## _wp12295

## Source details

**Canonical URL:** [_wp12295](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12295.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12295.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12295.pdf.json)

---

### Introduction
- Lipset (1959) reported a strong and positive correlation between income per capita and democracy.
- Przeworski et al. (2000) failed to find a significant relationship using Markov transition models; this finding was replicated by Acemoglu, Johnson, Robinson and Yared (2008) (AJRY).
- This paper challenges AJRY by:
  - Reporting a significant statistical relationship between income and democracy for countries that were neither consistent democracies nor consistent non-democracies over the past 50 years.
  - Finding the relationship to be negative.
- AJRY also reported, but did not comment upon, negative coefficients in their estimates.

### Additional background
- Two literatures related to this work:
  - The “rentier state” and “political resource curse” literatures (Mahdavy, 1970; Beblawi and Giacomi, 1987; Chaudry, 1994; Barro, 1996; Barro, 1999; Ross, 1999; Ross, 2001).
    - Mechanisms: large oil revenues can reduce political discontent via low taxes and high public benefits, increased internal security spending, and lower education/occupational specialization leading to failure to bring about democracy (Ross, 2001, p.336/337).
  - The literature on economic growth and political instability (Haggard and Kaufman, 1997; Geddes, 1999; Burke and Leigh, 2010; Alesina and Perrotti, 1994; Londregan and Poole, 1990).
    - Mechanisms: growth reversals and economic contractions can weaken authoritarian regimes, reduce opportunity costs of protest, and increase likelihood of government change or coups.

### Data and methods

- Data (main series and sample coverage):
  - Penn World Tables (PWT 6.3) chain weighted real GDP per capita series (log of real GDP per capita).
  - Polity IV democracy index ranging from -10 (perfect autocracies) to 10 (perfect democracies).
  - Sample includes 105 countries for main analyses; Figure 1 and Figure 2 show average trends 1960−2008.
  - Decomposition of income into resource and non-resource components uses United Nations Statistics Division National Accounts Database; data available for 104 out of 105 countries but run only from 1970–2007.
  - Countries excluded due to short time series or no time variation in the dependent variable are listed in the Appendix (examples provided in source).
- Key descriptive statistics (from Table B):
  - Democracy: Obs 5085, Countries 105, Mean -1.0, Std. Dev. 6.9, Min -10.0, Max 10.0
  - Log of Real Income per Capita: Obs 5076, Countries 105, Mean 8.0, Std. Dev. 1.0, Min 5.0, Max 11.5
  - Rents to GDP: Obs 3724, Countries 98, Mean 7.7, Std. Dev. 13.5, Min 0.0, Max 116.5
  - Log of Resource Income per Capita: Obs 3869, Countries 102, Mean 5.2, Std. Dev. 1.8, Min -2.4, Max 11.1
  - Share of Resource Income: Obs 3876, Countries 102, Mean 10.0, Std. Dev. 13.9, Min 0.07, Max 77.5

- Methods:
  - Estimation technique:
    - Augmented Pooled Mean Group (PMG) estimator (Pesaran et al., 1999) to allow intercepts, slope coefficients and error variances to vary across panel members; short-run coefficients vary across countries while long-run relationships are homogeneous.
    - Model augmented with cross-sectional averages of the dependent variable and regressors to proxy global/common factors (following Binder and Offermanns, 2007).
  - Concerns addressed:
    - Time-specific heterogeneity and cross-sectional dependence: captured via augmented PMG with cross-sectional averages.
    - Endogeneity and reverse causality:
      - Granger causality tests (Table 1) indicate reverse causality:
        - "Democracy does not Granger cause income": Observations 4532, Lags 3, F-stat 5.47, Probability 0.001
        - "Income does not Granger cause democracy": Observations 4532, Lags 3, F-stat 6.87, Probability 0.000
      - Addressed by augmenting PMG model with lags of regressors and dependent variable (ARDL framework); lag order chosen via AIC or SBC, constrained to a maximum of three lags.
      - Alternative heterogeneous panel cointegration techniques: group mean FMOLS (Pedroni, 2000) and DOLS (Pedroni, 2001).
  - Model specification: Error Correction Model with p lags; long-run slope coefficients denoted β, η, α for income, world income, and world democracy respectively; φ is the error correction coefficient indicating speed of adjustment.

