This paper explores the connection between text sentiment from Twitter and public opinion polls, focusing on consumer confidence and political opinion during 2008-2009. The study analyzes Twitter messages and public opinion surveys, finding strong correlations between sentiment word frequencies and poll data, with correlations as high as 80% in some cases. The results suggest that text sentiment can serve as a substitute or supplement to traditional polling. The study uses Twitter messages from 2008-2009, collected via the Twitter API and the Gardenhose stream, and public opinion surveys from multiple polling organizations. For consumer confidence, the study uses the Index of Consumer Sentiment (ICS) and the Gallup Economic Confidence Index. For political opinion, it uses Gallup's daily tracking poll for Obama's job approval and tracking polls from the 2008 U.S. presidential election. The analysis involves message retrieval and opinion estimation. Messages are filtered by topic keywords, and sentiment is determined by counting positive and negative words. The sentiment ratio is calculated as the ratio of positive to negative messages. To smooth the sentiment ratio, a moving average is applied over a window of past days. The study finds that text sentiment can predict future poll movements and that sentiment ratios correlate with poll data, with the strongest correlations for the ICS and Gallup data. The results suggest that text sentiment can be a leading indicator of public opinion, especially for long-term trends. However, the effectiveness of text sentiment varies depending on the smoothing window and lag parameters. The study also finds that topic frequencies correlate more strongly with polls than sentiment scores. For example, the volume of messages mentioning Obama correlates strongly with poll numbers, even when sentiment scores do not. The study concludes that text sentiment can be a useful tool for public opinion measurement, and that further research is needed to improve sentiment analysis techniques and understand how different signals reflect public opinion.This paper explores the connection between text sentiment from Twitter and public opinion polls, focusing on consumer confidence and political opinion during 2008-2009. The study analyzes Twitter messages and public opinion surveys, finding strong correlations between sentiment word frequencies and poll data, with correlations as high as 80% in some cases. The results suggest that text sentiment can serve as a substitute or supplement to traditional polling. The study uses Twitter messages from 2008-2009, collected via the Twitter API and the Gardenhose stream, and public opinion surveys from multiple polling organizations. For consumer confidence, the study uses the Index of Consumer Sentiment (ICS) and the Gallup Economic Confidence Index. For political opinion, it uses Gallup's daily tracking poll for Obama's job approval and tracking polls from the 2008 U.S. presidential election. The analysis involves message retrieval and opinion estimation. Messages are filtered by topic keywords, and sentiment is determined by counting positive and negative words. The sentiment ratio is calculated as the ratio of positive to negative messages. To smooth the sentiment ratio, a moving average is applied over a window of past days. The study finds that text sentiment can predict future poll movements and that sentiment ratios correlate with poll data, with the strongest correlations for the ICS and Gallup data. The results suggest that text sentiment can be a leading indicator of public opinion, especially for long-term trends. However, the effectiveness of text sentiment varies depending on the smoothing window and lag parameters. The study also finds that topic frequencies correlate more strongly with polls than sentiment scores. For example, the volume of messages mentioning Obama correlates strongly with poll numbers, even when sentiment scores do not. The study concludes that text sentiment can be a useful tool for public opinion measurement, and that further research is needed to improve sentiment analysis techniques and understand how different signals reflect public opinion.