Using PB to detect fraud on mexican elections?

ricardo · Post by **ricardo** » Mon Sep 11, 2006 4:57 pm

Maybe some of you read about mexican elections. One party claims that a fraud was done.

Now, the interesting part is that some mexican journalist friend of mine ask me to invite some coders to develope small tools (because the task is so easy) to verify if that fraud was real.

I invite to as many member of this community to develope some small code to achieve this.

the task is this one:

-Need to download the database (4 megas) of the mexican election from the Federal electoral institute (IFE), its a txt file with each voting-place in one line.

-Need to read it using PB and compare if there are any artimeitcal error as one party claims.

The idea is not read all the databse each time, but maybe some random voting-places (don't know how to say it in english) and check if there are matemathical errors and show the results. (Or if you want to verify the whole data its okay)

-------------------------------

If you are interested in participating:

The database can be downloaded from

http://www.ife.org.mx/documentos/proces ... ep2006.htm

Go to the PREP2006-Presidente.zip link and you will get the 4 megas file (its the official one!!)

The file has the data in this columns:

ESTADO|NOM_ESTADO|DISTRITO|SECCION|ID_CASILLA|TIPO_CASILLA|EXT_CONTIGUA|NUM_ACTA_IMPRESO|NUM_BOLETAS_RECIBIDAS|NUM_BOLETAS_SOBRANTES|TOTAL_CIUDADANOS_VOTARON|NUM_BOLETAS_DEPOSITADAS|PAN|PAN_VCAP|ALIANZA_POR_MEXICO|APM_VCAP|POR_EL_BIEN_DE_TODOS|PBT_VCAP|NUEVA_ALIANZA|NA_VCAP|ALTERNATIVA_SOCIALDEMOCRATA|ASDC_VCAP|NUM_VOTOS_CAN_NREG|NUM_VOTOS_NULOS|TOTAL_VOTOS|CASILLA|LISTA_NOMINAL|HORA_RECEPCION_CEDAT|HORA_CAPTURA_CEDAT|HORA_REGISTRO

That means: (The ones with * are the ones that must be verified if the aritmetical result is fine or not)

State
Name of the State
District
Section
ID of voting-place
Kinf of voting place
exterior
number of act
*Number of voting tickets receiben on that voting-place (the papers where people votes)
*Number of voting tickets that was not used at the end of the day
*Total of citizens that vote at that place
*Number of voting tickets found in the ballot box

Next 9 colums are for the votes of different partys

*Number of votes canceled
*Number of nule votes
*Total of votes
Voting-place
*Number of people allowed to vote at that voting place
Time of reception
time of capture
time of capture

Okay, thats the info.

here the important thing is NOT to verify how many votes receive each party, but to detect if there was MORE or LESS ballots in the ballot box that the oneas are suppoused to be, based on the number of voting ticktes received at the beggining of the day LESS the number of tickets that was not used at the end of the day, etc.

If you see carefully, many columns must match on each voting place. IF NOT its because something strange happends.

Okay, one party claims that there are plenty errors and that is very easy to anyone to notice it doing any elemental analysis.

Ive done this by myself and the result was that ITS A MESS, there are something like 30% of voting-places with MANY MANY errors.

I hope some of you get interested and develope some easy algorithm and publish it here and if you want publish here the results you find in your tests verifying the mexican election numbers.

*The database is stored as .txt file and each voting place is one line, and each column is separed by |
So it easy to split it.

I hope my explanation is not to badin english :roll:

This is a good cause and PB coders could do a good job.

If the results are interesting it will published on a newspaper and a credit will be showed to PB community!!

Thanks a lot!!

Killswitch · Post by **Killswitch** » Mon Sep 11, 2006 5:35 pm

I'd be interested in writing an application for this, but I'm not really sure what needs to be checked (sorry), could you clarify?

ricardo · Post by **ricardo** » Mon Sep 11, 2006 5:43 pm

Killswitch wrote:I'd be interested in writing an application for this, but I'm not really sure what needs to be checked (sorry), could you clarify?

