Empirical Exercise 13: Lobbyists

question, hypothesis and theory

1. the effect of political connections on influencing revenues

   1. Does a lobbyist’s revenue decrease when the politician they are connected to leaves office?

2. Hypothesis:  lobbyists who have more political connections will earn more revenue, as there are more people they can influence, more votes they can get, more decision making power

3. Theory: 

   1. Revolving door lobbyists are valuable as ex staffers generate rev for their firms by trading on their political connections

      1. Can exploit former experience in gov to gain desired policy outcomes of clients

   2. Looking at the impact of a serving politician’s exit on the lobbying revenues of her former staffers: when politician no longer serving in congress, political connection no longer has legislative influence

lobbyists, ex staffers definitions

1. professional paid advocates who aim to influence the decisions of legislators or other government officials

   1. Revolving door: lobbying firms attract individuals leaving gov positions with high paying offers and gain access to important politicians in office

      1. Can receive favorable legislation and to receive inside information on what is happening in government

      2. Focusing on ex staffers particularly : employed in offices of member of congress eg senator or house of representatives member

type of dataset, period of time, variables

1. -panel information on 

   • Revenue generated 

   • Whether connection to Senate, House, Committees, etc. 

   • Whether they lose connection to serving politicians

2. -panel of 1113 lobbyist and follow them over 11 years

   -balanced panel: observe every lobbyist in every 6 month period so would have 1112 x 22= 24464 observations but we only have 10418 so it is unbalanced panel

   -need 1112 dummies for 1113 observations so tell them to absorb lobbyist so tell them to estimate all fixed effects but doesn’t show all dummies in output 

   -I.period creates dummy variable for each 22 periods, leave out 1 for baseline group

   -need to cluster standard errors as error terms are correlated over time for each individual lobbyist as if they are earning high rev today, they will probably earn high rev in future

3. Main dependent variable is (log) revenue per lobbyist

   1. total value of all contracts you work on regardless whether you work on [unweighted measure] and focus on total value of contracts weighted by how much each lobbyist contributes [weighted measure]

   2. need to put in logs as all the values will be disproportionately affected as there are many outliers as lobbyists may earn average and others may earn way above 

4. Main independent variable of interest is connection

   1.  Currently serving politicians, but will leave within sample period, a lobbyist is linked to through his previous employment experience

   2. Variable goes from 2 to 1: politician lobbyist used to work for left office, connection is no longer 

controlling for individual fixed effects, is gaining connection exogenous

-remove everything from error term that is same for lobbyists but doesn’t change over time so expertise will be removed

-allows us to study the effect of only connections on revenue

-cannot gain any more connections as a lobbyist as you don’t work in politics anymore, but you can lose connections over time eg if a politician you knew loses in an re-election, retirement

-can gain a political connection due to expertise but cannot lose a connection over expertise of the lobbyist who used to work for them: gaining connections is endogenous, correlated with error term
-so connections you lose over time should not be correlated with expertise or anything else in error term: exogenous if you lose a connection

baseline population model

-individual lobbyist fixed effect: alpha i

-semester/time fixed effects: lambda t

-Pit: number of connections to currently serving officials
-while connection still in office Pit -1, at some point ex-boss will leave so Pit becomes 0

-use a fixed effects model as it removes expertise for error term [areg in STATA: absorb individual fixed effects for lobbyists: id_lobbyist vs normal reg]

-absorb and areg tells you there are individual fixed effects: alwys use areg for panel data as we want to include fixed effects

-42% of lobbyists are revolving door 

-3% are ex-congressman

-22% are ex-staffers: used to work in office of congressmen

-17% used to work in other parts of US gov 

-lobbying is a superstar industry as small amount of individuals earn more than 1 million, while majority earn much less 

-average rev in industry is 300k 

-ex congressman make 340k

-ex staffers make 350k

-revolving door lobbyists who used to work outside congress is 250k

-non revolving door lobbyists only make 170k

• Less successful 

• Lobbyists who work in politics in the past will earn more rev than lobbyists who have not worked in politics before 

-Being connected to a Representative at some point in your political career seems to be stronger than having been connected to a Senator

- are both statistically insignificant at the 10% 

-if we were to make an F-test of whether they are statistically different from each other, we would fail to reject the null hypothesis that they are not

would using fixed effects work with everhouse/eversenate

-don’t include individual fixed effects as dummies like ever connected to senate and house, don’t change over time, only take value 0 and 1

• Once you are connected that does not change over time so this will be absorbed by individual fixed effect: perfectly collinear, will be dropped from the regression

• everhouse and eversenate: are 0 as they are omitted 

• Controlling for fixed effects [absorb id_lobbyist] won't allow us to separate the effect of connection [everhouse/eversenate] which don’t change over time from the effect of other individual fixed effects like expertise as they also don’t change over time 

empirical model

 -variable being connected to senator has coefficient 0.23, significant at 5% level: 23% higher revenue being connected to current senator 

-every time you lose a connection to senators ie senate_exit goes down by 1 unit, revenue goes down by 23% vs periods when you would still have that connection

-variable being connected to current representative has coefficient 0.08, not significant: no significant effect of connections to house representative on revenue

