An official website of the U.S. government
U.S. flag
This site is in beta. We're updating this site on a rolling basis and we'd love your feedback! Contact us here.

Contracts over Time

How We Conducted This Analysis

This analysis was conducted using government-wide contract acquisition and spending data. An additional dataset was compiled identifying the timing of continuing resolutions and new appropriations enacted by Congress from fiscal year 2007 through the first quarter of 2018. Contract spending data for all agencies was used for the same time period.

Several models were estimated to determine if a relationship could be identified between congressional budget actions and contract spending across the government. The primary model estimated was a semi-log linear model which was used to estimate the effect that continuing resolutions and new appropriations had on total contract spending. The outcome measured was total contract spending by week. The model included several additional predictor variables including lagged values of total weekly contract spending and total yearly obligations government-wide, which was used to proxy the trend in overall spending over time. Due to the nature of the data certain corrections had to be implemented to ensure in the integrity of the estimates and results, this includes use of logged values for total spending and total obligations. Additionally, in the calculation of hypothesis test results, the Newey-West estimator was used to correct estimate standard errors for heteroskedasticity and autocorrelation.

Additional models were also estimated to examine more closely subsets of the contracts spending data and determine if the effects of congressional budget actions differed among specific subsets of contracts. Models were estimated to evaluate how passage of continuing resolutions and new appropriations affected the value of new contracts awarded and the value of contract modifications awarded. These models also used a semi-log linear model format and included total obligations as a proxy for yearly trend in spending.

Models were also developed to examine how congressional budget actions affected the specific types of goods and services that the government issued contracts to purchase. Contracts data was separated into one of seven product and services groups;

  • Facilities, Equipment and Construction;
  • Professional Services, Education and Training;
  • Information Technology and Electronics;
  • Miscellaneous Supplies and Equipment, Clothing and Textiles;
  • Transportation and Logistics Services;
  • Research and Development; and
  • Weapons and Ammunition.

Models were estimated to evaluate the impact of congressional budget actions on each group. These models followed the same structure, using a semi-log linear model. Not all models showed the same residual heteroskedasticity and potential for residual autocorrelation and therefore White-Huber or Newey-West corrected standard errors were used as appropriate to test the significance of the model results.

The model specifications and data transformations necessary to execute the analysis are discussed in more detail here: Methodology White Paper