Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Historical: Terms should end on January 3rd #7

Closed
konklone opened this issue Jan 1, 2013 · 37 comments
Closed

Historical: Terms should end on January 3rd #7

konklone opened this issue Jan 1, 2013 · 37 comments

Comments

@konklone
Copy link
Member

konklone commented Jan 1, 2013

Instead of December 31st. Terms both start and end on that day, since it ticks over at noon.

@JoshData
Copy link
Member

JoshData commented Jan 1, 2013

In the historical data, I think end dates are set to the date of adjournment of the House or Senate, whichever is later. It follows the session end dates in my data/us/sessions.tsv file. In the 112th, it was the best guess. The last adjournment is usually before the end of the year.

I see the benefit to following the Constitution-set dates (n.b. it wasn't always Jan 3), but I'm wary of people having two terms on the same day. I think that will foul people up. Unless we make them datetimes.

@dwillis
Copy link
Member

dwillis commented Jan 1, 2013

I think as long as the congress is set properly, then having people with two terms on the same day isn't a problem.

@JoshData
Copy link
Member

JoshData commented Jan 1, 2013

There's no congress field on terms right now. It's entirely by dates. But we could add that.

@dwillis
Copy link
Member

dwillis commented Jan 1, 2013

Oh, right. I think it makes sense to add it - that's what we do. It's a little bit redundant, I guess, but better to be specific.

@konklone
Copy link
Member Author

konklone commented Jan 2, 2013

Yeah, adding a congress field on terms makes sense, to distinguish them, though I think you could construct logic that properly guessed at congresses for each terms, even if we didn't add it and had two terms overlap on the same day. We could maybe also do datetimes for them, though I don't know where we'd get swearing in times for specially elected folks.

I think it's probably more correct to use the Constitution-set dates, because that person is still in power even when Congress is adjourned. Constituent services come to mind as a piece of legislative power that's orthogonal to adjournment.

@JoshData
Copy link
Member

JoshData commented Jan 2, 2013

Ok I'm down with it w/ a congress field. (A congress field would be helpful for me for other reasons anyway.)

JoshData pushed a commit that referenced this issue Jan 3, 2013
…,3) to be on Jan 3 rather than Dec 31, matching the end dates for the newly elected Members and addressing issue #7
@JoshData
Copy link
Member

JoshData commented Jan 3, 2013

For current legislators as of 113th Congress, fixed in 1f9bfdd. Do we want to revise in the historical data?

@konklone
Copy link
Member Author

konklone commented Jan 3, 2013

For everything from 1934 onwards? The amendment establishing January 3rd as
the date was ratified in late January, 1933:

http://en.wikipedia.org/wiki/Twentieth_Amendment_to_the_United_States_Constitution

So this would affect the ending date of the 1933-1934 term (placing it in
on 1935-01-03) and everything since.

-- Eric

On Wed, Jan 2, 2013 at 7:14 PM, Joshua Tauberer [email protected]:

For current legislators as of 113th Congress, fixed in 1f9bfddhttps://github.com/unitedstates/congress-legislators/commit/1f9bfdd.
Do we want to revise in the historical data?


Reply to this email directly or view it on GitHubhttps://github.com//issues/7#issuecomment-11830414.

Developer | sunlightfoundation.com

@konklone
Copy link
Member Author

konklone commented Jan 3, 2013

Also, thanks for taking care of this!

-- Eric
On Jan 2, 2013 7:29 PM, "Eric Mill" [email protected] wrote:

For everything from 1934 onwards? The amendment establishing January 3rd
as the date was ratified in late January, 1933:

http://en.wikipedia.org/wiki/Twentieth_Amendment_to_the_United_States_Constitution

So this would affect the ending date of the 1933-1934 term (placing it in
on 1935-01-03) and everything since.

