0.0.1 Document History

Original Publish Date: 21 July, 2020

Updated on: 10:08 PM – 03 August, 2021

In this post we further analyse the families. We categorised them into two - declining and rising. We track down all the families started in 1970s till 2007. Then we assess their performance in the elections held after 2007. This includes 3 General elections and 2 Assembly elections. If they have one at least one election in these 5 elections we call them rising dynasties and rest of them as declining dynasties.

The following table shows the difference the longevity of these two groups

1 Summary Stats

1.1 Elite types

1.1.1 Definitions

1.1.2 Break-up of elite categories

success_cat	count
Declining	90
Rising	122
Stable	110

1.1.3 turncoats and same party

election_type	success_cat	turncoat	same_party
AE	Declining	0.22	0.50
AE	Rising	0.20	0.54
AE	Stable	0.23	0.54
GE	Declining	0.18	0.62
GE	Rising	0.23	0.52
GE	Stable	0.18	0.72

1.1.4 delim change

## # A tibble: 4 x 4
## # Groups:   delim [2]
##   delim    constituency_type count total
##   <chr>    <chr>             <int> <int>
## 1 diss/new GEN                  16    18
## 2 diss/new SC                    2    18
## 3 old      GEN                 302   385
## 4 old      SC                   83   385

1.1.5 years to be a family

1.2 Viz & summary

1.2.1 A note on the life span metric

Life span is a metric that measures the years that a family or politicains have been active in politcs. Since our dynasty data data starts from 1974 and ends with politicinas who are in term till 2024, this duration of period will start from 2/3/5/6 and end at 50 years. So within this period, the maximum life span that a family can have is 50 years.

In simple words, this metric checks if a politician or a family was active ina certain year and adds up that year. We binary value is given to every year that is 1 for someone’s presence and 0 otherwise. Here we have to note that, we do not consider how many members are present at a time rather we just consider that the family was active in that particular year.

We define active in politics as being either runner-up or winner that election. Few examples:

Politician A contested in 1974, 2002 and his son contested in 2017. The life span of this family would be 15 years. See that we add a value of one for each year they are in power.
A politician and his wife contested together in 2012 and later on his wife resigned from assembly and contested in 2014 Loksabha election. In this case, this family’s life span would be 7 years

This metric also have some flaws, some of them are;

this does not diffrentiate between a winner and runner-up
this does not compensate for the years in which multiple members are cotesting together.

1.2.2 Ownership of edu categories wrt elite categories

1.2.3 Land wrt elite cat. and caste

we were using boxplot for this earlier, then we switched to box plot since the barplot was not showing any visible difference.

success_cat	mean(land, na.rm = TRUE, trim = 0.05)
Declining	259.8780
Rising	269.5455
Stable	287.3000

1.2.4 Industries

In first two tabs we look at the proportion of entities within that category who owns a specific business.

How it’s done:

In the manipulated dataset every business is coded as a binary variable. Take the case of petrol pumps(ind_c1), that entities column will be marked as 1 if they own a petrol pump, 0 otherwise.

To calculate the proportion of entities that own a specific business, we calculate the mean of that variable for all unique observations in that group.

1.2.4.1 fam v/s non-fam

Industries	% of owners among fam	% of owners among Non- fam
Petrol Pumps	0.46	0.12
Construction & Transport	0.43	0.18
Brick klin and Sand mining	0.27	0.08
Agri business	0.07	0.11
Shops and showrooms	0.05	0.03
Small Business	0.38	0.43
Unknown	0.04	0.40

1.2.4.2 Stable v/s declining

1.2.4.2.1 prev

Industries	% of owners among stable families	% of owners among declining families
Petrol Pumps	0.58	0.34
Construction & Transport	0.46	0.34
Brick klin and Sand mining	0.35	0.18
Agri business	0.06	0.04
Shops and showrooms	0.06	0.06
Small Business	0.29	0.50
Unknown	0.01	0.08

1.2.4.2.2 Invereted

family class	Petrol Pumps	Construction & Transport	Brick klin and Sand mining	Agri business	Shops and showrooms	Small Business	Unknown
Stable	0.67	0.62	0.71	0.64	0.58	0.42	0.12
Declining	0.33	0.38	0.29	0.36	0.42	0.58	0.88

1.2.4.3 Count

Industry ownership
Politician’s identity	Average number of ownersip
Declining	1.26
Rising	1.67
Stable	1.88

