The various dimensions of security threats in Nigeria pose diverse risks to investments. Cases of kidnapping have gained prominence in Nigeria (see Figure 2). People have been kidnapped all over the country recently (Ikezue, 2023). The most celebrated kidnapping was the 276 Chibok girls that were abducted from their school in Borno State, Nigeria in 2014 by the Boko Baram terrorist group. Since then, there have been a number of other significant kidnappings. The same terrorist organisation abducted 110 Dapchi girls in March 2018. After that, criminal groups began kidnapping people at random to extract ransom from the victims or their families.

According to the United Nations Development Programme (UNDP), Northeast Nigeria's insurgencies had taken the lives of nearly 350,000 people as of the end of 2020 (UNDP, 2021). Tanko (2021) reported that Boko Haram (a terrorist group in Nigeria) had seized several territories and levied taxes on agricultural products, they now control the once, booming international fish market in the Chad Basin. The challenge is made more difficult by Nigeria's ungoverned spaces--remote areas that are largely ignored, where various militant groups have seized to torment rural communities without fear of reprisal (Awa, 2020; Ebipre & Wilson, 2020). In the Southeast, the Indigenous People of Biafra (IPOP) gave a sit-at-home order (every Monday) to demand the release of its leader, Nnamdi Kanu, from prison, resulting in the loss of lives, properties of residents and businesses (Odeniyi, 2023). The violent agitators for resource control in the Niger Delta include the activities of various militia groups such as the Movement for Emancipation of the Niger Delta (MEND), the Ijaw Youth Council (IYC), Mgbonyenbi and Emeni (2020) likened these struggles to the ‘resource curse’, the tendency of natural resource reliance to hinder growth, increase inequality and poverty for a larger majority of the populace. Figure 3 depicts the number of casualties from the various security challenges between 2014 and 2021 in Nigeria.

Source: International Centre for Investigative Reporting (ICIR) https://www.icirnigeria.org

In response to the rising state of insecurity in Nigeria, the government has been allocating much of its resources to defence and security-related expenditure (refer Figure 4). These allocations represent large sums of money that should have been invested in key economic sectors like education and health (Onyele & Ariwa, 2020). The exorbitant defence expenditure, which primarily relies on foreign procurement, has a negative economic impact and further devalues the Naira (Ayodele & Tomisin, 2021). The country has also witnessed how agricultural produce in Benue (the food basket of the nation) and a number of other states has been badly depleted by the incessant herdsmen-farmers clashes. This protracted security challenge has affected businesses who that, due to the uncertainty created by insecurity, put off investment decisions in Nigeria. Under such an economic climate, creativity is discouraged as innovation takes a back seat despite being a driving force for economic advancement, leading to a high rate of brain drain through mass emigration to western countries for safety (Aderemi, 2019). As such, the government’s inability to put an end to the security challenges has created a feeling of hopelessness and helplessness, especially in the areas that are vulnerable (Omole, 2020).

Source: https://www.cfr.org/nigeria/nigeria-security-tracker/p29483

𝑟𝑡𝑛𝑒𝑜 = is the domestic interest rate associated with capital per worker ratio

𝐴𝑡 𝑓 ′ (𝑘𝑡 ) = is the net depreciation of marginal product of capital per worker.

Imagine that capital is completely movable across national boundaries, allowing enterprises to tap into international savings without home savings restricting investments. Here, the scenario of a small open economy dealing with an indefinitely elastic supply of capital at the global interest rate (rw) is considered. Foreign capital would flow into a nation if the initial level of the capital stock to labour ratio's related rate of return on capital was higher than the world interest rate. Neoclassical theory, which is based on liberal capital movements, states that significant capital transfers from industrialised to developing countries should be seen as a result of the capital's diminishing returns (Solow, 1956).

