Showing posts with label Statistics. Show all posts
Showing posts with label Statistics. Show all posts

Thursday, December 22, 2016

Electoral forensics on the 2016 parliamentary elections

In order to help monitor the fidelity of the October 2016 parliamentary election results, CRRC-Georgia has carried out quantitative analysis of election-related statistics within the auspices of the Detecting Election Fraud through Data Analysis (DEFDA) project. Within the project we used methods from the field of election forensics. Election forensics is a field in political science that attempts to identify election day issues through looking at statistical patterns in election returns. This blog post reports the results of our analysis of the 2016 proportional election results. The full report of the analysis is available here.

Our analysis suggests that the results of the 2016 elections were roughly equivalent to the 2012 proportional list elections.

Before going further into the results, two caveats and a note on methods are needed. To start with the two caveats:

  • Results are probabilistic. A test may return a statistically anomalous result, and this suggests that a given result is highly unlikely to have occurred by chance alone. The way in which we calculate the test statistics is likely to provide 1 false positive for every 100 tests performed.
  • If a test does suggest a statistical anomaly, it does not necessarily mean that election-related malfeasance caused the result, but that it may have. Statistical anomalies can be caused by benign activities such as strategic voting or divergent voting patterns within a region. Electoral malfeasance does often cause a positive test result, however. Hence, substantive knowledge and judgment of each positive test are required to determine whether malfeasance actually did occur.

When it comes to methods, to be frank, they are relatively complex. Rather than dive into the details here, we recommend that interested readers see Hicken and Mebane, 2015, here. Below we present the results of the following election forensics tests:

  • Mean of second digit in turnout;
  • Skew of turnout;
  • Kurtosis of turnout;
  • Means of the final digit in turnout;
  • Frequency of zeros and fives in the final digit in turnout;
  • Unimodality test of turnout distribution.

Results

In 2016, three of the six tests were set off:



By comparison, in 2012 two of six tests were also set off. However, one test – of the second digit mean – was exceptionally close to being set off. Due to the nature of the method – bootstrapping uses resampling with replacement – this test just as well could have been set off if run again.


Given the borderline nature of the 2012 tests, providing a conclusive comparison of the two elections is somewhat difficult. However, since the test results are roughly equivalent, the tests are indicative rather than definitive, and the elections by most accounts have been considered broadly free and fair, despite having clear issues, and the 2012 elections were considered to be broadly free and fair, despite also having clear issues, we consider the 2016 election results to also be broadly free and fair.
For more on the subject, take a look at our final report for the DEFDA project, available here.

Note: The DEFDA project is funded by the Embassy of the United States of America in Georgia, however, none of the views expressed in the above blog post represent the views of the US Embassy in Georgia or any related US Government entity.

Monday, December 19, 2016

Number of logical inconsistencies in 2016 election protocols decline

Following the 2016 parliamentary elections, a number of politicians questioned the results based on logical inconsistencies on election protocols. Some of the election protocols, which summarize election results for individual voting stations, reported that more voters had come to the polls than actually cast ballots while others reported that more votes had been cast than voters came to the polling station.  While both did happen, the Central Election Commission has made dramatic improvements compared to Georgia’s 2012 parliamentary elections.

In the 2012 parliamentary elections, according to an analysis of data the Central Election Commission provided, in the proportional list elections alone there were over 30,000 more voters that came to the polls than cast ballots. In 2016, there were less than 3000 such voters – a clear improvement.

Not only were there more voters than votes in many precincts – there were more votes cast than voters that came to the polls, again according to the official record. In the 2012 parliamentary elections, there were 696 more votes than signatures for those votes. By comparison, in 2016 there were 76 – again a clear improvement.

A third logical inconsistency present in the data is declining turnout. In the 2012 elections, in 8 precincts, there were more votes at 12PM than at 5PM. That is to say that the precincts recorded declining turnout. In 2016, by contrast, only one precinct reported declining turnout, again, a clear improvement.



While the CEC has clearly improved its recording of the vote in 2016, and small mismatches are bound to happen, any voter may reasonably ask themselves – if the CEC cannot make election protocols add up, how do I know my vote counted? Thus, we strongly recommend that the CEC make efforts to minimize the number of logical inconsistencies in future elections. Some recommendations on how the CEC might do so are available in our report on the 2016 elections.

Note: The DEFDA project is funded by the Embassy of the United States of America in Georgia, however, none of the views expressed in the above blog post represent the views of the US Embassy in Georgia or any related US Government entity.


