Introduction Clinical registration of the cause of death is available for less than one-third of the global newborn deaths, but the need for good quality data on causes of death for public health planning has renewed the interest in the Verbal Autopsy (VA). We aimed to determine the cause of neonatal deaths by VA in Morang district of Nepal. Methods Caretakers of the deceased were interviewed using a semi-structured VA questionnaire by female community health volunteers. The cause of death was assigned by two senior paediatricians independently and disagreements in ascertaining the proximate cause of death were resolved by consensus. Results The proximate causes of deaths were infections (41 %), birth asphyxia (37.2 %), prematurity (11.5 %), and low birth weight related causes (6.9 %). There was no significant statistical difference in deaths due to infection seen in non-institutional deliveries (43.5 %) than institutional deliveries (34.6 %). More than half of the deaths (58.5 %) occurred within the first three days of life where the predominant cause of death was birth asphyxia (60.7 %). Conclusions Analysis of verbal autopsies demonstrates that the major causes of death still are infections and birth asphyxia. The timing of deaths suggests that neonatal interventions should be aimed at the first week of life. There is no comparative advantage between institutional deliveries at below district level institutions and non-institutional deliveries to prevent neonatal infection. Thus, further study on the quality of care at institutes below the district level should be conducted. Disparities still occur in deaths, with most deaths in Morang occurring in non-institutional deliveries and in disadvantaged groups.