Good!!

Yes, what need to be clarified is if the elemental math on each ballot place are okay.

Per example, verify if the number betwen the ballots that was not used at the end of the day (column 10), the number of ballots received at the begging of the day (column 9) and the total of ballots counted as votes (coumn 12) matchs.

Also if the number of citizens that votes matchs against all this numbers.

Believe it or not, in that basic arithmetics i found a big percent of errors. Per example, many voting-places have at the end of the day MORE VOTES that the ones suppoused to have if we follow some artimetic.

Per example:

One voting-place receive 400 ballots at the beggining of the elections day.
230 citizens go to vote.
But they have 190 not used ballots at the end of the day (and they should have only 170!!) Where does that 20 extra votes comes from? Numbers don't match!!!.

That shows that something goes wrong there... if this happends in an important number of voting-places, then we can feel that the whole election is not much cleaner.

What need to be verifyed is the elemental math only and check if there are errors or not.

Killswitch · Post by **Killswitch** » Mon Sep 11, 2006 7:18 pm

Here we go. Place in the same folder as the database.

I think it's working correctly - it outputs to a new file called Results.txt.

I hope I got this right, but this is what it checks:

(Column 9 - Column 10) <> Column 12 ;If true, it's wrong

This code is easily modifyable as you can use GetColumn to get the value of any particular column, so you can extend it to check as many different results as you want.

Code: Select all

Procedure.s GetColumn(Column,String.s)
  For t=1 To Column-1
    Pos=FindString(String,"|",Pos)+1
  Next t 
  ProcedureReturn Mid(String,Pos,FindString(String,"|",Pos)-Pos)
EndProcedure

Structure Votes
  ID.l
  Is.l
  ShouldBe.l
EndStructure

NewList BoxTally.Votes()

If OpenFile(0,"PREP2006-Presidente.txt")
  
  ReadString(0) : ReadString(0) ;Ignore first two lines
  t=2
  While Eof(0)=#False
    
    t+1
    
    String.s=ReadString(0)
    
    ID=t
    
    Received=Val(GetColumn(9,String))
    
    NotUsed=Val(GetColumn(10,String))
    
    InBox=Val(GetColumn(12,String)) 
    
    
    If (Received-NotUsed)<>InBox
      Wrong+1
      AddElement(BoxTally())
      BoxTally()\ID=ID
      BoxTally()\Is=InBox
      BoxTally()\ShouldBe=(Received-NotUsed)
    EndIf
    
  Wend
  
  If CreateFile(0,"Results.txt")
    WriteStringN(0,"RESULTS")
    WriteStringN(0,"")
    WriteStringN(0,"  Box Errors")
    WriteStringN(0,"    ;Number of votes in box does not match ballets at station subtract ballets used")
    WriteStringN(0,"    ;Line: InBox / Should Be")
    ForEach BoxTally()
      WriteStringN(0,"    "+Str(BoxTally()\ID)+": "+Str(BoxTally()\Is)+" / "+Str(BoxTally()\ShouldBe))
    Next 
  Else  
    MessageRequester("Error","Could not create Results.txt")
    End
  EndIf
  
  MessageRequester("Done!",Str(Wrong)+"/"+Str(t-2)+" incorrect.")
  
Else
  MessageRequester("Error","Could not open database.")
EndIf

ricardo · Post by **ricardo** » Mon Sep 11, 2006 7:43 pm

Killswitch

Great!!

Its very similar to the one i develope my self and get same results, more than 50 k of error on 117 k

So, ATM it seams that the election was a fraud!!

Can you share the results you get in your PC please?

Killswitch · Post by **Killswitch** » Mon Sep 11, 2006 8:35 pm

I'll upload them later, they're on my laptop.

I don't think this is conclusive though, I mean ballots can get lost, or get thrown out etc, etc.

ricardo · Post by **ricardo** » Mon Sep 11, 2006 8:45 pm

Killswitch wrote:I'll upload them later, they're on my laptop.