### Estimation and results

- Reproducing AJRY:
  - Extended dataset (153 countries annual; 129 countries for five- and ten-year panels; 1960–2000) reproduces AJRY: pooled OLS positive and significant income coefficient; inclusion of country and year fixed effects renders income coefficient insignificant (Table 2).
  - PMG sample reproduces AJRY findings when using their model on that sample.

- Main PMG results (Table 3; Overall Sample N=105; 1955–2007):
  - Long-run PMG coefficients (column 1):
    - Log income per Capita: -1.239*** (standard error 0.153)
    - World democracy: 0.800*** (standard error 0.029)
    - World output: 3.059*** (standard error 0.547)
    - Error correction coefficient: -0.264*** (standard error 0.029)
  - Interpretation examples:
    - A 10 percent increase in per capita income leads in the long run to a roughly 0.12 unit decrease on the polity scale (log-linear interpretation).
    - When global democracy score increases by one unit, democracy score improves by an average of 0.8 units.
    - A 10 percent increase in world income improves the democracy score by 0.3 units.
  - Robustness:
    - Error correction coefficient suggests about 26 percent of disequilibrium corrected in a single period.
    - Results robust to AIC vs. SBC, number of lags, cross-sectional demeaning, and exclusion/inclusion of Sub-Saharan Africa.

- Alternative estimators (Table 4; N=105; 1955–2007):
  - DOLS: Log of income per capita coefficient -2.293*** (standard error 0.189) with country and year FE = YES
  - FMOLS: Log of income per capita coefficient -0.503** (standard error 0.197) with country and year FE = YES

- Pooled error correction OLS on PMG sample (Table 5; 1960–2007; N=105):
  - Long-run coefficients across lag specifications (Lags=2 to Lags=5):
    - Log of income per capita: -1.325* (p-value=0.07), -1.236 (p-value=0.11), -1.595* (p-value=0.07), -1.683* (p-value=0.06)
    - Error correction coefficient: -0.141*** to -0.144*** across specifications
    - R-squared: 0.85
  - When the pooled ECM is estimated on the AJRY (bigger) sample, long-run coefficients on income per capita are insignificant regardless of number of lags.

- Interpretation of differences with AJRY:
  - Differences arise from:
    1. Estimation methods exploiting dynamics and heterogeneity.
    2. Sample choice excluding consistent autocracies and consistent democracies.
  - Restricting the sample to countries with time variation in income and polity allows detection of a significantly negative relationship between income and democracy.
  - Sub-period check: Results for sub-period 1985–2007 (not reported) maintain the negative relationship; with the PMG model global output becomes significantly negatively related to democracy.

### Digging deeper: sectoral composition and regional patterns

- Resource rents and democracy (Table 6; N=98; 1970–2007):
  - Long-run PMG coefficients:
    - Log income per capita: -0.651*** (standard error 0.157)
    - Natural resource rents to GDP: -0.021*** (standard error 0.005)
    - World democracy: 1.211*** (standard error 0.035)
    - World output: -8.554*** (standard error 1.268)
    - World rents: 0.201*** (standard error 0.038)
    - Error correction coefficient: -0.355*** (standard error 0.039)
  - Finding: natural resource rents to GDP negatively and significantly related to democracy.

- Decomposing national income (resource vs non-resource) (Table 7; N=102; 1970–2007):
  - Long-run PMG coefficients (selected specifications):
    - Log Income per capita (baseline): -1.228*** (standard error 0.259)
    - Log of resource GDP per capita: -0.306*** (standard error 0.081)
    - Log of non-resource GDP per capita: 0.710** (standard error 0.271) in one specification and 0.518* (standard error 0.272) in another
    - Error correction coefficient reported across specifications (e.g., -0.305***, -0.248***, 0.245***, -0.307*** with standard errors)
  - Conclusion: the negative relationship between income and democracy is driven by the resource component of income; resource proportion of income per capita is negatively and significantly related to democracy, whereas non-resource income shows a positive relationship.

