# Step 287: Israeli fixed investment excluding dwellings

This pilot separates dwellings and general government on the same annual sample. The inputs in `data/processed/israel_investment_cbs_v287.csv` are transcribed from [CBS National Accounts 1995–2023](https://www.cbs.gov.il/he/publications/DocLib/2025/1969/e_print.pdf), tables 1 (GDP), 8 (total and residential fixed investment), and 8B (general government, nominal row, cross-checked against table 24). Inputs are NIS millions. The printed tables used here cover 2014–2023. Unemployment comes from the archived v282 WDI panel (ILO modeled estimate). Run `python scripts/israel_investment_v287.py` to reproduce.

For each outcome, regress its annual change in percentage points of nominal GDP on current and lagged unemployment changes and current and lagged log growth of real GDP, with an intercept. Report the sum of the unemployment coefficients, with the v282 approximate Newey–West interval (two lags). Two lags of levels are needed, leaving eight matched annual observations (2016–2023). Leave-one-out ranges refit the coefficients after dropping each outcome year in turn while preserving the original lag definitions.

| Outcome | Slope | Approximate 95% interval | Leave-one-out range |
|---|---:|---:|---:|
| Total fixed investment | −0.43 | [−1.89, 1.04] | [−2.06, 1.46] |
| Fixed investment less dwellings | −0.40 | [−1.83, 1.02] | [−1.95, 1.45] |
| Total less dwellings and general government | −0.21 | [−1.11, 0.70] | [−1.19, 0.95] |

Dwellings are an asset category and government is an institutional sector. The last row subtracts both aggregates, so any government investment in dwellings would be subtracted twice. It is an arithmetic residual and may include nonprofit and other nonbusiness investment. Subtraction is applied only to current-price data.

Removing dwellings barely changes the point estimate. The further subtraction of government investment attenuates the estimate, but all signs can reverse when dropping a single year. With only three residual degrees of freedom in the baseline, the reported intervals are diagnostic. No Israeli model parameter is selected from this sample.

For a longer replication, use the [OECD annual GFCF by asset](https://data-explorer.oecd.org/vis?df%5Bag%5D=OECD.SDD.NAD&df%5Bid%5D=DSD_NAMAIN10%40DF_TABLE1_EXPENDITURE_GFCF_ASSET&df%5Bvs%5D=2.0&lc=en) for a harmonized dwellings subtraction, and verify actual Israeli coverage. Identifying business investment separately requires an asset-by-institutional-sector cross-tabulation or a direct series with a documented sector boundary.