-- Eric

On Wed, Jan 2, 2013 at 7:14 PM, Joshua Tauberer [email protected]:

For current legislators as of 113th Congress, fixed in 1f9bfddhttps://github.com/unitedstates/congress-legislators/commit/1f9bfdd.
Do we want to revise in the historical data?


Reply to this email directly or view it on GitHubhttps://github.com//issues/7#issuecomment-11830414.

Developer | sunlightfoundation.com

@GPHemsley
Copy link
Member

I think the terms for everyone serving from 1933 onward should be updated. (That's the middle of the 73rd Congress plus every full Congress starting with the 74th.) Per the 20th Amendment, all terms are from January 3rd to January 3rd except in member-specific situations.

It is arguable what the terms were prior to 1933, but I believe the BioGuide suggests March 4 to March 3. Wikipedia has had a debate or two as to whether that is indeed the case, or whether it should instead be March 4 to March 4. (I suppose the debate is whether the term ends at noon or at midnight.)

In all cases, I don't think it makes sense to list the Congress session times as the term delineations, because the people were members of Congress even when Congress was not in session. (If the President called an emergency session of Congress, who would show up? The people who were then serving terms as members of Congress.)

@konklone
Copy link
Member Author

konklone commented Jan 7, 2013

Before 1933, it was way more chaotic than that. Take a look at the first 86 years of the House and Senate:
http://memory.loc.gov/ammem/amlaw/lwsjlink.html

People in these sessions were elected the November before, sometimes started a session the March after, but sometimes not until December - 13 months after their election date. I honestly am not certain whether or not they were sworn in on March with regularity, which could affect it, but I also don't know how we would tell.

I think all of us are in agreement on updating the term end dates from 1933 onwards, but I don't know that we've reached consensus on whether the start date should always be the 3rd, or whether it should be the day of swearing in. (I also don't know where we'd get the data to delineate the two.)

I'm fine with only dealing with the end date for now - it was what just bit us, and can make a practical difference in deciding what happened on January 1 or 2 of a year. The start date choice makes a lot less practical difference, if any.

@JoshData
Copy link
Member

JoshData commented Jan 7, 2013

Let's put the question of start dates into a separate issue and leave this open for fixing the end dates in historical data.

These date changes create an enormous headache for me when loading the data into GovTrack.

@GPHemsley
Copy link
Member

Where does this stand?

I think it would make sense that any person continuously serving in Congress should have term dates that are continuous. The start date for their first term and the end date for their last term might be debatable, though.

@konklone
Copy link
Member Author

@bchartoff has done some work in #131 (now closed) relevant to closing this issue.

@bchartoff
Copy link
Contributor

I'd like to address this issue. Talking with @konklone, we think the following makes sense:

  • Add a "congress" field for all terms in current and historical data
  • Update post 1933 end dates to Jan 3rd (excluding deaths/retirements/party changes)
  • Leave start dates and pre-1933 dates untouched.
    @tauberer , would this end up being a huge headache in re: govtrack? Additionally, I see merit in including your sessions.tsv file in the repo. Thoughts?

@JoshData
Copy link
Member

Because of senators it'll have to be either "congresses" or (perhaps better) "starting_congress"?

The rest sounds good.

@GPHemsley
Copy link
Member

Is there a reason why continuing members (post-1933) can't have their start dates updated to January 3rd? Re-swearing-in doesn't really matter for those already serving, does it?

@konklone
Copy link
Member Author

I didn't think there was an issue with start dates, but if they're not reliably on January 3rd, they should be updated to be so.

I like the idea of just putting the starting Congress for senators' terms a lot. We can just call it "congress" though, and document that in the context of Senators' terms, it's their starting term. Combined with the senate_class field, that's enough to programmatically understand the scope of their term.

@GPHemsley
Copy link
Member

Out of curiosity, what is the benefit of having only a single Congress for a Senator's term?