### Conclusion
- Main substantive findings:
  - For the group of countries that were neither consistent democracies nor consistent non-democracies over the sample period, there is a significant and negative long-run relationship between income and democracy.
  - The negative relationship is concentrated in resource-rich countries; in countries with little or no income from resources, the relationship between income and democracy is positive and significant.
- Methodological contribution:
  - Exploiting short-run cross-country heterogeneity in responses to income shocks and variation in income composition reveals relationships overlooked by prior work that relied on fixed effects models purging informative variation.

*Source: _wp12295 - INTRODUCTION (authors’ calculations and tables as provided in the supplied content).*

### Introduction ...........................................................................................................

### _wp12295 - Introduction ...........................................................................................................

### Major sections
- Introduction ................................................................................................................................3
- I. Additional Background...........................................................................................................4
- II. Data and Methods ..................................................................................................................4
  - A. Data ...........................................................................................................................5
  - B. Methods .....................................................................................................................7
- III. Estimation and Results .......................................................................................................10
- IV. Digging Deeper ..................................................................................................................14
- V. Conclusion ..........................................................................................................................16
- References ................................................................................................................................23

### Tables listed
- Table 1. Granger Causality Test ..........................................................................................................8
- Table 2. Reproducing AJRY with Pooled OLS and Fixed Effects; Annual Data 1960–2000.........11
- Table 3. Augmented PMG Estimation; Overall Sample (N=105); 1955–2007 .................................12
- Table 4. Pooled Error Correction Model by OLS With Country and Year Fixed Effects on the PMG Sample (1960–2007; N=105) Granger Causality Test .............................................................13
- Table 5. Augmented PMG Estimation; Overall Sample (N=98); 1970–2007 ...................................15
- Table 6. Augmented PMG Estimation; Overall Sample (N=102); 1970–2007 .................................16

### Figures listed
- Figure 1. Democracy and Income Global Averages, 1960–2008 .........................................................5
- Figure 2. Regional Average Polity Scores, 1960–2008 ........................................................................6

### Appendix items
- Table A ......................................................................................................................................18
- Table B ......................................................................................................................................19
- Table C ......................................................................................................................................20
- Table D ......................................................................................................................................21
- Figure A .....................................................................................................................................22

*Source: _wp12295 - Introduction (source PDF filename: _wp12295 - Introduction ...........................................................................................................)*

### INTRODUCTION

### _wp12295 - INTRODUCTION

### Introduction
- Lipset (1959) reported a strong and positive correlation between income per capita and democracy.
- Przeworski et al. (2000) failed to find a significant relationship using Markov transition models; this finding was replicated by Acemoglu, Johnson, Robinson and Yared (2008) (AJRY).
- This paper challenges AJRY by:
  - Reporting a significant statistical relationship between income and democracy for countries that were neither consistent democracies nor consistent non-democracies over the past 50 years.
  - Finding the relationship to be negative.
- AJRY also reported, but did not comment upon, negative coefficients in their estimates.

### Additional background
- Two literatures related to this work:
  - The “rentier state” and “political resource curse” literatures (Mahdavy, 1970; Beblawi and Giacomi, 1987; Chaudry, 1994; Barro, 1996; Barro, 1999; Ross, 1999; Ross, 2001).
    - Mechanisms: large oil revenues can reduce political discontent via low taxes and high public benefits, increased internal security spending, and lower education/occupational specialization leading to failure to bring about democracy (Ross, 2001, p.336/337).
  - The literature on economic growth and political instability (Haggard and Kaufman, 1997; Geddes, 1999; Burke and Leigh, 2010; Alesina and Perrotti, 1994; Londregan and Poole, 1990).
    - Mechanisms: growth reversals and economic contractions can weaken authoritarian regimes, reduce opportunity costs of protest, and increase likelihood of government change or coups.