I don't think this is conclusive though, I mean ballots can get lost, or get thrown out etc, etc.

Yes, i know.

but if you check, many times ballots appers from nowhere!!

Not all differences means ballots lost, but many many times means ballots magical appear!! he he

However, close to 50% errors, sound very bad

the winning party winds with less then 0.5% difference... sounds very suspicious!

Killswitch · Post by **Killswitch** » Mon Sep 11, 2006 10:52 pm

Here's the results I got running the above code:

http://www.btinternet.com/~douglas.marsh/Results.txt

But, I just ran a test for when there are more ballots in the box than were used and the results are still pretty bad: 17849 - 15% - that's terrible.

BTW I realised a small error in the code I pasted above, when it reports the number of incorrect items / total items the first two lines of the file count (which are just formatting, really) so I've edited it above. Nothing major.

John Bedlam · Post by **John Bedlam** » Mon Sep 11, 2006 10:55 pm

I probably made some mistakes in this, but if I understood the structure of the database correctly, and my code is right, there are close to two million votes more than there are people who have voted?

Code: Select all

If OpenFile(0, "G:\PREP2006-Presidente\PREP2006-Presidente.txt")
  RecordsRead.l = 0
  WrongRecords.l = 0
  Discrepancy = 0
  Repeat
    RecordsRead + 1
    Record.s = ReadString(0)
    If (Val(StringField(Record, 9, "|"))-Val(StringField(Record, 10, "|"))) <> Val(StringField(Record, 12, "|"))
      WrongRecords + 1
      ;Debug Record
      Discrepancy + (Val(StringField(Record, 11, "|"))-Val(StringField(Record, 12, "|")))
    EndIf
  Until Eof(0)
  Debug RecordsRead
  Debug WrongRecords
  Debug Discrepancy
  CloseFile(0)
EndIf

I must've made a mistake somewhere.

ricardo · Post by **ricardo** » Mon Sep 11, 2006 11:19 pm

John Bedlam wrote:I must've made a mistake somewhere.

I will show my numbers as soon as i finish, but some preliminar test gives me something like 1.9 million difference (if not make a mistake too!).

mike74 · Post by **mike74** » Mon Sep 11, 2006 11:37 pm

John Bedlam wrote:I must've made a mistake somewhere.

When I run the program below on Killswitch's results, it seems that the cumulative difference in lines where the number of ballots were greater than it should have been is about 1 million and the cumulative difference in lines where the number of ballots were less than it should have been is about 3 million. The net is about 2 million.

Code: Select all

Enumeration
  #votes
EndEnumeration
OpenFile(#votes, "votes.txt")
intot = 0: shouldtot = 0: posdiff = 0: negdiff = 0: poscount = 0: negcount = 0
While Eof(#votes) = 0
    linein = 0: lineshould = 0
    line.s = ReadString(#votes)
    colpos = FindString(line, ":", 1)
    slashpos = FindString(line, "/", 1)
    linein = Val(Mid(line, colpos + 1, slashpos - colpos - 1))
    intot = intot + linein
    lineshould = Val(Mid(line, slashpos + 1, Len(line) - colpos - 1))
    shouldtot = shouldtot + lineshould
    If linein > lineshould: poscount + 1: posdiff = posdiff + linein - lineshould: Else: negcount + 1: negdiff = negdiff + lineshould - linein: EndIf
Wend
   Debug shouldtot
   Debug intot
   Debug intot - shouldtot
   Debug Posdiff
   Debug Negdiff
   Debug poscount
   Debug negcount

mike74 · Post by **mike74** » Mon Sep 11, 2006 11:42 pm

Wait I think got it backwards.

mike74 · Post by **mike74** » Mon Sep 11, 2006 11:52 pm

mike74 wrote:Wait I think got it backwards.

