מסמך טכני מיומן המחקר · באנגלית, כפי שנכתב בשלב העבודה
Step 288: quarterly Israeli fixed investment excluding dwellings
The matched panel covers 2014Q1–2023Q4. GDP and fixed investment are from tables 35 and 38 of CBS National Accounts 1995–2023, with the Hebrew print copy used to resolve two differences in the searchable transcription. Investment and nominal GDP are in current NIS millions. Real GDP is chained at 2015 prices. Unemployment is the OECD quarterly seasonally adjusted rate via FRED, series LRUNTTTTILQ156S. Nominal quarterly investment and GDP reconcile with the v287 annual inputs to within NIS 1 million per year. The published quarterly real GDP values are used as printed.
The outcome is the change in the investment share of nominal GDP. Nonhousing investment subtracts residential construction from total fixed investment and includes government investment outside dwellings. The baseline uses four-quarter differences, with contemporaneous and one-year lagged changes in unemployment and real GDP log growth. The alternative uses one-quarter differences with one-quarter lags and quarter indicators for seasonality. Both report the sum of the two unemployment coefficients, with approximate 95% Newey–West intervals using four quarterly lags. The coefficients are conditional associations at different horizons.
| Changes and endpoint | Observations | Total fixed investment | Excluding dwellings |
|---|---|---|---|
| Four quarters, through 2023Q4 | 32 | 0.19 [−0.81, 1.18] | −0.29 [−0.89, 0.30] |
| Four quarters, through 2023Q3 | 31 | −0.34 [−0.89, 0.22] | −0.47 [−1.00, 0.07] |
| Four quarters, through 2019Q4 | 16 | −1.57 [−2.40, −0.74] | −1.33 [−2.20, −0.47] |
| One quarter, through 2023Q4 | 38 | 0.34 [−0.72, 1.39] | 0.22 [−0.48, 0.92] |
| One quarter, through 2023Q3 | 37 | 0.11 [−0.56, 0.79] | 0.14 [−0.46, 0.75] |
| One quarter, through 2019Q4 | 22 | 1.00 [−0.54, 2.55] | 1.16 [−0.41, 2.72] |
Full-sample nonhousing slopes differ in sign by horizon. The pre-2020 four-quarter estimate is much more negative, but it uses 16 overlapping observations. The endpoint checks do not identify the effects of the pandemic or the war. The sample and horizon sensitivity leave the Israeli calibration unresolved. A longer matched series would allow the horizon comparison across additional business cycles.
Reproduce with python scripts/israel_investment_quarterly_v288.py.