@JoshData
Copy link
Member

On start dates: Everyone is re-sworn in every Congress (in the Senate too?). If they miss the swearing in, they're not congressmen yet (in some respects at least). So I think we should keep start dates as they are, which is the best guess of the swearing in dates.

Listing only a single Congress will look nicer in the YAML and will be easier to maintain.

And just to clarify, for senators appointed mid-term, their starting congress would be the congress during which they were appointed, not the starting congress of the term they're filling, right? (I'm not sure it matters.)

@konklone
Copy link
Member Author

Agreed on start dates, @tauberer, that sounds reasonable. And yeah, the single Congress field will be lots easier to maintain and work with.

I guess for Senators appointed mid-term, their starting congress should be the congress they were appointed; otherwise we'd be nonsensically saying the person was a Senator in a congress they weren't in. But it does mean the senate_class can't be used the way I described, it'd have to depend on dates. But that's fine.

@JoshData
Copy link
Member

If my class is X such that I can only have been elected in congresses 1, 4, 7, and my "start" congress is 3, then I must be filling the original 1,2,3 term (except in the 1790s). Slightly more cumbersome, but possible.

@konklone
Copy link
Member Author

Good point.

@GPHemsley
Copy link
Member

Ah, I see your point about the term start dates. (I keep neglecting the fact that Congressional terms are not as continuous as presidential terms.)

But it seems to me that, given the number of exceptions to such a calculation for a Senate term's Congress (short terms in the 1790s, filling a vacancy during the same Congress, filling a vacancy during a subsequent Congress, etc.), it would be easier to just include all 3 Congresses with a Senator's term. Otherwise, the addition of the field in the first place doesn't get you much (IMO).

@JoshData
Copy link
Member

The value for me is that it's sort of a primary key (in combination with other fields). I just need something stable to tie the records to before the date changes (because I use the dates now as a sort of primary key).

@konklone
Copy link
Member Author

I agree with you (@GPHemsley) that the edge cases around Senate terms are annoying. But I think even a single scalar congress field does get you a lot, it handles most common questions. For precision, you always need to rely on the dates.

But @tauberer, could you just calculate the congress on the fly using the rules you described above, and use that as your primary key, regardless of whether it appears in this YAML? After all, this would just be a convenience field, always derivable from the dates we keep.

@bchartoff
Copy link
Contributor

I'm a bit torn, but I think I'd either argue to:
A. not add any congress or congresses field, but update the functions in utils.py to more accurately return a list of congresses based on start and end dates (or tie it more concretely to the data and pass it a legislator/ term number?) Right now the function is great for timestamped data, but since our dates have no timestamps, it seems a bit out of place?

B. add an array of congresses to each legislator, and update the appropriate scripts (which update start/ end dates and terms) to include congress arrays.

The advantage I see to A is that it removes duplication of data in the yamls and eases future maintenance. If we're going to add congress terms, I'd rather go with all of them -- as @konklone says, it's really a convenience field and, as a frequent user of this data, any application which I can think of which needs congress #s would need arrays rather than a starting_congress field. Sure, I could derive the list, but if I'm going to derive it I might as well derive from the start/end dates (which I've always done w/ minimal problems). And, perhaps more importantly, less experienced/ savy users of the data might not be able to derive this term themselves. @tauberer, could you use the first item of the array + class as a pk?

If I implement B, I'd be happy to also implement A.

@JoshData
Copy link
Member

I think we're going in circles here.

@konklone
Copy link
Member Author

All of these data refactoring threads end up with us stalemating in circles! :) @bchartoff, feel free to break the stalemate with working code, and we can move from there.

@GPHemsley
Copy link
Member

When worrying about primary keys and whatnot, remember to be able to handle cases (such as Strom Thurmond in 1956) where an individual is appointed to continue their own term in the same Congress.

@konklone
Copy link
Member Author

Why is Congress so crazy??