Yeah, I did. I just edited my first post and the program for clarity. It appears that there were 17,849 cases where there were more ballots than there should have been, and 39,808 cases where there were less.

ricardo · Post by **ricardo** » Tue Sep 12, 2006 12:12 am

Using Kohn idea, but trying to be more precise in the different kind of error i found there could be 4 different kind of errors on each voting place:

-More deposited ballots that was suppoused to be (received - left)
-Less deposited ballots that was suppoused to be (received - left)
-More ballots deposited than citizens that vote on that place
-Less ballots deposited than citizens that vote on that place

Then i change the code to this:

Code: Select all

If OpenFile(0, "PREP2006-Presidente.txt") 
    RecordsRead.l = 0 
    Record.s = ReadString():Record.s = ReadString();first two lines
    Repeat 
        RecordsRead + 1 
        Record.s = ReadString() 
        uBallotsReceived = Val(StringField(Record, 9, "|"))
        uBallotsLeft = Val(StringField(Record, 10, "|"))
        uCitizensThatVote = Val(StringField(Record, 11, "|"))
        uDepositedBallots = Val(StringField(Record, 12, "|"))
        
        If uBallotsReceived-uBallotsLeft < uDepositedBallots
            ;More ballots that was suppoused to be
            WrongRecords1 + 1
            Discrepancy1 + (uDepositedBallots-(uBallotsReceived-uBallotsLeft))
        ElseIf uBallotsReceived-uBallotsLeft > uDepositedBallots
            ;Less ballots that was suppoused to be
            WrongRecords2 + 1
            Discrepancy2 + ((uBallotsReceived-uBallotsLeft)-uDepositedBallots)
        EndIf

        If uCitizensThatVote > uDepositedBallots
            ;More ballots deposited than citizens
            WrongRecords3 + 1
            Discrepancy3 + (uCitizensThatVote-uDepositedBallots)
        ElseIf uCitizensThatVote < uDepositedBallots
            ;Less ballots deposited than citizens
            WrongRecords4 + 1
            Discrepancy4 + (uDepositedBallots-uCitizensThatVote)
        EndIf
        
    Until Eof(0) 
    Debug RecordsRead 
    Debug ""
    Debug "More Ballots thats suppoused to be: " + Str(WrongRecords1 ) + " places"
    Debug "Less Ballots thats suppoused to be: " + Str(WrongRecords2) + " places"
    Debug "More citizens that ballots: " + Str(WrongRecords3 ) + " places"
    Debug "Less citizens that ballots: " + Str(WrongRecords4) + " places"
    Debug ""
    Debug "More ballots than difference (received - left): " + Str(Discrepancy1) + " ballots"
    Debug "Less ballots than difference (received - left): " + Str(Discrepancy2) + " ballots"
    Debug "Less deposited ballots than citizens: " + Str(Discrepancy3) + " ballots"
    Debug "More deposited ballots than citizens: " + Str(Discrepancy4) + " ballots"
    Debug ""
    
    CloseFile(0) 
EndIf

I don't know if made some mistake since results show TOO MANY errors in the results of the election!!

117296

More Ballots thats suppoused to be: 17849 places
Less Ballots thats suppoused to be: 39808 places
More citizens that ballots: 32129 places
Less citizens that ballots: 19905 places

More ballots than difference (received - left): 966559 ballots
Less ballots than difference (received - left): 3127090 ballots
Less deposited ballots than citizens: 3419291 ballots
More deposited ballots than citizens: 1597709 ballots

Can anybody see if im wrong here please?

ricardo · Post by **ricardo** » Tue Sep 12, 2006 12:15 am

mike74 wrote:
mike74 wrote:Wait I think got it backwards.
Yeah, I did. I just edited my first post and the program for clarity. It appears that there were 17,849 cases where there were more ballots than there should have been, and 39,808 cases where there were less.

Same that i get using modified John's code!!

Did you count how many ballots did we get on each case (not only how many cases with discrepancy, but the total of ballots on each discrepancy case)?

My results are very big ones, its hard for me to believe that there could be such amount of ballots missed/added.