The modeling of nonmotorized travel demand has mostly been conducted at the large spatial level (e.g., city, county, or census tract level) by using data from the Bureau of the Census and the National Household Travel Survey. This paper introduces a modeling approach for estimating the mode share of nonmotorized trips by using data from multiple sources at a finer spatial scale. The correlations between a number of socioeconomic, environmental, and infrastructural factors and the nonmotorized share of the daily commute are analyzed at the level of the census block group. A neighborhood analysis concept is developed to take the length of non-motorized trips into consideration. Multiple regression analysis shows that employment density, the percentage of the student population, median household income, and average sidewalk length together provide the strongest power for prediction of the nonmotorized mode share. The potential applications of the methodology and the implications for data collection are also discussed.