Contrarily, Lucas (1990) noted that capital moved from developing to developed nations, refuting the neoclassical postulation. The literature argues that the neoclassical assumptions do not apply in practice because market failure and distortions may prevent efficient capital allocation, which explains the paradoxical behaviour of capital flows. As a result, the capital influx to poor and developing economies is less than the neoclassical model can account for. In the words of Alfaro et al. (2008), theoretical explanations for the Lucas paradox are divided into those that emphasise flaws in the capital market and those that focus on basic production issues. Numerous academics have focused on Lucas' claims, and an extensive empirical study on the capital flow allocation conundrum has been conducted. Lucas (1990) outlined his justification using the following one-sector model. Let equation 3 represent the production function, with the variables labor (L) and capital (K) standing for the output:

Also, let 𝑝 represent the price of goods, and 𝑟 𝑎𝑛𝑑 𝑤 denote returns on capital and labour, respectively. Then, firms profit maximisation gives:

The domestic price is equalised between nations with unrestricted trade flow. According to the theory of diminishing marginal products, countries with lower capital intensity should have higher returns (r). Lucas gave the example that the return on capital in developing nations needs to be larger than that in developed nations. Lucas argued for significant capital transfers from rich (developed) to poor

(developing) countries with a return difference in favour of the latter. The Lucas paradox refers to the fact that this claim is false in practice. Several approaches, such as seeing human capital as a novel production element and taking sovereign risk into account, include thinking of a worker in a rich country as effectively equivalent to many workers in a poor country (Ju & Wei, 2006). The dynamics of macroeconomic elements and security threats, however, may be important aspects that support the Lucas paradox, according to further studies.

There are several empirical works on the subject discussed in this paper. Many of these studies provide evidence as to why sufficient capital is not flowing from rich countries to poor countries, as predicted by Lucas's (1990) paradox. The review contains studies that show how macroeconomic conditions affect capital inflows, while the others show how insecurity affects capital inflows. The major gap found in the empirical review is the fact that most of the Nigerian studies reviewed did not consider the effect of insecurity, and the few studies like Ayoola (2022); Essien et al. (2015) and Igbadoo et al. (2023) that looked at the effect of insecurity did not consider the macroeconomic perspective in their analysis. Hence, the present study considered the collective effect of insecurity and macroeconomic variables on capital flows to Nigeria. Also, most of the studies for Nigeria largely considered FDI, which is only an aspect of inward capital flows1: hence, the current study focused on the aggregate figure of inward capital flows to Nigeria. A summary of these studies highlighted in Table 1.

This study used quarterly data spanning from 2014Q1 to 2021Q2 to assess the relationship between macroeconomic dynamics and inward capital flows to Nigeria in the face of security challenges. The justification for using quarterly data was hinged on the fact that the incessant Nigerian security challenges were more pronounced in 2014 with the adoption of the Chibok girls in Borno State. Therefore, we used quarterly data to ensure sufficient observations between 2014 and 2021, which aligned with the dynamics of time series analysis.

The predictor variables were insecurity and macroeconomic variables such as the exchange rate, industrial production capacity, consumer price index, and loan rate. The predicted variable (the dependent variable) was the total inflow of capital. Data for inward capital flows was extracted from the National Bureau of Statistics' (NBS, 2021) Quarterly Capital Importation Report (various issues), the data for insecurity (number of deaths) was sourced from the Armed Conflicts Locations and Events database, and data for exchange rate, industrial production capacity, consumer price index, and lending rate were sourced from the Quarterly Economic Report (2021) and CBN Statistical Bulletin (2021). Table 2 contains the list of data sources and their descriptions.

To ensure reliable and consistent empirical results, the data were transformed into a natural logarithm (LN) which was applied to solve the potential problems of time and growth. The functional relationship of the models was specified thus:

LNCPF = f(LNINS, LNEXR, LNIPC, LNCPI, LNLDR)

Before the model execution, it was mandatory to scrutinise the time-series properties of the data. As a result, the stationarity condition of the data set was ascertained using the Augmented Dickey-Fuller (ADF), Phillips-Perron (PP) and Kwiatkowski-Phillips-Schmidt- Shin (KPSS) approaches for unit root testing (Zhong, 2015). As some variables were stationary at I(0) and others are I(1), the study was directed to the application of the Autoregressive Distributed Lag (ARDL) bounds test (Pesaran et al., 2001). The ARDL model was preferred over other traditional estimation methods of for testing cointegration for some reasons. Firstly, this estimation approach technique can be applied when variables are mixed between I(0) and I(1) levels of integration. Secondly, the ARDL method allows for simultaneous estimation of the model's short-run and long-run linkages. Additionally, the ARDL model considers endogeneity problems by adding lags of predicted as well as predictor variables to the model. The ARDL model was specified as: 𝑛1 𝑛2 𝑛3