@JoshData
Copy link
Member

JoshData commented Sep 8, 2013

Just for curiosity/reference, here's an old joint resolution setting the start date for the subsequent Congress:
http://www.gpo.gov/fdsys/granule/STATUTE-84/STATUTE-84-Pg1880-2/content-detail.html

@JoshData
Copy link
Member

I was just looking this over again. Our data is really bad for any Senate appointments. Those terms seemed to all have gotten squashed with the subsequent term. I either messed this up in the original import, or maybe it was originally incomplete in the Bioguide.

We need to re-parse bioguide to find all of the mistakes. Fixing by hand will take forever. Meh.

@konklone
Copy link
Member Author

Well, I've at least got a bioguide walker/parser already, and @wilson428 extended it once for other purposes...

JoshData added a commit that referenced this issue Aug 4, 2015
…o fix term end dates to be the Constitutionally-defined end of a term.

In summary, this commit does a simple string substitution to correct the following dates:

	  2nd Congress 1793-03-02 => 1793-03-03
	 22nd Congress 1833-03-02 => 1833-03-03
	 73rd Congress 1934-06-18 => 1935-01-03
	 74th Congress 1936-06-20 => 1937-01-03
	 75th Congress 1938-06-16 => 1939-01-03
	 77th Congress 1942-12-16 => 1943-01-03
	 78th Congress 1944-12-19 => 1945-01-03
	 79th Congress 1946-08-02 => 1947-01-03
	 80th Congress 1948-12-31 => 1949-01-03
	 81st Congress 1951-01-02 => 1951-01-03
	 82nd Congress 1952-07-07 => 1953-01-03
	 83rd Congress 1954-12-02 => 1955-01-03
	 84th Congress 1956-07-27 => 1957-01-03
	 85th Congress 1958-08-24 => 1959-01-03
	 86th Congress 1960-09-01 => 1961-01-03
	 87th Congress 1962-10-13 => 1963-01-03
	 88th Congress 1964-10-03 => 1965-01-03
	 89th Congress 1966-10-22 => 1967-01-03
	 90th Congress 1968-10-14 => 1969-01-03
	 91st Congress 1971-01-02 => 1971-01-03
	 92nd Congress 1972-10-18 => 1973-01-03
	 93rd Congress 1974-12-20 => 1975-01-03
	 94th Congress 1976-10-01 => 1977-01-03
	 95th Congress 1978-10-15 => 1979-01-03
	 96th Congress 1980-12-16 => 1981-01-03
	 97th Congress 1982-12-23 => 1983-01-03
	 98th Congress 1984-10-12 => 1985-01-03
	 99th Congress 1986-10-18 => 1987-01-03
	100th Congress 1988-10-22 => 1989-01-03
	101th Congress 1990-10-28 => 1991-01-03
	102th Congress 1992-10-09 => 1993-01-03
	103th Congress 1994-12-01 => 1995-01-03
	104th Congress 1996-10-04 => 1997-01-03
	105th Congress 1998-12-19 => 1999-01-03
	106th Congress 2000-12-15 => 2001-01-03
	107th Congress 2002-11-22 => 2003-01-03
	108th Congress 2004-12-09 => 2005-01-03
	109th Congress 2006-12-09 => 2007-01-03
	111th Congress 2010-12-22 => 2011-01-03

When I first created this dataset back in 2009, I used the sine die adjournment date of Congress as term end dates (as best as I could quickly identify those dates, and the later of the House or Senate). It has been a bit of a research project to figure out the actual term end dates, especially prior to 1933. Constitutional term end dates are more appropriate since a Member is a Member, and can come back into session, until the term actually ends, not when Congress last meets. In addition, I have found that historical roll call data from Voteview and historical legislative data from THOMAS/Congress.gov indicates there was legislative activity beyond the end dates that I have used --- so they may be incorrect anyway.

(Note that this doesn't address term start dates, which we've been storing as swearing-in dates, where known.)