Monday, August 17, 2015

An interesting implication of the 2014 census: Georgia is likely an upper middle income country


While Georgia has yet to be officially declared an upper middle income country by the World Bank, as a result of the 2014 census, it’s likely to be labeled one after the final census results are published in April of 2016. Interestingly, Georgia likely became one in 2013. Why is this the case and what are the implications? Let’s take a look using the 2014 preliminary census data and a population model developed by Ilia State University’s Giorgi Tsuladze published in a 2014 United Nations Population Fund (UNPF) and International School of Economics at Tbilisi State University (ISET) report.

The 2002 census was way off

In 2002, the Georgian government carried out a population census and found that there were 4.37 million Georgians. This number though was and is widely considered to be suspect. According to the 2014 UNPF report (and notably, Geostat employees at the time), the main problem with the 2002 census was its method of counting the migrant population. Specifically, the 2002 population count included 114,000 migrants who may have been permanently settled abroad rather than temporarily. This number may have been even higher considering that an estimated one million Georgians left the country between 1990 and 2002. Their family members who were interviewed for the census may have been reluctant to report that their relatives had gone abroad and instead reported them as temporary migrants or still in the country for a variety of reasons.

Not only was the census off, but so too were the civil registries which count birth and death registration. Between censuses, governments update population counts based on birth and death registrations, but because many births in Georgia happened and to a certain extent still happen outside of hospitals, births are not always registered. Also important are the lack of death registrations.

Recognizing these problems, Giorgi Tsuladze, a Professor at Ilia State University, made a downward adjustment of the population figure from the 2002 census, an upward adjustment to the birth rate, and a decrease in the estimate of the average life expectancy to estimate the population. In turn, his estimates of the population are quite close to what the preliminary 2014 census results tell us about the Georgian population – there are about 3.7 million people in Georgia (excluding South Ossetia and Abkhazia).


Geostat population estimate (thou.) Tsuladze population estimate (thou.)
2002 4,372 4,001
2003 4,343 3,966
2004 4,315 3,931
2005 4,322 3,899
2006 4,401 3,869
2007 4,395 3,839
2008 4,382 3,814
2009 4.385 3,797
2010 4,436 3,790
2011 4.469 3,786
2012 4,498 3,777
2013 4,484 3,768

Source: Tsuladze, G.; N. Maglaperidze and A. Vadachkoria. 2002. Demographic Overview of Georgia. Tbilisi, UNFPA. Cited in
Hakkert, Ralph, Gulnara Kadyrkulova, Nata Avaliani, Eduard Jongstra, Lasha Labadze, Maka Chitanava, and Nino Doghonadze. Population Situation Analysis (PSA) 2014. Rep. Tbilisi: United Nations Population Fund, 2015. Print.

Income classifications

The second important part of this story is understanding how countries are classified into income groups. The World Bank classifies countries by Gross National Income per capita (slightly different than Gross Domestic Product per capita – see here for exact definitions).

Countries with a per capita GNI of less than $1,045 are considered low income countries. Countries with greater than $1045, but less than $4,125 GNI/capita are classified as lower-middle income countries. Countries below $12,736 but above $4,125 GNI/capita are considered upper middle income countries, and countries above the $12,736 mark are considered to be upper income countries.

Since, a country’s income classification is based on the size of its population, and as we saw above, Georgia’s official population size was inflated by a fairly sizable margin for the past twelve years, Georgia’s GNI per capita (as well as GDP per capita) was underestimated.

Georgia probably moved from the lower-middle income to the upper-middle income grouping in 2013 when GNI per capita moved from from $3914 in 2012 to $4240 in 2013 (based on Tsuladze’s population estimates). In 2014, using the preliminary census data, Georgia’s GNI was $4489/capita. Hence Georgia has very likely moved income groups, barring a major miscount of the preliminary census data of roughly 330,000 people.


Why does this matter?

Well, it is good and bad news for Georgia.

To start with the bad, aid is sometimes distributed based on a country’s economic status. There are many other important factors at play (see here for a discussion of the subject), but nonetheless it is often considered in aid decisions. Hence, Georgia may expect lower levels of aid in the coming years as its per capita economic statistics are adjusted upward following the finalization of the 2014 census in 2016.

When it comes to the good news for the country, Georgians are doing better than the numbers suggested. This does not change the facts on the ground and widespread poverty experienced in Georgia, but in the long run it can lead to a number of benefits. For instance, foreign private capital flows may increase as the country may be perceived as a more enticing investment environment, having moved to a higher income category.

The upward adjustment will also be important for Georgia’s Euro-Atlantic integration prospects. One of the key factors which the EU has identified as a barrier to prospective membership for countries in its neighborhood is low income levels, and as Georgia’s income level gradually increases, it will make Georgia a more attractive partner country. Notably, the lower population also means that per capita income is increasing at a higher rate than previously thought. In the short term, it may also help ease fears over migrant flight from Georgia (which is likely an inhibiting factor at present for Georgia in the EU visa liberalization process). It is important to note that if income inequality persists in Georgia, economic growth is unlikely to deter migrants from attempting to make their way to the EU, though a fuller treatment of this subject is beyond the scope of this post.

