A before-and-after analysis was performed at eleven intersections where a bike box was installed in Portland, Oregon to explore the safety effects of the treatment. Video data were gathered prior to installation at 14 intersections where a bike box installation was planned by the Portland Bureau of Transportation. Cameras were set up to capture three full twenty-four hour days (72 hours) of data for each intersection from Tuesday through Thursday. Of the 14 original selected intersections, 11 intersections actually received the bike box treatment. Video data were again gathered for these intersections after the installation of the bike box for another three full twenty-four hour days (72 hours) between Tuesday and Thursday. One day of data (24 hours) was selected for observation from both the before and after periods in the analysis for each study intersection during midweek. Safety effects were evaluated by three metrics: 1) observed conflicts; 2) observed cyclist behavior for all conflicts as measured by head or shoulder checks; and 3) reported crash data. To develop the conflict data, a log was created of each motor vehicle and bicycle passing through the intersection for approximately 528 hours of video. All conflicts that were observed during the period were further reviewed by an expert panel that scored conflicts by severity. Following this review, a total of 18 conflicts were observed during the before period. The total exposure in the before period was 39,497 motor vehicles in the vehicle lane adjacent to the bike lane (10,454 of which were right-turning) and 7,849 bicycles. A total of 19 conflicts were observed during the after period. Total exposure was 42,381 motor vehicles in the vehicle lane adjacent to the bike lane (11,053 of which were right-turning) and 5,852 bicycles. The sample size of observed conflicts was insufficient to draw statistically significant conclusions for any of the specific intersections that were treated. When taking in account the total amount of conflicts, the limited data suggest a slight increase in the rate of conflicts when normalized against a product of right-turning vehicles and bicycles observed in the intersection. The data also suggest that the installation of a bike box at an intersection reduces the rate of conflicts per hundred motor vehicles and increases the rate of conflicts per hundred bicyclists. Data regarding head-checks from the bicyclist shows an increase in bicyclists observing the possibility of conflicts approaching from behind as they pass through the intersection. A review or crash data at each of the intersections shows an increase at three of the observed intersections and a decrease at the remaining five.