Where, 𝜀𝑡 stands for the white noise term and LNCPF (natural log of capital inflows), LNINS (natural log of insecurity), LNEXR (natural log of exchange rate), LNIPC (natural log of industrial production capacity), LNCPI (natural log of consumer price index), and LNLDR (natural log of lending rate) are the model's explanatory (predictor) variables. The bounds test technique for cointegration requires the ARDL model to undergo an F-test with appropriate lag lengths. The Akaike Information Criterion limited the model to a maximum lag duration of three (3) days. (AIC). To confirm the existence of cointegration, the conventional F-test was performed, which displayed two sets of crucial values, i.e., lower and upper bound values (Pesaran et al., 2001). The lower and upper critical values capture the assumption that all variables are either I(0) or I(1) and none is I(2) which makes the ARDL estimation technique most appropriate. Hence, the critical values provided a restricted bound that incorporates all possible categories of the variables. The null hypothesis of no cointegration (i.e., no long-run relationship) was rejected if the upper bound critical value lies below the F-statistic generated from the bounds test, and vice versa, the test fails to reject the null hypothesis. On the other hand, the test is deemed inconclusive if the F-statistic value falls in between the lower and upper bounds. Once a long- run relationship (cointegration) is ascertained, the short-run dynamics are captured by transforming Equation (6) into an error correction model (ECM) as follows: 𝑛1 𝑛2 𝑛3

𝐸𝐶𝑀𝑡−1 indicates the speed of adjustment, which measures how quickly a model returns to long-run equilibrium following the occurrence of short-run shocks that cause disequilibrium (Shrestha & Bhatta, 2018). The ECM's sign must be negative, statistically significant, and have a coefficient between -1 and 0, signifying rapid and flawless convergence following a shock to the mechanism, to guarantee long- run convergence.

Model diagnostic tests need to verify the validity of some important ARDL assumptions, such as serial independence, homoscedasticity, and normal distribution. Therefore, the Jarque-Bera test was used to test for normality, the Breusch-Godfrey serial correlation LM test was used to check for serial independence, and the ARCH test was used to check for heteroskedasticity in the model. To determine whether there was model misspecification, the Ramsey reset test was used. To determine whether the model parameters are stable, the recursive CUSUM and CUSUM of squares are applied (Turner, 2010). In the analysis, the outcome of the diagnostic tests showed that the ARDL model was free from problems associated with serial correlation, heteroskedasticity, abnormal distribution of the residuals, and model misspecification. The ARDL technique encapsulates the dynamics of the short-term connection between variables, and the distributed lag component represents the lagged values of the explanatory variables. The ARDL model captures the lagged effects of the explanatory variables on the dependent variable. This paper also applied the Modified Wald test (MWALD), as recommended by Toda and Yamamoto (1995), to give an understanding of the direction of causation among the variables, with regard to the relationships between macroeconomic dynamics, insecurity, and inward capital flows. The Toda and Yamamoto (1995) approach uses a standard vector autoregression (VAR) model and level series, implying that there is minimal risk of wrongly identifying the order of integration of the quarterly time series data under consideration.

To show the statistical properties of the data, the descriptive statistic presented in Table 3 was deemed necessary. The series, especially NINS, IPC and LDR showed moderate variation from their mean values as seen by the standard deviation, suggesting characteristics of an abnormal distribution.

The skewness results confirmed this, revealing that the aforementioned variables' skewness values significantly exceeded the threshold (0) average. This is an indication that the distribution of the series was highly skewed, either negatively or positively, and nearly abnormal, with the exception of CPF, EXR and CPI, which were only slightly skewed. The Jarque-Bera test revealed that while INS, IPC, and LDR were not normally distributed, CPF, EXR, and CPI all support the null hypothesis of normal distribution. The fact that the variables were not normally distributed led to a natural logarithmic transformation of the data to make the moderately skewed data more normally distributed (to achieve constant variance).