### Data and methods

A. Data
- Main series:
  - Penn World Tables (PWT 6.3) chain weighted real GDP per capita series (log of real GDP per capita).
  - Polity IV democracy index ranging from -10 (perfect autocracies) to 10 (perfect democracies).
- Sample and coverage:
  - Sample includes 105 countries for main analyses; Figure 1 and Figure 2 show average trends 1960−2008.
  - When decomposing income into resource and non-resource components, United Nations Statistics Division National Accounts Database used; data available for 104 out of 105 countries but run only from 1970–2007.
  - Countries excluded due to short time series or no time variation in the dependent variable are listed in the Appendix (e.g., Armenia, Azerbaijan, Belarus, ...; consistent democracies or non-democracies excluded include Australia, Austria, Belgium, ... United States, Vietnam).
- Key descriptive statistics (from Table B):
  - Democracy: Obs 5085, Countries 105, Mean -1.0, Std. Dev. 6.9, Min -10.0, Max 10.0
  - Log of Real Income per Capita: Obs 5076, Countries 105, Mean 8.0, Std. Dev. 1.0, Min 5.0, Max 11.5
  - Rents to GDP: Obs 3724, Countries 98, Mean 7.7, Std. Dev. 13.5, Min 0.0, Max 116.5
  - Log of Resource Income per Capita: Obs 3869, Countries 102, Mean 5.2, Std. Dev. 1.8, Min -2.4, Max 11.1
  - Share of Resource Income: Obs 3876, Countries 102, Mean 10.0, Std. Dev. 13.9, Min 0.07, Max 77.5

B. Methods
- Estimation technique:
  - Augmented Pooled Mean Group (PMG) estimator (Pesaran et al., 1999) employed to:
    - Allow intercepts, slope coefficients and error variances to vary across panel members.
    - Allow short-run coefficients to vary across countries while restricting long-run relationships to be homogeneous.
    - Incorporate country-specific fixed effects via heterogeneous intercepts.
  - Model augmented with cross-sectional averages of dependent variable and regressors to proxy global/common factors (following Binder and Offermanns, 2007).
- Concerns addressed:
  - Time-specific heterogeneity and cross-sectional dependence: augmented PMG with cross-sectional averages used to capture common unobserved global shocks.
  - Endogeneity and reverse causality:
    - Granger causality tests (Table 1) indicate reverse causality:
      - "Democracy does not Granger cause income": Observations 4532, Lags 3, F-stat 5.47, Probability 0.001
      - "Income does not Granger cause democracy": Observations 4532, Lags 3, F-stat 6.87, Probability 0.000
    - Addressed by:
      - Augmenting PMG model with lags of regressors and dependent variable (ARDL framework); lag order chosen via AIC or SBC, constrained to a maximum of three lags.
      - Alternative heterogeneous panel cointegration techniques: group mean FMOLS (Pedroni, 2000) and DOLS (Pedroni, 2001).
- Model specification (Error Correction Model with p lags): long-run slope coefficients denoted β, η, α for income, world income, and world democracy respectively; φ is the error correction coefficient indicating speed of adjustment.

### Estimation and results
- Reproducing AJRY:
  - Extended dataset (153 countries annual; 129 countries for five- and ten-year panels; 1960–2000) reproduces AJRY: pooled OLS positive and significant income coefficient; inclusion of country and year fixed effects renders income coefficient insignificant (Table 2).
  - PMG sample (smaller) reproduces AJRY findings when using their model on that sample.
- Main PMG results (Table 3; Overall Sample N=105; 1955–2007):
  - Long-run coefficients (PMG, column 1):
    - Log income per Capita: -1.239*** (standard error 0.153)
    - World democracy: 0.800*** (standard error 0.029)
    - World output: 3.059*** (standard error 0.547)
    - Error correction coefficient: -0.264*** (standard error 0.029)
  - Interpretation examples provided in text:
    - A 10 percent increase in per capita income leads in the long run to a roughly 0.12 unit decrease on the polity scale (log-linear interpretation).
    - When global democracy score increases by one unit, democracy score improves by an average of 0.8 units.
    - A 10 percent increase in world income improves the democracy score by 0.3 units (text statement).
  - Robustness:
    - Error correction coefficient suggests about 26 percent of disequilibrium corrected in a single period.
    - Results robust to AIC vs. SBC, number of lags, cross-sectional demeaning, and exclusion/inclusion of Sub-Saharan Africa.
- Alternative estimators (Table 4; N=105; 1955–2007):
  - DOLS: Log of income per capita coefficient -2.293*** (standard error 0.189) with country and year FE = YES
  - FMOLS: Log of income per capita coefficient -0.503** (standard error 0.197) with country and year FE = YES
- Pooled error correction OLS on PMG sample (Table 5; 1960–2007; N=105):
  - Long-run coefficients across lag specifications (Lags=2 to Lags=5):
    - Log of income per capita: -1.325* (p-value=0.07), -1.236 (p-value=0.11), -1.595* (p-value=0.07), -1.683* (p-value=0.06)
    - Error correction coefficient: -0.141*** to -0.144*** across specifications
    - R-squared: 0.85
  - When the pooled ECM is estimated on the AJRY (bigger) sample, long-run coefficients on income per capita are insignificant regardless of number of lags.
- Interpretation of differences with AJRY:
  - Differences arise from:
    1. Estimation methods exploiting dynamics and heterogeneity.
    2. Sample choice excluding consistent autocracies and consistent democracies.
  - Restricting the sample to countries with time variation in income and polity allows detection of a significantly negative relationship between income and democracy.
- Sub-period check:
  - Results for sub-period 1985–2007 (not reported) maintain the negative relationship; with the PMG model global output becomes significantly negatively related to democracy.