According to the [Constitution Annotated, Amendment 20](https://www.congress.gov/content/conan/pdf/GPO-CONAN-REV-2014-10-21.pdf), quoting the Senate report on the Twentieth Amendment (1 S. Rep. No. 26, 72d Cong., 1st Sess. 2, 4, 5, 6 (1932)):

> The commencement of the terms of the first President and Vice President and of Senators and Representatives
composing the First Congress was fixed by an act of [the Continental] Congress adopted September 13, 1788, and that act provided ‘that the first Wednesday in March next to be the time for commencing proceedings under the Constitution.’ It happened that the first Wednesday in March was the 4th day of March, and hence the terms of the President and Vice President and Members of Congress began on the 4th day of March.

This explains the frequency of March 3 as adjournment dates in our term end data through 1933 (72nd Congress), with only anomalies for the 2nd Congress (1793-03-02), 22nd Congress (1833-03-02), and 69th Congress (1927-03-04). I've verified that the 69th Congress did end on a March 4, and per further analysis in the Constitution Annotated (see below), it seems terms ended at some time on March 4. Nevertheless, I've left the March 3 end dates where they are (a total of 69 Congresses had that date), and I've left the 69th Congress's March 4 date alone. I've corrected the 2nd Congress and 22nd Congresses to March 3's (for consistency with the other Congresses in that time period).

The [20th Amendment](http://www.archives.gov/exhibits/charters/constitution_amendments_11-27.html) changed the start and end dates of legislative (and executive) terms. The amendment was ratified on January 23, 1933. Per section 5 of the amendment, section 1 takes effect "on the 15th day of October following the ratification of this article", i.e. October 15, 1933. Section 1 states that legislative terms start and end "at noon on the 3d day of January, of the years in which such terms would have ended if this article had not been ratified." The 73rd Congress was in session. We have its adjournment date as 1934-06-18. But its actual end date was changed by the 20th Amendment from 1935-03-03 (or -04) to 1935-01-03 at noon. So this commit changes the term end date from 1934-06-18 to 1935-01-03.

The Constitution Annotated clarifies:

> [The 20th Amendment] shortened, by the intervals between January 3 and March 4, the terms of Senators elected for terms ending March 4, 1935, 1937, and 1939 . . . It also shortened the terms of Representatives elected to the Seventy-third Congress, by the interval between January 3 and March 4, 1935.

Only a handful of Congresses since then have actually adjourned on January 3 (76th, 110th, 112th Congresses), so most of the term end dates starting with the 73rd Congress are updated by this commit to January 3.

Beginning with the 113th Congress we've been using January 3, so the last Congress updated is the 11th Congress.

I made these changes using `sed`. Anywhere the date occurs in an `end: ` is assumed to be the end of the term at the end of a Congress. Any ends of terms on other dates not listed above are left alone because they're resignations, deaths, etc.
JoshData added a commit that referenced this issue Aug 4, 2015
…o fix term end dates to be the Constitutionally-defined end of a term.

In summary, this commit does a simple string substitution to correct the following dates:

	  2nd Congress 1793-03-02 => 1793-03-03
	 22nd Congress 1833-03-02 => 1833-03-03
	 73rd Congress 1934-06-18 => 1935-01-03
	 74th Congress 1936-06-20 => 1937-01-03
	 75th Congress 1938-06-16 => 1939-01-03
	 77th Congress 1942-12-16 => 1943-01-03
	 78th Congress 1944-12-19 => 1945-01-03
	 79th Congress 1946-08-02 => 1947-01-03
	 80th Congress 1948-12-31 => 1949-01-03
	 81st Congress 1951-01-02 => 1951-01-03
	 82nd Congress 1952-07-07 => 1953-01-03
	 83rd Congress 1954-12-02 => 1955-01-03
	 84th Congress 1956-07-27 => 1957-01-03
	 85th Congress 1958-08-24 => 1959-01-03
	 86th Congress 1960-09-01 => 1961-01-03
	 87th Congress 1962-10-13 => 1963-01-03
	 88th Congress 1964-10-03 => 1965-01-03
	 89th Congress 1966-10-22 => 1967-01-03
	 90th Congress 1968-10-14 => 1969-01-03
	 91st Congress 1971-01-02 => 1971-01-03
	 92nd Congress 1972-10-18 => 1973-01-03
	 93rd Congress 1974-12-20 => 1975-01-03
	 94th Congress 1976-10-01 => 1977-01-03
	 95th Congress 1978-10-15 => 1979-01-03
	 96th Congress 1980-12-16 => 1981-01-03
	 97th Congress 1982-12-23 => 1983-01-03
	 98th Congress 1984-10-12 => 1985-01-03
	 99th Congress 1986-10-18 => 1987-01-03
	100th Congress 1988-10-22 => 1989-01-03
	101th Congress 1990-10-28 => 1991-01-03
	102th Congress 1992-10-09 => 1993-01-03
	103th Congress 1994-12-01 => 1995-01-03
	104th Congress 1996-10-04 => 1997-01-03
	105th Congress 1998-12-19 => 1999-01-03
	106th Congress 2000-12-15 => 2001-01-03
	107th Congress 2002-11-22 => 2003-01-03
	108th Congress 2004-12-09 => 2005-01-03
	109th Congress 2006-12-09 => 2007-01-03
	111th Congress 2010-12-22 => 2011-01-03

When I first created this dataset back in 2009, I used the sine die adjournment date of Congress as term end dates (as best as I could quickly identify those dates, and the later of the House or Senate). It has been a bit of a research project to figure out the actual term end dates, especially prior to 1933. Constitutional term end dates are more appropriate since a Member is a Member, and can come back into session, until the term actually ends, not when Congress last meets. In addition, I have found that historical roll call data from Voteview and historical legislative data from THOMAS/Congress.gov indicates there was legislative activity beyond the end dates that I have used --- so they may be incorrect anyway.

(Note that this doesn't address term start dates, which we've been storing as swearing-in dates, where known.)

According to the [Constitution Annotated, Amendment 20](https://www.congress.gov/content/conan/pdf/GPO-CONAN-REV-2014-10-21.pdf), quoting the Senate report on the Twentieth Amendment (1 S. Rep. No. 26, 72d Cong., 1st Sess. 2, 4, 5, 6 (1932)):

> The commencement of the terms of the first President and Vice President and of Senators and Representatives composing the First Congress was fixed by an act of [the Continental] Congress adopted September 13, 1788, and that act provided ‘that the first Wednesday in March next to be the time for commencing proceedings under the Constitution.’ It happened that the first Wednesday in March was the 4th day of March, and hence the terms of the President and Vice President and Members of Congress began on the 4th day of March.

This explains the frequency of March 3 as adjournment dates in our term end data through 1933 (72nd Congress), with only anomalies for the 2nd Congress (1793-03-02), 22nd Congress (1833-03-02), and 69th Congress (1927-03-04). I've verified that the 69th Congress did end on a March 4, and per further analysis in the Constitution Annotated (see below), it seems terms ended at some time on March 4. Nevertheless, I've left the March 3 end dates where they are (a total of 69 Congresses had that date), and I've left the 69th Congress's March 4 date alone. I've corrected the 2nd Congress and 22nd Congresses to March 3's (for consistency with the other Congresses in that time period).

The [20th Amendment](http://www.archives.gov/exhibits/charters/constitution_amendments_11-27.html) changed the start and end dates of legislative (and executive) terms. The amendment was ratified on January 23, 1933. Per section 5 of the amendment, section 1 takes effect "on the 15th day of October following the ratification of this article", i.e. October 15, 1933. Section 1 states that legislative terms start and end "at noon on the 3d day of January, of the years in which such terms would have ended if this article had not been ratified." The 73rd Congress was in session. We have its adjournment date as 1934-06-18. But its actual end date was changed by the 20th Amendment from 1935-03-03 (or -04) to 1935-01-03 at noon. So this commit changes the term end date from 1934-06-18 to 1935-01-03.