A system of Vector Autoregression (VAR) was formed to generate the optimum lag. The ideal lag was three (3), which was in line with all the selection criteria such as AIC (Akaike Information Criteria), SC (Schwarz Information Criteria) and HQ (Hanna-Quinn Information Criteria). As a result, the optimal lag length of three was suggested by all the information criteria.

The ARDL bounds test for cointegration was also carried out. At the 1 percent (5.230000) level, the reported F-statistic (5.998478) was greater than the upper bound, or I(1). As a result, the cointegration precondition was established, and the null hypothesis of no cointegration (i.e., no long-run link) was rejected. It was then ascertained that there was a long-term connection between inward capital flows, macroeconomic dynamics, and insecurity in Nigeria. The implication of cointegration was the presence of an equilibrium correction representation of the variables. The existence of cointegration also clarified the problem of spurious regression, in which intrinsically unrelated time series data are highly correlated. The number(s) of lag(s) chosen for LNCPF, LNINS, LNEXR, LNIPC, LNCPI, and LNLDR, respectively, were implied by the ARDL specification of (1, 1, 2, 1, 2, 1).

The ECM confirmed that capital inflows, macroeconomic dynamics, and insecurity were linked in a long-run equilibrium path. The results of the ECM and short-run estimates were captured in Table 4. The ECM coefficient of the ARDL equation turned out to be -0.471775, meaning that the speed of adjustment of lags from previous period’s errors was approximately 47 percent before the long term variables converged to the long-term equilibrium path. This suggests that the current quarter has remedied approximately 47 percent of the inward capital flows (CPF) disequilibria from the previous quarter. Thus, further indicated that it took approximately two (2) quarters (𝑡ℎ𝑎𝑡 𝑖𝑠 1⁄𝐸𝐶𝑀) for inward capital flows to re-adjust to long-run equilibrium after short-run shocks arising from macroeconomic dynamics and insecurity issues.

The ECM results also indicated that in the short run, LNCPF was negatively and significantly affected by LNINS, implying that an increase in the number of deaths arising from insecurity had an immediate diminishing effect on inward capital flows. Also, a period lag of exchange rate, denoted as D(LNEXR(- 1), had a negative and significant effect on LNCPF, meaning that a rise in the naira – dollar exchange rate in the previous period caused a decreasing effect on the current period’s inward capital to Nigeria. Again, the short-run coefficient of LNIPC was found to be positive and non-significant, indicating that an increase in industrial production capacity had a marginal short-term effect on inward capital flows. LNCPI was seen to have had a positive and statistically significant lag effect on LNCPF, implying that the previous period’s increases in industrial production capacity attracted inward capital to Nigeria in the current period. The negative and statistical significance of LNLDR suggested that an increase in the lending rate discouraged the inflows of capital in the short run.

The adjusted R-squared value for the model's overall goodness of fit was 0.750289, which meant that the predictor variables (LNINS, LNEXR, LNIPC, LNCPI and LNLDR) collectively accounted for around 75 percent of the variation in the predicted variable (LNCPF). Based on the probability, the adjusted R-squared value was 0.750289 at a 5 percent significance level. (F-statistic) of 0.000023, it was implied that the collective effect of the independent variables (LNINS, LNEXR, LNIPC, LNCPI and LNLDR) on the dependent variable (LNCPF) was statistically significant. This indicated that the interactions between macroeconomic variables and insecurity had a significant influence on Nigeria's inward capital flows.

The long-run estimated coefficients of the model and residual diagnostic tests are shown in Panel A of

Nigeria during the previous quarter had a significant impact on capital inflows in the current quarter. The long-run coefficient of LNINS was found to be negative and insignificant, implying that a 1 percent increase in insecurity caused inward capital to decrease marginally by 5.3 percent. Also, the long-run coefficient of LNEXR turned out positive and statistically significant, implying that a percentage increase in the naira-dollar rate led to a considerable change of approximately 5.5 percent in inward capital. The estimated coefficient of LNIPC in the long-run was positive and significant, implying a percentage increase in the industrial production capacity caused a considerable increase of approximately 5.7 percent in inward capital flows. The long run coefficient of LNDR was found to be negative and statistically non-significant, implying that a percentage increase in lending rate caused inward capital flows to decrease marginally.