### Digging deeper: sectoral composition and regional patterns
- Resource rents and democracy (Table 6; N=98; 1970–2007):
  - Long-run PMG coefficients:
    - Log income per capita: -0.651*** (standard error 0.157)
    - Natural resource rents to GDP: -0.021*** (standard error 0.005)
    - World democracy: 1.211*** (standard error 0.035)
    - World output: -8.554*** (standard error 1.268)
    - World rents: 0.201*** (standard error 0.038)
    - Error correction coefficient: -0.355*** (standard error 0.039)
  - Finding: natural resource rents to GDP negatively and significantly related to democracy.
- Decomposing national income (resource vs non-resource) (Table 7; N=102; 1970–2007):
  - Long-run PMG coefficients:
    - Log Income per capita (baseline): -1.228*** (standard error 0.259)
    - Log of resource GDP per capita: -0.306*** (standard error 0.081)
    - Log of non-resource GDP per capita: 0.710** (standard error 0.271) in one specification and 0.518* (standard error 0.272) in another
    - Error correction coefficient ranges: -0.305***, -0.248***, 0.245***, -0.307*** (standard errors reported)
  - Conclusion: the negative relationship between income and democracy is driven by the resource component of income; resource proportion of income per capita is negatively and significantly related to democracy, whereas non-resource income shows a positive relationship.

### Conclusion
- Re-examination of Lipset’s law using augmented PMG and heterogeneous dynamic panel methods yields:
  - For the group of countries that were neither consistent democracies nor consistent non-democracies over the sample period, there is a significant and negative long-run relationship between income and democracy.
  - The negative relationship is concentrated in resource-rich countries; in countries with little or no income from resources, the relationship between income and democracy is positive and significant.
- Methodological contributions:
  - Exploiting short-run cross-country heterogeneity in responses to income shocks and variation in income composition reveals relationships overlooked by prior work that relied on fixed effects models purging informative variation.

*Source: _wp12295 - INTRODUCTION (authors’ calculations and tables as provided in the supplied content).*