On the grand scheme of things, the adjustment is good as well. While not necessarily good for Georgia, countries in more dire straits may receive more aid that would have been aimed at Georgia. Better decisions about what kind of aid the country receives may also result from the more accurate data and income categorization.

Although we should not expect to see the official income categorization change to upper-middle until after Geostat has published the final census numbers and adjusted its population estimates for 2002-2014, it should occur in the next few years.

To take a look through the preliminary 2014 census results, take a look here, and for the estimates of the population size which this blog is based on as well as other interesting data and analysis on Georgia’s demographic situation, take a look at the UNPF/ISET report, here. Notably, Georgia is not the first and surely not the last country to have a major economic indicator readjustment based on something besides economic growth. Ghana and Nigeria both have had large changes to their economic indicators in recent years caused by how GDP was calculated rather than growth with interesting implications. Listen to this Planet Money story to find out more.

Wednesday, October 21, 2009

Survey Documentation and Analysis with South Caucasus data

Earlier this month, CRRC launched Survey Documentation and Analysis (SDA), a web-based interface for statistical analysis. SDA was designed by the Association for Computer Assisted Survey at the University of California, Berkeley. Through SDA you can for example calculate frequencies, make cross tabulations, comparison of means and comparison of correlations. CRRC has now loaded its data, based on interviews carried out with more than 6 000 respondents in Georgia, Armenia and Azerbaijan, into the SDA platform. As a result, it is now possible for anyone to find out information on everything from language knowledge to perceptions of the Russian-Georgian war.

In comparison to several other statistical software programs, SDA does not require any prior knowledge of statistics. Extracting data is an easy and fast process, as the program provides the user with explanations for the different functions. In addition, there is no need to download any software. You simply visit http://www.crrccenters.org/sda/ and start exploring CRRC’s data. Having reliable, up-to-date and easily accessible data on an extensive number of topics is now also possible for those of us that have earlier refrained from using statistical data due to its sometimes rather complex nature.

Thursday, October 09, 2008

South Caucasus Data 2007 on Unemployment

Unemployment clearly is one of the pressing issues in the South Caucasus. But there is a lack of reliable data on people being without and looking for a job. This blog, based on CRRC’s Data Initiative 2007, provides a snapshot on these numbers.

According to CRRC’s dataset, about 25% of the adult population in Armenia and Georgia, and 20% of Azerbaijan’s citizens say they are unemployed. Further analyzing these numbers shows that 18% in Georgia, 14% in Armenia and 12% in Azerbaijan are actually interested in looking for a job.

[Note: excluded are “students", "housewives", "disabled" or "retired" - even if they are looking for a job.]

Yet the data shows sizeable differences across the countries, depending on whether you ask in rural areas, urban environments, or the capital. Let's look at what people say when asked whether they consider themselves to be employed. Note that housewives, pensioners, disabled and students are also considered "not employed".

Do you consider yourself to be employed? This employment may be part-time or full-time, you may be officially employed, informally employed, or self-employed, but it brings you monetary income.


If you analyze the data of by settlement type, it reveals that of those that describe themselves as not employed a relatively low number of people look for work in Baku (22%), compared with Tbilisi (29%) and Yerevan (32%). Besides, about the same share of people (again, of those describing themselves as not employed with monetary income) in the three countries look for a job in rural areas (nearly 30%).

However, the data impressively illustrates that the major interest -- among those that are not employed -- in a workplace can be found in urban areas, where about 40% of Armenians and Georgians, and almost 50% Azerbaijanis try to find work. This figure powerfully underlines the desolation of Caucasian cityscapes.

Of those that are not employed, what percentage is looking for a job?

Finally, the DI statistics show that the same number (once you factor in the margin of error) of people is unemployed and interested in a job, but not currently looking: 6% in Armenia, and 5% in Georgia and Azerbaijan. A slightly lower number of the unemployed is not looking for a job at all. Have those already given up?

Now the definitions of unemployment always are a little complicated (are pensioners looking for work considered unemployed?), but here is an article that can help. If you are interested to check the datasets yourself , please download it from CRRC’s homepage. For more information on the Data Initiative project, please click here.

Friday, July 18, 2008

PFA Report on “Armenia’s 2008 Presidential Election”

For those who have been far from Armenia or who have not actively followed the plethora of developments that have occurred in the country for the past six months, the report encompassing a nearly full picture of the current situation in Armenia has finally become available. “Armenia’s 2008 Presidential Election: Select Issue an Analysis” is a report recently released by Policy Forum Armenia (PFA), a newly founded association.