Source. EViews Output. Note. *** denote 1% significant level

The results of the diagnostic tests in Panel “B” of Table 5 suggest that the ARDL model's residuals were uncorrelated and homoscedastic. Also, since the model passed the Ramsey RESET test, the null hypothesis was accepted, and the long-run ARDL coefficients had little or no misspecification error. The Jarque-Bera test confirmed the normal distribution of the residuals in the ARDL model.

In the long-run and short-run, insecurity exhibited a diminishing effect on inward capital flows to Nigeria. This was in consonance with the earlier established fact that heightened security challenges causes investors to divest to other countries that are more secure. However, the effect of insecurity was significant in the short run but non-significant in the long run. This implies that a sudden rise in insecurity had a considerable instantaneous impact on capital flows to Nigeria, but its non-significance in the long run could mean that foreign capital flows to Nigeria could be influenced by other factors such as macroeconomic dynamics over time. A plausible reason for this discrepancy could be that foreign capital flows to Nigeria were targeted at Nigerian states where insecurity was minimal. Also, this could imply that investors had a short memory of the security challenges, which explains the long­ run insignificance of insecurity. The observed negative effect of insecurity on inward capital flows aligned with prior studies such as Le et al. (2023); Ayoola (2022); Bardwell and Igbal (2021); Hogetoorn and Gerritse (2021); Essien et al. (2015) who had affirmed that a rise in insecurity hindered capital inflows.

Exchange rate in the short run did not have an instantaneous significant effect on inward capital flows, but its lagged effect was found to be negative and statistically significant, which varied with the long- run outcome of positive and statistical significance. The lagged effect implied that the effect of the exchange rate in determining the direction of capital flows was largely based on investors’ perceptions of previous period changes in the naira-dollar rate. This could also mean that foreign investors were at some point motivated to invest in Nigeria for other reasons, such as opportunities to make huge profits or for resource exploitation. This finding is in consonance with Dembo and Nyambe (2016), who found that the effect of the exchange rate on capital inflows varied with time.

With regards to industrial production capacity, it was realised that the variable maintained a positive coefficient in both the short-run and long-run but its significance was recorded only in the long run. This could imply that consistent improvement in industrial production capacity (which implies higher domestic productivity) could attract foreign inflows of investible funds in the long run, meaning that improvements in the production capacity of domestic industries could pull foreign investments. The findings of Nwagu (2023); Wijaya et al. (2020); Nwosa and Adeleke (2017) lend credence to this study.

Although the short-run coefficient of the consumer price index yielded a positive and non-significant effect on inward capital flows. However, its one-period lagged effect was negative and statistically significant, implying that investors’ perception of the previous quarter’s rise in domestic prices continued to diminish the potentials of attracting inward capital in the current quarter. This could also imply that the loss of purchasing power previously experienced by investors does not encourage them to invest more in the future. Studies such as Wijaya et al. (2020); Nwokoye and Oniore (2017); Waqas et al. (2015) have come up with findings that are in consonance with this study.

There was a negative and significant effect of the lending rate on inward capital flows in the short run but it turned out to be negative and non-significant in the long run. This indicated that a high lending rate generally causes a decline in inward capital flows, especially in the short run. The marginal effect of the lending rate in the long run could be attributed to monetary adjustments (like a downward review of the monetary policy rate) aimed at stabilising the macroeconomic to attract investors (Al-Smadi, 2018). The negative effect of the lending rate found in this study aligned with the studies of Nwagu (2023); Wijaya et al. (2020); Tellez-Leon and Ibarra (2019); Nwokoye and Oniore (2017) that an increase in the lending rate could lead to macroeconomic instability and reduce inward capital flows.

Compared to previous tests of causality, the Toda Yamamoto test is better since it may be applied in situations where the variables are cointegrated in a random order, the same order, or not at all. The ARDL co-integration bounds test, which illustrates the short- and long-term relationships between the variables, does not reveal the direction of the causal relationships. The estimated results are therefore not complete. This improves the estimation's results since it verifies the causal influence between the variables. The findings of the T-Y causality test confirmed the association between macroeconomic factors, inward capital, and insecurity, as indicated in Table 6.

Note. *** and ** denote 1% and 5% significant levels respectively