 | |
| Formation | 1980 |
|---|---|
| Founder | Doug Lonnstrom |
| Location |
|
Director | Donald P. Levy |
| Affiliations | Siena College |
| Website | scri |
| Remarks | Conducts expert and public opinion polls, focusing on New York State and the United States, on issues of public policy interest |
Siena College Research Institute (SCRI) is an affiliate ofSiena College, located originally in Friars Hall and now in Hines Hall on the college's campus, inLoudonville, New York, in suburbanAlbany.[1][2][3] It was founded in 1980.[4] Statistics and finance professorDoug Lonnstrom was its founding director.[5] Donald P. Levy is its current director.[6][7]
It conducts both expert and public opinion polls, focusing on New York State and the United States, on issues ofpublic policy interest. They include education, health care, andconsumer confidence, and explores business, economic, political, voter, social, educational, and historical issues.[4][8] SCRI conducted surveys on New Yorkers' sentiments towards the creation of theCordoba House Mosque near theWorld Trade Center site inlower Manhattan, and theArizona Immigration Law.[9][10]
Among other things, starting in 1982 SCRI has polled presidential scholars in an effort to rate theUnited States presidents,[11][12] as well as theFirst Ladies.[1][13][14][15] It has also conducted polls as to America's most notable women,[16] television's most memorable moment,[17] and consumer confidence.[18][19]
In July 2013, SCRI andThe New York Times began a polling partnership, with its first collaborative poll conducted for the2013 New York City mayoral election.[20]
In 2018,The New York Times/Siena College partnership delivered the first real time Midterm Election polls.[21][22] In its review of polling conducted during the2022 midterm elections, FiveThirtyEight found thatThe New York Times/Siena College partnered polls had the loweststatistical error of all pollsters it measured, tied withSuffolk University.[23] It currently holds a 3-star rating fromFiveThirtyEight, their highest possible rating for a pollster, based on measurements of error, bias, and transparency in methodology. As of 2024, FiveThirtyEight describesThe New York Times/Siena College Poll as the most accurate pollster in America.[24][25]
The New York Times/Siena College Poll contacts arandom sample of respondents, with sampling frames drawn from a national voter file of registered voters from data company L2 and supplemented with voter file-matched cellular numbers from market research company Marketing Systems Group. It utilizes live interviewers calling both landlines and cellphones in a dual frame model. For recent polls conducted byThe New York Times/Siena College, over 90% of contacts are reached through cell phones and calls are aimed to last under 15 minutes. The poll aims to reach a representative sample of American voters and adjusts forparticipation bias viastratified sampling, utilizing demographic information from its voter file such as race, party affiliation, and regional location. The poll attempts more contacts with respondents from groups that are less likely to respond. As of 2024, overallresponse rates toThe New York Times/Siena College Poll calls are usually under 2%. A 2022 incentivized poll conducted byThe New York Times as an experiment in partnership withIpsos achieved a 30% response rate while finding similar results to the lower response rate Siena College partnered polls.[26][27][28][29]The New York Times/Siena College Poll has also variously used a turnout model to prioritize calls to a "likely electorate", which is gauged based on trends detected inThe New York Times/Siena polls, past election results, and demographic data. This model was utilized for sampling in the2018 midterm election but not the2016 presidential election.[28][30]
Completed surveys are weighted for probability of selection by stratum. Demographic weights of likely voters are weighted primarily with modeled turnout estimates from the poll's overall voter file, with adjustments made based on 12 demographic variables. Weighting of registered voter results, instead of likely voters, is conducted directly through characteristics derived from the voter file, without using the turnout model. Demographic information on education is not available in the voter file, and is weighted separately with theAmerican Community Survey andCurrent Population Survey from theUnited States Census Bureau. Poll questions are published verbatim in the reports of each poll.[27][31][28][30]
When nearing election days, the poll emphasizes results among likely voters over registered voters overall, which is determined through a respondent's voting history and self-reported intention to vote.[27]