1.2.4.4 rent

industry ownership wrt rent extraction type
Politician’s identity	rent-type	Proportion
Declining	non-rent-thick	0.55
Declining	rent-thick	0.45
Stable	non-rent-thick	0.31
Stable	rent-thick	0.69

1.2.5 variable sumamry

Success	land	industires	N_minister	Minister duration (months)	school	college
Declining	275.67	1.26	0.43	9.37	0.27	0.24
Rising	292.21	1.67	0.65	20.07	0.33	0.33
Stable	306.64	1.88	0.72	26.73	0.46	0.43

1.2.6 political diversification wrt family class

1.2.7 political diversification wrt caste

1.2.8 Median bars

Rising category not included.

1.2.8.1 caste - pol & econ div

wrt caste groups

econ div

1.2.8.2 land - pol & econ div

1.3 Life span

1.3.1 wrt success cats

Average life span
Success	Life span
Declining	19
Rising	22
Stable	33

1.3.2 life span graphs

1.3.3 life span regression

\[{Life \ span \ of \ family} = \alpha +\beta_{1}Political \ diversification {_i} + \beta_{2} Economic \ diversification{_i} + \\ \beta_{3}Land \ ownership{_i} + \beta_{4} Years \ to \ second \ member \ entry{_i} + \\ \beta_{5} Election \ won \ by \ first\ member{_i} + Z + \epsilon \]


	Dependent variable:

	fam_life

Political diversification	1.799^***
	(0.507)

Economic diversification	1.439^***
	(0.363)

Land ownership	-0.450
	(0.391)

Years to second member entry	-0.153^**
	(0.060)

Elections won by founder	3.509^***
	(0.257)


Caste control	Yes
Decade control	Yes
Observations	322
R²	0.570
Adjusted R²	0.550
Residual Std. Error	7.260 (df = 307)

Note:	p<0.1; p<0.05; p<0.01

1.3.4 family formation

yar	prop
2009	0.07
2012	0.41
2014	0.06
2017	0.42
2019	0.05

1.4 caste

1.4.1 caste composition

1.4.2 economic diversification

caste_groups	mean	median
Upper Caste	2.00	2
Yadav	2.15	2
Non-Yadav OBC	1.96	2
Dalit	1.74	2
Muslim	1.69	2

1.4.3 Land

1.4.4 land fam

land_cat	success_cat	count	pc
small	Declining	18	0.42
small	Stable	25	0.58
medium	Declining	26	0.47
medium	Stable	29	0.53
large	Declining	29	0.42
large	Stable	40	0.58
very large	Declining	11	0.44
very large	Stable	14	0.56

1.4.5 Life span wrt caste

caste_groups	mean	median
Dalit	25	23
Muslim	25	26
Non-Yadav OBC	26	26
Others	19	20
Upper Caste	24	23
Yadav	27	26

2 ADR

2.1 assets

2.2 movable+ immovable - stacked

2.3 criminaltiy

3 Regressions

3.1 Multinom.

This is a multinomial logit model that regresses three success categories. Rising families are the reference ones. Results do not have any constituency variable here since this is regressed at family level.

3.1.1 Declining ref.

\[ \left(logit \right){Political \ family \ class} = \alpha +\beta_{1}Political \ diversification {_i} + \beta_{2} Economic \ diversification{_i} + \\ \beta_{3}Land \ ownership{_i} + Z + \epsilon \]

## # weights:  30 (18 variable)
## initial  value 353.753157 
## iter  10 value 328.883869
## iter  20 value 326.145913
## final  value 326.143264 
## converged


	Dependent variable:

	Stable	Rising
	(1)	(2)

Political diversification	0.669^***	0.490^***
	(0.189)	(0.183)

Economic diversification	0.548^***	0.284^**
	(0.142)	(0.136)

Land ownership	-0.174	-0.122
	(0.147)	(0.139)


Caste control	Yes	Yes
Akaike Inf. Crit.	688.287	688.287

Note:	p<0.1; p<0.05; p<0.01

3.1.2 Stable ref.

## # weights:  30 (18 variable)
## initial  value 353.753157 
## iter  10 value 330.002696
## iter  20 value 326.184559
## final  value 326.143264 
## converged


	Dependent variable:

	Rising	Declining
	(1)	(2)

Political diversification	-0.180	-0.669^***
	(0.153)	(0.189)