The Constitution Annotated clarifies:

> [The 20th Amendment] shortened, by the intervals between January 3 and March 4, the terms of Senators elected for terms ending March 4, 1935, 1937, and 1939 . . . It also shortened the terms of Representatives elected to the Seventy-third Congress, by the interval between January 3 and March 4, 1935.

Only a handful of Congresses since then have actually adjourned on January 3 (76th, 110th, 112th Congresses), so most of the term end dates starting with the 73rd Congress are updated by this commit to January 3.

Beginning with the 113th Congress we've been using January 3, so the last Congress updated here is the 111th Congress.

I made these changes using `sed`. Anywhere the date occurs in an `end: ` is assumed to be the end of the term at the end of a Congress. Any ends of terms on other dates not listed above are left alone because they're resignations, deaths, etc.
@konklone
Copy link
Member Author

konklone commented Aug 4, 2015

A note for anyone still subscribed to this thread, technically or emotionally, that @JoshData has filed a fix for review at #305.

@JimHarperDC
Copy link

I am emotionally subscribed to this thread, but not technically. ('Scuse the interruption... Had to.)

JoshData added a commit that referenced this issue Aug 7, 2015
…o fix term end dates to be the Constitutionally-defined end of a term.

In summary, this commit does a simple string substitution to correct the following dates:

	  2nd Congress 1793-03-02 => 1793-03-03
	 22nd Congress 1833-03-02 => 1833-03-03
	 73rd Congress 1934-06-18 => 1935-01-03
	 74th Congress 1936-06-20 => 1937-01-03
	 75th Congress 1938-06-16 => 1939-01-03
	 77th Congress 1942-12-16 => 1943-01-03
	 78th Congress 1944-12-19 => 1945-01-03
	 79th Congress 1946-08-02 => 1947-01-03
	 80th Congress 1948-12-31 => 1949-01-03
	 81st Congress 1951-01-02 => 1951-01-03
	 82nd Congress 1952-07-07 => 1953-01-03
	 83rd Congress 1954-12-02 => 1955-01-03
	 84th Congress 1956-07-27 => 1957-01-03
	 85th Congress 1958-08-24 => 1959-01-03
	 86th Congress 1960-09-01 => 1961-01-03
	 87th Congress 1962-10-13 => 1963-01-03
	 88th Congress 1964-10-03 => 1965-01-03
	 89th Congress 1966-10-22 => 1967-01-03
	 90th Congress 1968-10-14 => 1969-01-03
	 91st Congress 1971-01-02 => 1971-01-03
	 92nd Congress 1972-10-18 => 1973-01-03
	 93rd Congress 1974-12-20 => 1975-01-03
	 94th Congress 1976-10-01 => 1977-01-03
	 95th Congress 1978-10-15 => 1979-01-03
	 96th Congress 1980-12-16 => 1981-01-03
	 97th Congress 1982-12-23 => 1983-01-03
	 98th Congress 1984-10-12 => 1985-01-03
	 99th Congress 1986-10-18 => 1987-01-03
	100th Congress 1988-10-22 => 1989-01-03
	101th Congress 1990-10-28 => 1991-01-03
	102th Congress 1992-10-09 => 1993-01-03
	103th Congress 1994-12-01 => 1995-01-03
	104th Congress 1996-10-04 => 1997-01-03
	105th Congress 1998-12-19 => 1999-01-03
	106th Congress 2000-12-15 => 2001-01-03
	107th Congress 2002-11-22 => 2003-01-03
	108th Congress 2004-12-09 => 2005-01-03
	109th Congress 2006-12-09 => 2007-01-03
	111th Congress 2010-12-22 => 2011-01-03

When I first created this dataset back in 2009, I used the sine die adjournment date of Congress as term end dates (as best as I could quickly identify those dates, and the later of the House or Senate). It has been a bit of a research project to figure out the actual term end dates, especially prior to 1933. Constitutional term end dates are more appropriate since a Member is a Member, and can come back into session, until the term actually ends, not when Congress last meets. In addition, I have found that historical roll call data from Voteview and historical legislative data from THOMAS/Congress.gov indicates there was legislative activity beyond the end dates that I have used --- so they may be incorrect anyway.