-senate_exit: connections to senators who will exit at some point in future, interpret as percentage increase

-connections will only go down over time

-house_exit: connections to congressman currently in house 

-controlling for individual fixed effects here allows us to separate out the effect of connection (senate_exit/house_exit: which varies over time) from the effect of expertise (which is included in the individual fixed effects since it is time-invariant)

-these variables are 0 when the lobbyist is working for senate/house and switch to 1 when they leave: time variant

political power effect on revenue: population model, variables

-measuring whether how powerful politician you know will affect revenue

-connection to a lot of power should generate a lot of rev/ connection to little power should generate little revenue

-senators are more powerful than representatives 

-politicians in important committees are very powerful eg

• Appropriations 

• Ways and Means 

• Finance

connections to senators on finance committee

-connections to senators sitting in appropriation

-not finance or appropriation is: cmte_not_fa_s

-connections to house of representatives sitting on ways and means committee

-connections to house of representatives sitting on appropriations committee

-not ways and means or appropriation is: cmte_not_wma_h

-connected to finance increases rev by 36% and appropriations committee in senate increases rev by 45%, not connected to politicians is not significant so rev doesn’t change

-suffer losses of 36/45% in rev when those senators leave office

-connected to ways/means increases rev by 35% and connected to appropriations committee in house increases rev by 6% 

-suffer losses of 35% in rev when those senators leave office

-politicians in neither committees do not affect affiliated lobbyists when they leave: p values are insignificant

-connected to finance increases rev by 36% and appropriations committee in senate increases rev by 45%, not connected to politicians is not significant so rev doesn’t change

-suffer losses of 36/45% in rev when those senators leave office

-connected to ways/means increases rev by 35% and connected to appropriations committee in house increases rev by 6% 

-suffer losses of 35% in rev when those senators leave office

-politicians in neither committees do not affect affiliated lobbyists when they leave: p values are insignificant

why do we need to control for other variables

-after controlling for lobbyist fixed effects, need to make sure no third variable correlated with lobbyist revenue and is correlated with connection with currently serving politician

-party-time dummies allows us to control for different time effects for lobbyists connected to politicians in different parties (i.e. Democrats versus Republicans)

-controlling for party-chamber-time allow for different time effects for lobbyists connected to politicians in different party/chamber combinations 

controlling for how badly a party does

-variable like party is doing badly in error term could be something linked to lower revenue so could cause OVB

-number of connections could go down as politicians will retire and will be losing connections

-if it is a good year for democrats, republican is more likely to leave office as they will defeated in re-election

-when democrats are doing well, republicans will defeated and if the lobbyist is republican, it will be harder for them to influence government, leading to loss in revenue

-if it is a good year for democrats, democrat ex-boss connection is more likely to stay in office as they will win in re-election

-when democrats are doing well, democrats will do better and if the lobbyist is democrat, it will be easier for them to influence government, leading to more revenue

empirical model for controlling for party + time fixed effects

-need more time period fixed effects: effects for republicans and time effects for democrats vs one time period dummy like we had before

-now have 44 time fixed effects as i.period x which party you worked for [22 for democrats, 22 for republicans]

-connected to senate increases revenue by 22% and is significant, connected to house representative is not significant effect on revenue: same result as before controlling for party fixed effects

party chamber time: interacting with senate/house

-can also have party x chamber x time fixed effect as how badly party does in chamber ie specifically in the house or senate so earning low rev and lose connections in only that on chamber and not the other 

-connected to house of representatives or senate or both for each political group: 6 fixed effects

-22 x 6 =132 fixed effects 

-connected to senate increases revenue by 21% and is significant, connected to house representative has no significant effect on revenue: same result

controlling for experience of lobbyist

-more connections if you have just entered lobbying: so OVB also caused by experience of lobbyist

-if you have been working as lobbyists for longer, you have lost more connections

-so still have time x party x chamber fixed effects but also controlling for experience

-connected to senate increases revenue by 24% [slightly higher than previous] and is significant

-connected to house representative increase revenue by 10%, p value significant at 10% level so minorly significant effect on revenue 

-smaller increase in revenue for every house politician you work for vs senate

 
-it is experience squared because loads of experience means revenue goes down when you get older , revenue goes down at very high levels of experience 

conclusion of empirical study

-connections to serving politicians are an important cause of revenue generation for lobbyists 

-better connections generate more revenue

-yes having more connections allows lobbyists to earn more money, but only if they are high quality politicians who sit on budget committees 
- or conversely drop in revenue is strongest when departed politician was serving on senate finance/appropriations or house ways and means

internal validity

• Panel data with lobbyist fixed effects: Controls for time-invariant individual characteristics (e.g., innate skill).

• Exogenous variation in politician exits: Many exits (e.g., retirement, death, scandals) are not influenced by lobbyists and thus help identify causal effects.

• Use of time-varying connection indicators: Exploits within-lobbyist variation over time.

• Controls for experience, time, and party effects to reduce omitted variable bias.

external validity

Focused on U.S. federal lobbyists from 1998–2008: results may not generalize to:

• Other countries with different lobbying systems

• State-level or local lobbying

• More recent political dynamics (e.g., post-2016 polarization)