Economic diversification	-0.264^**	-0.548^***
	(0.120)	(0.142)

Land ownership	0.051	0.174
	(0.130)	(0.147)


Caste control	Yes	Yes
Akaike Inf. Crit.	688.287	688.287

Note:	p<0.1; p<0.05; p<0.01

3.1.3 Rising ref.

## # weights:  30 (18 variable)
## initial  value 353.753157 
## iter  10 value 328.686559
## iter  20 value 326.156532
## final  value 326.143264 
## converged


	Dependent variable:

	Declining	Stable
	(1)	(2)

Political diversification	-0.490^***	0.180
	(0.183)	(0.153)

Economic diversification	-0.284^**	0.264^**
	(0.136)	(0.120)

Land ownership	0.122	-0.051
	(0.139)	(0.130)


Caste control	Yes	Yes
Akaike Inf. Crit.	688.287	688.287

Note:	p<0.1; p<0.05; p<0.01

3.2 OLS

3.2.1 All family class: All families


	Dependent variable:

	declining	rising	stable
	(1)	(2)	(3)

Political diversification	-0.081^***	0.011	0.069^**
	(0.030)	(0.033)	(0.032)

Economic diversification	-0.068^***	-0.007	0.075^***
	(0.021)	(0.023)	(0.023)

Land ownership	0.023	-0.002	-0.021
	(0.023)	(0.025)	(0.024)


Caste control	Yes	Yes	Yes
Observations	322	322	322
R²	0.134	0.105	0.106
Adjusted R²	0.106	0.077	0.077
Residual Std. Error (df = 311)	0.425	0.467	0.456

Note:	p<0.1; p<0.05; p<0.01

3.2.2 Declining


	Dependent variable:

	Family decline

Political diversification	-0.124^***
	(0.039)

Economic diversification	-0.095^***
	(0.028)

Land ownership	0.012
	(0.030)

Years to second member entry	0.022^***
	(0.007)

Elections won by founder	0.007
	(0.021)


Caste control	Yes
Decadecontrol	Yes
Observations	200
R²	0.291
Adjusted R²	0.242
Residual Std. Error	0.434 (df = 186)

Note:	p<0.1; p<0.05; p<0.01

3.3 summary

Summary of family data
Measure	Elite catagory	Political diversification	Minister duration(Months)	Economic diversification	Land ownership(Deciles)	Years for second member
Length	Declining	90.00	90.00	90.00	90.00	90.00
Length	Rising	122.00	122.00	122.00	122.00	122.00
Length	Stable	110.00	110.00	110.00	110.00	110.00
Unique-N	Declining	2.00	25.00	4.00	5.00	18.00
Unique-N	Rising	2.00	32.00	4.00	5.00	25.00
Unique-N	Stable	2.00	31.00	4.00	5.00	15.00
Mean	Declining	0.04	4.49	1.26	3.24	7.21
Mean	Rising	0.18	9.93	1.67	3.25	10.53
Mean	Stable	0.17	10.53	1.88	3.35	5.35
SD	Declining	0.21	10.55	1.01	1.18	5.18
SD	Rising	0.39	17.72	1.02	1.18	9.88
SD	Stable	0.38	17.90	0.92	1.06	4.19
Min	Declining	0.00	0.00	0.00	1.00	0.00
Min	Rising	0.00	0.00	0.00	1.00	0.00
Min	Stable	0.00	0.00	0.00	1.00	0.00
Max	Declining	1.00	54.00	3.00	5.00	25.00
Max	Rising	1.00	78.50	3.00	5.00	43.00
Max	Stable	1.00	86.00	3.00	5.00	22.00

–>

–> –> –> –>

–> –>

–>

4 survival analysis 00

You might note that there are only 200 families in the survival analysis. This is because of the fact that, a main proponent of the survival analysis the variable dead which tells you which all families still alive or not.

In our analysis, we have decided to consider families that are not active for last 10 year (2009-2019) as no longer alive. Hence we can only include families that have at least entered before 2009. As you would know, a huge chunk of entities only formed families post delimitation, so that leaves out ~122 families from the survival analysis.