(Note that this doesn't address term start dates, which we've been storing as swearing-in dates, where known.)

According to the [Constitution Annotated, Amendment 20](https://www.congress.gov/content/conan/pdf/GPO-CONAN-REV-2014-10-21.pdf), quoting the Senate report on the Twentieth Amendment (1 S. Rep. No. 26, 72d Cong., 1st Sess. 2, 4, 5, 6 (1932)):

> The commencement of the terms of the first President and Vice President and of Senators and Representatives composing the First Congress was fixed by an act of [the Continental] Congress adopted September 13, 1788, and that act provided ‘that the first Wednesday in March next to be the time for commencing proceedings under the Constitution.’ It happened that the first Wednesday in March was the 4th day of March, and hence the terms of the President and Vice President and Members of Congress began on the 4th day of March.

This explains the frequency of March 3 as adjournment dates in our term end data through 1933 (72nd Congress), with only anomalies for the 2nd Congress (1793-03-02), 22nd Congress (1833-03-02), and 69th Congress (1927-03-04). I've verified that the 69th Congress did end on a March 4, and per further analysis in the Constitution Annotated (see below), it seems terms ended at some time on March 4. Nevertheless, I've left the March 3 end dates where they are (a total of 69 Congresses had that date), and I've left the 69th Congress's March 4 date alone. I've corrected the 2nd Congress and 22nd Congresses to March 3's (for consistency with the other Congresses in that time period).

The [20th Amendment](http://www.archives.gov/exhibits/charters/constitution_amendments_11-27.html) changed the start and end dates of legislative (and executive) terms. The amendment was ratified on January 23, 1933. Per section 5 of the amendment, section 1 takes effect "on the 15th day of October following the ratification of this article", i.e. October 15, 1933. Section 1 states that legislative terms start and end "at noon on the 3d day of January, of the years in which such terms would have ended if this article had not been ratified." The 73rd Congress was in session. We have its adjournment date as 1934-06-18. But its actual end date was changed by the 20th Amendment from 1935-03-03 (or -04) to 1935-01-03 at noon. So this commit changes the term end date from 1934-06-18 to 1935-01-03.

The Constitution Annotated clarifies:

> [The 20th Amendment] shortened, by the intervals between January 3 and March 4, the terms of Senators elected for terms ending March 4, 1935, 1937, and 1939 . . . It also shortened the terms of Representatives elected to the Seventy-third Congress, by the interval between January 3 and March 4, 1935.

Only a handful of Congresses since then have actually adjourned on January 3 (76th, 110th, 112th Congresses), so most of the term end dates starting with the 73rd Congress are updated by this commit to January 3.

Beginning with the 113th Congress we've been using January 3, so the last Congress updated here is the 111th Congress.

I made these changes using `sed`. Anywhere the date occurs in an `end: ` is assumed to be the end of the term at the end of a Congress. Any ends of terms on other dates not listed above are left alone because they're resignations, deaths, etc.
JoshData added a commit that referenced this issue Sep 3, 2015
Correct term end dates in historical terms, per #7
@JoshData
Copy link
Member

JoshData commented Apr 6, 2016

This was fixed by the above commit (428d714).

@JoshData JoshData closed this as completed Apr 6, 2016
konklone pushed a commit that referenced this issue Oct 25, 2017
Peter Welch IG peterwelchvt => reppeterwelch
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

6 participants