They are given parameters by the ones who make the request. They are told what to include in their reports, what to omit, and often told which sources they can use as a basis for them.
The CBO is not allowed to broaden the scope of a report, even if in doing so, it would significantly alter their findings.
Although CBO is given the specific question to answer, CBO is not told how to answer that question. CBO's response is independent.
That's why their reports include so many "what if's" and explanations for their findings.
CBO, like any entity trying to make a forecast, has to work with incomplete information, make assumptions about the current and future environment, etc. Laying out scenarios and mentioning variables that might have a significant impact makes good sense.
We saw this with Obamacare. Their reports prior to it's passing that were requested by the administration and democrats on capitol hill, painted a very rosy economic picture for America if the legislation were enacted. After the bill was passed, they were requested to do several other, more broad studies on the Obamacare bill, and every one that has been released so far, has contradicted or greatly reduced the positive benefits they reported initially. That's because their scope was severely limited before by the administration.
As more information becomes available, initial ideas can change. Long-range forecasting is very difficult and imprecise. For example, if one is trying to project the budget deficit for 2020, one needs to have a reasonable idea concerning the macroeconomic environment.
What will the GDP be in 2020? Historically, one can assume at least one and perhaps two additional recessions from now until then. But how deep? Keeping current law fixed, one could then estimate revenue and expenditures? But what if there is a geopolitical shock? Certain expenditures might rocket well beyond what current law would have them. What will interest rates be? Given the nation's sizable and growing debt, one needs to understand the structure of the nation's debt and relevant interest rates in order to assess interest payments.
If one looks closely at CBO's forecasts, one finds that beyond the short-term, CBO reverts toward what is believed to be stable state figures i.e. nominal GDP grows by 4.4%, 4.3%, and 4.3% in 2018-20; Real GDP grows by 2.3%, 2.2%, and 2.2% in that timeframe; Core CPI is at 2.0% each of those three years; the unemployment rate is at 5.0% during those three years, etc.
Why? Because errors are likely to be smaller if one assumes a stable state toward what are believed to be sustainable figures.
Nonetheless, sizable errors could still result. For a simple illustration, let's say tax revenue amounts to about 18% of nominal GDP under the current tax law framework. Nominal GDP was about $14 trillion in 2009. Over the past 20 years, nominal GDP expanded at an average rate of 4.8% per year.
What would happen to tax revenue if annual growth amounted to +/- 0.5% from the above average?
5.3% annual growth: $4.45 trillion tax revenue
4.8% annual growth: $4.22 trillion tax revenue
4.3% annual growth: $4.00 trillion tax revenue
In short, average annual growth of +/- 0.5% from the 20-year annualized figure would create a difference of $450 billion in 2020 revenue.
Needless to say, CBO's models are more complex. Nonetheless, assumptions toward a stable state are made as one extends farther out along the time horizon.
In reality, CBO has erred on the side of optimism (lower net costs/greater net benefits/policy makers will deliver on program constraints). That is a human bias, not something that is unique to CBO. In practice, that has led to CBO's often underestimating the growth of major programs. The recent "doctor's fix" in which both parties overwhelmingly voted to waive a very modest budget-saving mechanism provides just the latest example of policy makers choosing not to be bound by fiscal constraints. Hence, an approach that followed the current law would have missed the extra increase in the budget deficit that resulted under a bipartisan move to avoid modest fiscal constraints.
Indeed, given past history, CBO should probably print a disclaimer highlighting budget constraints under existing law and explaining that policy makers have typically failed to remain bound by those constraints.