In a comment thread below, Tom linked to a fine article by Kevin D. Williamson at the National Review Online, which I thought should be highlighted here. Williamson cites three areas where conservatives fail to engage the middle-of-the-road voter: (1) the best way to address risk, (2) the real value and dangers of economic inequality, and (3) how to rely on growth instead of on redistribution of a finite pie. On the first point, he reminds us that segments of the population who historically were systematically excluded from the formal economic system will be hard sells on the notion that accepting economic risk is the best path to prosperity; we'll have to acknowledge their legitimate suspicion of the game.
Regarding inequality, he cautions against arguing that "merit and merit alone accounts for the diverging prospects of the very well off and the rest." A free market doesn't ensure that merit will triumph, only that individuals' preferences will have more clout than those of bureaucrats. A conservative's desire to favor individuals over bureaucrats doesn't rest on a conviction that all individuals are better judges than any bureaucrat. It rests in part on a philosophical preference for individual autonomy, and in part on an empirical conviction that, although masses of individuals can make appalling choices, their inevitable failures pale before the even more appalling choices of bureaucrats.
On the subject of growth vs. redistribution, Williamson points out that the "people as useless mouths to feed" cant of Malthusian liberals sometimes raises its ugly head equally in the hearts of conservatives who back trade barriers and oppose immigration. He recommends a focus on people as the engines of future growth and prosperity, and on the education and healthcare policies most likely to make that possible.
He closes with an encouraging look at recent conservative reforms in Sweden, all achieved without outraging the compassionate or liberal instincts of most voters in that very collectivized state.