Some might be sceptical about that ten years in particular. We show that it is unfound by showing that all variables are still significant even after varying the censoring years in section censoring sesitivity

4.1 Base survival plots

4.1.1 all entities

4.1.2 won

4.1.3 all fam

4.1.4 after became fam

4.1.5 Econ diversification grid

combination of education and industry

4.1.6 pol diversification grid

4.2 coax -proportional hazard regression

**Coax hazard regression**

	Dependent variable:

	Family decline

Political diversification	-0.608^***
	(0.202)

Economic diversification	-0.454^***
	(0.105)

Land ownership	0.047^**
	(0.022)

Years to second member entry	0.051
	(0.099)

Elections won by founder	-0.306^***
	(0.087)


Caste control	Yes
Observations	200
R²	0.326
Max. Possible R²	0.988
Log Likelihood	-399.137
Wald Test	64.500^*** (df = 11)
LR Test	78.845^*** (df = 11)
Score (Logrank) Test	71.808^*** (df = 11)

Note:	p<0.1; p<0.05; p<0.01

4.3 year entry sensitivity

Here we are running models on families a three different models for families that are started in those particular period in last 70 years. Aim of these models is to show that variables are still relevant irrespective of the families commencenmt year and age

**Coax hazard regression**

	Dependent variable:

	Family decline
	(1)	(2)	(3)

Political diversification	-0.646^***	-0.636^*	-2.300^***
	(0.206)	(0.350)	(0.804)

Economic diversification	-0.476^***	-0.654^***	-1.357^***
	(0.103)	(0.198)	(0.526)

Land ownership	-0.003	-0.045	-0.076
	(0.018)	(0.044)	(0.099)

Years to second member entry	0.111	0.098	0.504^*
	(0.099)	(0.181)	(0.294)

Elections won by founder	-0.463^***	-0.515^***	-1.210^***
	(0.086)	(0.155)	(0.415)


Start	All	Post 80	Post 90
Caste control	Yes	Yes	Yes
Decade control	Yes	Yes	Yes
Observations	322	188	121
R²	0.280	0.230	0.266
Max. Possible R²	0.947	0.816	0.650
Log Likelihood	-421.184	-134.520	-44.914
Wald Test (df = 12)	82.490^***	42.230^***	21.580^**
LR Test (df = 12)	105.636^***	49.109^***	37.370^***
Score (Logrank) Test (df = 12)	92.773^***	45.822^***	26.884^***

Note:	p<0.1; p<0.05; p<0.01

4.4 Censoring sensitivity

In this section we provide three different cut off to assess whether a family is dead or not. for eg: If a family is not has to been contesting in last x years we would consider them as dead. The number of families included in the model also change according to the cut-off because we can only include the family that has formed on or before the cut off years. The break up looks like this:

Cut off	5 years (2014)	10 years (2009)	15 years (2004)	20 years (1999)
alive	128	110	91 \|	88
dead	130	90	65 \|	31
total	258	200	156 \|	119

**Coax hazard regression**

	DV: Family decline

	(1)	(2)	(3)	(4)

Political diversification	-0.553^***	-0.332^**	-0.621^**	-0.655^*
	(0.196)	(0.152)	(0.250)	(0.385)

Economic diversification	-0.452^***	-0.354^***	-0.444^***	-0.637^***
	(0.103)	(0.084)	(0.123)	(0.184)

Land ownership	0.037^*	0.051^***	0.021	-0.066
	(0.022)	(0.016)	(0.026)	(0.048)

Years to second member entry	0.052	0.037	0.072	0.152
	(0.097)	(0.082)	(0.117)	(0.159)

Elections won by founder	-0.318^***	-0.344^***	-0.297^***	-0.258
	(0.086)	(0.068)	(0.111)	(0.188)


Cut-off years	2009	2014	2004	1999
Caste control	Yes	Yes	Yes	Yes
Decade control	Yes	Yes	Yes	Yes
Observations	200	258	156	119
R²	0.331	0.318	0.297	0.282
Max. Possible R²	0.989	0.995	0.980	0.910
Log Likelihood	-412.742	-645.460	-279.591	-123.745
Wald Test (df = 12)	66.660^***	82.000^***	46.030^***	25.150^**
LR Test (df = 12)	80.492^***	98.618^***	55.036^***	39.467^***
Score (Logrank) Test (df = 12)	73.470^***	90.355^***	51.156^***	31.560^***

Note:	p<0.1; p<0.05; p<0.01

4.5 Econ and pol diversification grid

life span variable in both plots are the ones that arecounted only after they ebcame a family.

Families

2020-02-07