John Harwood joined The Wall Street Journal in 1991 as White House correspondent. He subsequently covered Congress and national politics, and became National Political Editor in 1997. He has reported on each of the last five American presidential elections. Read John Harwood's bio

Q: Can you describe to what degree the American public may be buying into the supposed security of re-electing an incumbent president in a time of war? How did similar conditions play-out in past elections? What role did security play in those elections?

The boost that George W. Bush received at his convention, and the fact that part of it appears to have endured, suggests that security broadly defined may be playing a larger role in the president's favor than it did earlier this year. This spring, as the Iraqi insurgency flared and the Abu Ghraib prisoner abuse tarred America's image around the world, the security issue was defined mainly by bad news from Baghdad. Now, thanks to a sustained period in which Iraq faded from the headlines and Mr. Bush cast the conflict as part of the broader anti-terror war, Americans are thinking differently about security. That plays to the president's strengths as the nation's post-9/11 leader. One key question now is whether the upper hand Mr. Bush enjoys at the moment survives the upsurge in insurgent violence we've seen in the last few days.

Americans tend to gravitate to safe choices in war time, and to define incumbents as their safe choices; President Wilson was re-elected in 1916 on the brink of America's entry into World War I, and Franklin Roosevelt was re-elected amid World War II in 1940 and 1944. That tendency didn't prevent Harry Truman from falling so far in popularity that he declined to seek re-election in 1952; he watched WWII hero Dwight Eisenhower win election to succeed him.

More recently, security has played a key role in two incumbent re-election victories. In 1964, Lyndon Johnson swamped Barry Goldwater in a campaign notable for Johnson's "Daisy" ad suggested that Mr. Goldwater might trigger nuclear war. During the Cold War 20 years later, Ronald Reagan's advertising against Walter Mondale reminded Americans that there was a Soviet "bear in the woods."

By contrast, the end of the Cold War, and the corresponding shift from security to domestic issues, helped Bill Clinton oust George H.W. Bush, the current president's father, in 1992 as the economy struggled to recover from recession.

Q: President Bush's National Guard service record made headlines again last week. Please explain its relevance to this election cycle?

That's a very good question, which some in both parties are asking. Mr. Bush's National Guard record mattered little in 2000, when its impact was potentially greatest because Americans were less familiar with him. Now Democrats will have a tougher time using it to harm the president, since Americans have had to form opinions of his four years as commander in chief -- three of them during wartime.

Strategists allied with Mr. Kerry acknowledge the limited utility of this issue. But they say the controversy has had some effect in recent days nevertheless by generating media coverage that interferes with the president's ability to communicate with voters. After the president's success over the last several weeks, Democrats welcome any static like that.

Q: What do we actually know at this point about the relationship between 9/11, the war on terror, and the war in Iraq? In your reporting, how do you make sure you don't play into the conflation?

That's a tricky issue for which clear answers are hard to come by. Many Democrats argue that the Iraq war can't be reasonably tied to the war on terror since Iraq had no demonstrated role in the 9/11 attacks. Indeed, they say that the Iraq war damaged American success in the war on terror by diverting resources from the effort to vanquish Osama bin Laden and al Qaeda.

The Bush administration argues that the Iraq war CAN be tied to the war on terror, even though Secretary of State Colin Powell has acknowledged in recent days that Iraq appears not to have been directly connected to 9/11. That's because they consider Saddam Hussein's propensity to develop and use advanced weaponry to have been a threat, which could have armed international terrorists and threatened the U.S.

Our goal in reporting is to make as clear as possible what we know and what we don't. That usually requires cast each side's argument as assertion rather than fact -- while offering from best possible sources what evidence is available to support or impeach those assertions.

Q: What factors go into deciding when to initiate a Wall Street Journal poll, and how seriously do you treat the poll numbers? Can a poll really indicate the outcome of an election? What do voters need to know about the myriad of polls/poll numbers that come out in the final several weeks of an election?

I love our Wall Street Journal/NBC News poll, and would be happy to have one once every couple of weeks. Alas, polls cost money and our budget allows for one just every five weeks or so. The timing of those polls is influenced to some extent by events and by when our poll might stand out. We chose to poll immediately before the Democratic and Republican Conventions, for example, but not immediately after since we knew so many other news organizations would do post-convention polling.

I take our numbers pretty seriously, while realizing that all polls are snapshots that can be subject to sampling error, bias in question wording, and distortions imposed even by the order in which questions are asked. Our poll was conducted for years by veteran Democrat Peter Hart and veteran Republican Bob Teeter. After Bob passed away earlier this year, he was succeeded by Bill McInturff. All our aces have an outstanding track record on both the science, such as selection of respondents, and the art, such as question phrasing and analysis, of polling. People who work in politics tell me they consider our poll exceptionally sound and reliable.

For most of an election year, polls can NOT tell you who will win an election. But they can tell with a high degree of accuracy where public opinion is at any given moment. Very close to an election, a well-conducted poll may closely resemble the election outcome; exit polls conducted on election day VERY closely resemble election outcomes, though as we saw in 2000 extremely close elections are difficult for anybody to call until votes are counted.

Polling has gotten more difficult to conduct in recent years, in part because so many Americans now refuse to respond to telephone surveys. The 2004 campaign is introducing another complication: the intense voter turnout drives being conducted by both Democratic and Republican partisans. Because no one can be sure how successful those drives will be, pollsters may have more difficulty than usual determining who in fact is a "likely voter." This uncertainty may be especially challenging -- and confusing to readers -- at the end of the campaign when pollsters attempt to focus most stringently on likely voters.

Next week on Reporter's Notebook: Jeanne Cummings, political correspondent of the Wall Street Journal, talks about money and politics. Click here to read the article.