### REFERENCES

### REFERENCES

### Cited works
- Acemoglu D, Johnson S, Robinson JA, Yared P. 2008. Income and Democracy. American 
            Economic            Review            98:808-842.  
- Alesina A, Perotti R. 1994. The Political Economy of Growth: A Critical Survey of the 
            Recent            Literature.            World Bank Economic Review 8: 351–371. 
- Barro R. 1996. Democracy and Growth. Journal of Economic Growth 1: 1–27. 
- Barro R. 1999. Determinants of Democracy. Journal of Political Economy 107: 158–182. 
- Beblawi H, Giacomi L. 1987. The Rentier State. London, Croom Helm. 
- Binder M, Offermanns CJ. 2007.  International Investment Positions and Exchange Rate 
 Dynamics: a dynamic panel analysis. CFS Working Paper No. 2007/23. 
- Boix C,  Stokes S. 2003. Endogenous Democratization. World Politics 55: 517–549. 
- Burke PG, Leigh AK. 2010. Do Output Contractions Trigger Democratic Change? 
American Economic Journal: Macroeconomics 2, 124–157. 
- Chaudhry KA. 1994. Economic Liberalization and the Lineages of the Rentier State. 
 Comparative Politics 27, 1–25. 
- Coakley J, Fuertes AM, Smith R. 2006. Unobserved Heterogeneity in Panel Time Series 
            Models.            Computational Statistics & Data Analysis 50: 2361–2380. 
- Epstein D, Bates RH, Goldstone J, Kristensen I, O’Halloran S. 2006. Democratic Transitions. 
 American Journal of Political Science 50: 551-569. 
- Geddes B. 1999. What Do We Know About Democratization After Twenty Years? Annual 
 Review of Political Science 2: 115–144. 
- Gundlach E, Paldam M. 2008. Income and Democracy: A Comment on Acemoglu, Johnson, 
 Robinson, and Yared, Kiel Working Paper No. 1458. 
- Haggard S, Kaufman R. 1997. The Political Economy of Democratic Transitions. 
 Comparative Politics 29: 263–283. 
- Hausman J. 1978. Specification Tests in Econometrics. Econometrica 46: 1251–1271. 
- Im KS, Pesaran MH, Shin Y. 2003. Testing for Unit Roots in Heterogeneous Panels. 
Journal of Econometrics 115: 53–74. 
- Lipset SM. 1959. Some Social Requisites of Democracy: Economic Development and 
            Political            Legitimacy.            The American Political Science Review 53:  69–105.  
- Londregan J, Poole K. 1990. Poverty, the Coup Trap, and the Seizure of Executive Power. 
 World Politics 42: 151–183. 
- Mahdavy H. 1970. The Patterns and Problems of Economic Development in Rentier States: 
The Case of Iran. In M. A. Cook (Ed.), Studies in Economic History of the Middle 
East. London: Oxford University Press, 1970. 
- Nickell S. 1981. Biases in Dynamic Models with Fixed Effects. Econometrica 49: 1417–
            1426.            
- Pedroni, P. 2000. Fully Modified OLS for Heterogeneous Cointegrated Panels. In B.H. 
Baltagi    (Ed.), Nonstationary Panels, Cointegration in Panels and Dynamic Panels. 
Amsterdam: Elsevier. 
- Pedroni, P. 2001. Purchasing Power Parity Tests in Cointegrated Panels. Review of 
Economics and Statistics, 83: 727-731. 
- Pedroni P. 2004. Panel cointegration: aymptotic and finite sample properties of pooled time 
series tests with an application to the PPP hypothesis. Econometric Theory 20: 579–
625. 
- Pesaran MH. 1997. The Role of Economic Theory in Modelling the Long Run. The 
Economic Journal 107: 178–191. 
- Pesaran MH, Smith R. 1995. Estimating Long-run Relationships from Dynamic 
Heterogeneous Panels. Journal of Econometrics 68: 79–113. 
- Pesaran MH, Shin Y. 1999. An Autoregressive Distributed Lag Modelling Approach to 
Cointegration Analysis. In Strom, S. (Ed.), Econometrics and Economic Theory in 
the 20th Century: the Ragnar-Frisch  Centennial Symposium. Cambridge University 
Press.   
- Phillips, P.C.B., Hansen, B.E. 1990. Statistical Inference in Instrumental Variables 
Regression with I(1) Processes. Review of Economic Studies, 57: 99-125. 
- Pesaran MH, Shin Y, Smith R. 1999. Pooled Mean Group Estimation of Dynamic 
Heterogeneous Panels. Journal of the American Statistical Association 94: 621–634. 
- Phillips PCB, Moon HR. 1999. Linear Regression Limit Theory for Nonstationary Panel 
Data. Econometrica, 67: 1057–1112. 
- Przeworski A. Alvarez M.A. Cheibub J.A. Limongi F. 2000. Democracy and Development:   
Political Institutions and Well-Being in the World, 1950–1990, Cambridge 
University Press: Cambridge UK. 
- Ross ML. 1999. The Political Economy of the Resource Curse. World Politics 51: 297–322. 
- Ross ML. 2001. Does Oil Hinder Democracy? World Politics 53: 325–361. 
- Stock, J.H., Watson, M. 1993. A Simple Estimator of Cointegrating Vectors in Higher Order 
Integrated Systems. Econometrica, 614: 783-820.

*REFERENCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12295.pdf_