter(residly601 + np.mean(data_d['ly60']), \n",
    "           (residgr1 + np.mean(data_d['ly85'] - data_d['ly60']))*100/25)\n",
    "ax[1].set_xlabel('Log GDP Per Worker in 1960')\n",
    "ax[1].set_ylabel('Growth Rate: 1960 - 1985')\n",
    "ax[1].set_title('Conditional on saving and population growth')\n",
    "\n",
    "\n",
    "ax[2].scatter(residly602 + np.mean(data_d['ly60']), \n",
    "           (residgr2 + np.mean(data_d['ly85'] - data_d['ly60']))*100/25)\n",
    "ax[2].set_xlabel('Log GDP Per Worker in 1960')\n",
    "ax[2].set_ylabel('Growth Rate: 1960 - 1985')\n",
    "ax[2].set_title('Conditional on human capital, saving and population growth')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}