Despite increasing interest for the carbon footprint of higher education institutions, little is known about the carbon footprint associated to research activities. Air travel and attendance to conferences concentrate recent data and debates. Here we develop a hybrid method to estimate the greenhouse gas emissions (GHG) associated to research purchases. To do so, we combine macroeconomic databases, research-centered companies footprints and life-cycle analysis to construct a public database of monetary emission factors (EF) for research purchases. We apply such EFs to estimate the purchases emissions of hundreds of research laboratories in France gathering more than 20000 staff, from all disciplines. We find that purchases dominate laboratory emissions, with a median of 2.3 tCO2 e/pers, accounting for more than 50% of emissions, and 3-fold higher than the individual contribution from travel, commutes and heating. Electricity emissions are 5-fold lower in our dataset of laboratories using low carbon electricity but they become preponderant for high carbon electricity mixes (3.5 tCO2 e/pers). Purchases emissions are very heterogeneous among laboratories, but are strongly correlated with budget, with an average carbon intensity of 0.33+/-0.07 kg CO2e/euro and differences between research domains. Finally, we quantify the effect of a series of demand-driven mitigation strategies obtaining a maximum reduction of 15 % in total emissions, suggesting that effectively reducing the carbon footprint of research activities calls for systemic changes.