John Fleck found a great paper by Karl T. Ulrich on the ultimate energy costs of using bicycles versus cars. The bottom-line argument is that cycling saves energy, but because you live longer your lifetime emissions will be greater. Such fun. However: Those who adopt the bicycle as a means of transportation could potentially develop an increased awareness of the environmental impact of their actions and may over their lifetimes reduce energy consumption substantially in their other, non-transportation activities.

Some tit-bits:

• Despite the inefficiencies in the energy conversion processes to generate human power, the bicycle remains about 6-9 times more efficient per km traveled than the single-occupant automobile.
• Warburton et al. (2006) and Lee and Skerrett (2001) review the literature on the dose-response relation between volume of physical activity and all-cause mortality. Several dozen studies provide clear evidence that energy expenditure of about 4200 kJ/wk (1000 kcal/wk) is associated with a 20-35 percent reduction in risk of all-cause mortality.
• Fatalities due to accidents while bicycling present mortality risks relative to automobile use. Pucher and Dijkstra (2003) estimate that the fatality rate for bicycling in the U.S. is 72 fatalities per billion kilometers, whereas for automobile travel it is just 6 fatalities per billion kilometers.
• I estimate the change in average
  longevity that would result from a decrease in mortality due to increased physical activity and an increase in mortality due to the increased accident risk from bicycling… Based on this analysis… each year of bicycling increasing longevity by an average of 0.029 years (10.6 days),
• Spillman and Lubitz (2000) find that increasing longevity does not change fundamentally the pattern of end-of-life healthcare requirements. Instead, an increase in longevity tends to extend the healthy “middle years” of life. I infer from this finding that increasing longevity due to physical activity would not change fundamentally the pattern of energy use at the end of life, but essentially also extend the middle years of energy use.
• a year of bicycling reduces energy use for transportation substantially, -11 GJ for the US context and -7 GJ for
  the OECD context. However, for each year of bicycling, longevity is increased by 0.029 years (10.6 days) which corresponds to an increase in lifetime energy use of 10 GJ for the US
• As a society, we value longevity more than long-term environmental impact. If we did not, we might provide incentives for risky behaviors such as smoking, drug abuse, and driving without seat belts.
• Given that each person-year of human-powered transportation by a previously sedentary individual offers an average increase in longevity of 0.029 years, we might ask how efficient human-powered transportation could be as a public health instrument. At a value of 60,000 USD/QALY, 0.029 years of additional longevity should be worth 1740 USD. Could individuals be converted from a sedentary lifestyle to the use of bicycles for transportation at a cost of less than 1740 USD/yr?

Most amusingly: The overall system energy consumption of a personal electric vehicle such as an electric scooter or electric bicycle is very similar to that of a human-powered bicycle (Ulrich 2006), and yet the personal electric vehicle does not increase longevity. In fact, an open vehicle like an electric bicycle is likely to increase fatality risk without providing offsetting health benefits. As a result, transportation by personal electric vehicle offers substantial environmental benefits, but not the health benefits of human-powered transportation.