The report is the first of its kind following the February 19, 2008 Presidential Election of Armenia since it provides a full description of the pre-election and post-election events. The report includes both qualitative and quantitative analysis. The information provided in the qualitative sections of this report is mainly based on Armenian local and international newspaper articles, reports released by international organization as well as blogs on the internet. Unlike the primarily technical reports by OSCE/ODIHR, the report does not just limit itself to describing the 2008 presidential elections, but also presents it in the larger context of political and social developments of post-Soviet Armenia.

For those who have closely followed the political developments in Armenia, a particularly interesting section of the report may be the section on the “Statistical Analysis of the Official Election Outcome”. In this section the authors utilize a number of tests developed in the 1990s by Sobianin and Sukhovolskiy, and later revised by other scholars such as Gelman, Kaiunov, Michael Myagkov (University of Oregon), Peter Ordeshook (California Institute of Technology), and their co-authors. The analysis of the election results through this methodology indicates inconsistencies in the 2008 Presidential Election. For example, the report reveals that there was much higher voter turnout in the regions outside of the capital city-Yerevan. This is unlikely, as generally considerably more civic activism is observed in Yerevan as opposed to the rural areas. Out of the 1,923 polling stations in Armenia more than 129 polling stations had higher than 90 percent voter turnout (p. 21, see figure below). Such turnout levels are highly unlikely, especially given the high migration levels. The other tests reveal inconsistencies within the distribution of individual candidates’ votes, in the relationship between the candidates’ votes and voter turnout, and within the distribution of invalid ballots. The report is careful to specify that the statistical findings do not provide definitive proof of election fraud, but only an (albeit powerful) indication.

Finally, the report concludes with a full section devoted to the civil society awakening in Armenia in connection with the 2008 Presidential Election. More specifically, the final section discusses the increase in the activism especially among women and the youth, as well as the rise in information sharing and networking through the internet. While the report does not provide any innovative recommendations to mediate the post-election discontent in Armenia, it provides a solid ground for policy makers to put the events of the past six months into perspective, assess what the available tools are and put the February 19, 2008 elections into the wider political context of Armenia’s newly independent history.

On a side note, similar studies of election fraud and perception were conducted by CRRC fellows in 2005 by Dr. Masis Poghosyan and Sergey Harutyunyan.

For more detail, check out the report itself at http://www.pf-armenia.org/

Wednesday, March 14, 2007

Divorce Rates: the seven-year-itch?

According to popular lore, marriages often break up after around seven years. What does the Georgian data say? How long have those who are getting divorced typically been married?

Georgian data does not support the seven-year itch hypothesis. Divorces seem to be pretty equally distributed across the years, with some fluctuations, year-by-year.

In some ways, a fairly high number of divorces still takes place after 15 years (according to the data, the number of couples breaking up after more than 20 years is the single largest group of divorcees, but it consolidates all 20+ data).

As for the total divorce number, it has remained stable from 2001 to 2005, at around 1900 divorces per year. By comparison, around 13.000 people married annually, with an upward tendency more recently (2005 a bumper year, with 18.012 marriages).

According to one divorce league table (some inconsistency here), Georgia has about 12% divorces per marriage, Azerbaijan 15%, Armenia at 18%. Matrimonial harmony, or at least stubborn persistence, compared to Kyrgyzstan (25), Kazakhstan (39), United States (41), Russia (65) or Belarus (69). Take another league table in which divorces are listed by 1000 people, and the US comes first in divorces (4.95), Russia third (3.36), and Georgia would still come below Syria, with less than 0.5 divorces per 1000 people.

The Georgian data, and much more engaging information, is available on these pages of the Georgian Department of Statistics.

Tuesday, November 07, 2006

Marriage statistics -- food for thought, hunger for data

I would like to know more about this: are more people getting married, or are just more couples getting registered? And is it really the case that more of a third of the registered marriages are between Georgians and foreign citizens? We should get comparative data from Armenia (which would include diaspora-marriages) and Azerbaijan. Data, data, data....
-------------------------------
"During January-September 2006, the State Registry issued marriage certificates to 4004 Georgian citizens marrying foreigners. Of those, 3611 citizens were sent documents to register their marriage from abroad and 393 Georgians married a foreigner in Georgia. Georgians marry foreigners primarily from Russia, Greece, the US, Israel, Turkey and Germany, reports the newspaper Akhali Versia."

Total registered weddings:
2005 -- 18,012
2004 -- 14,866
2000 -- 12,870

Source: Department of Statistics
, World of weddings, By M. Alkhazashvili, The Messenger, 6 Nov, 2006