Remittances and economic growth in developing Asia and the Pacific countries

UNIVERSITY OF ECONOMICS HO CHI MINH CITY VIETNAM INSTITUTE OF SOCIAL STUDIES THE HAGUE THE NETHERLANDS VIETNAM – NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS REMITTANCES AND ECONOMIC GROWTH IN DEVELOPING ASIA AND THE PACIFIC COUNTRIES A thesis submitted in partial fulfilment of the requirements for the degree of MASTER OF ARTS IN DEVELOPMENT ECONOMICS By PHAM THI HANG Academic Supervisor: Dr PHAM KHANH NAM HO CHI MINH CITY, December 2014 Table of Contents List of Figures CHAPTER I: INTRODUCTION Introduction Research objectives CHAPTER II: LITERATURE REVIEW Remittance definition Remittance in Growth model 10 Consequence of remittances 12 3.1 Remittances and capital accumulation 12 3.2 Remittance and labor force growth 13 3.3 Remittance and total factor productivity growth 14 Factors effect economic growth 15 4.1 Remittances 15 4.2 Investment 20 4.3 Fiscal balance 22 4.4 Trade openness 25 4.5 Inflation 27 4.6 Population growth 30 4.7 Human capital formation 32 CHAPTER III: METHODOLOGY 35 Analytical framework 35 Estimation technique 39 Data source 41 CHAPTER IV: EMPIRICAL RESULTS 42 Overview of remittances in developing Asia and the Pacific countries 42 Remittances and growth 46 2.1 Non-parametric analysis 46 2.2 Parametric analysis 48 CHAPTER V: CONCLUSION 53 References 57 Appendix 66 List of Tables Table 1: Data description 37 Table 2: Remittances to developing countries, 2010 -2013 (US$ billion) .42 Table 3: Summary statistics of variables 46 Table 4: Correlation matrix .46 Table 5: Estimation results 52 Table 6: List of countries and remittances (share of GDP, 2000-2012) 66 List of Figures Figure 1: Remittances and other resource flows to developing countries Figure 2: Share of remittances by region in Asia and the Pacific countries, 2013 43 Figure 3: Growth rate of remittances by region in Asia and the Pacific countries 43 Figure 4: Top 10 remittance-receiving developing countries in Asia and the Pacific, 2013 44 Figure 5: Scatter plot of growth and remittances 47 ABSTRACT Over the past three decades, remittance inflows have increasing dramatically and become the main source of foreign exchange both in absolute terms and as a percentage of GDP in many developing countries However, the growth effect of remittance is still not well understood This study attempts to investigate the impact of remittance inflows on economic growth in developing Asia and the Pacific countries Moreover, it examines whether remittances can effect on the impact of labor and capital on growth in remittance-receiving countries The study uses a balanced panel data on remittance flows to 25 developing countries in Asia and the Pacific for the period 2000-2012 Endogeneity problem is controlled by system GMM estimator The results find no evidence suggesting the significant relationship between remittances and growth when remittance is considered as an explanatory variable in a standard growth regression Taking into account interaction terms, this paper comes to conclusion while population growth and remittances is complementary, human capital development and remittances are used as substitutes to promote growth CHAPTER I: INTRODUCTION Introduction Remittance is one of the most crucial parts of total international capital flows It is transferred through official and unofficial channels For instance, in 2013, official recorded worldwide remittance flows reached nearly $550 billion (World Bank, 2013) The unrecorded remittance flows is believed to be as large as 20 to 200 percent of total official remittance flows (Aggarwal et al., 2006; World Bank, 2006) International remittance inflows to developing countries are expected to increase to 8.4 percent in 2014-2016 and can reach to $516 billion in 2016 (World Bank, 2013) This forecast is calculated according to the outlook for GDP growth rate in key remittance-sending countries and remittance growth rate in the past In other words, the outlook for remittances remains optimistically Moreover, these flows are expected to get three times larger than official development assistance and become more stable than private debt and portfolio equity Figure 1: Remittances and other resource flows to developing countries Sources: Migration and Development Brief 22, World Bank Because of the dramatic increase in size, remittance flow has attracted scrupulous attention of academics and policy makers They have believed that the money the migrants send back to their relatives and friends in home country may impact that country macroeconomics conditions in many aspects For example, on the one hand, the migrants of skilled and educated labor raise serious doubts about the brain drain and effect on sustained economic growth of migrant-sending countries (Docquire & Schiff, 2009) On the other hand, the academic and policy circles believe that remittances can be seen not only as the main income for the low and middle-income households in developing countries but also as the crucial financial supply for domestic investment Another advantage of remittances is that it is sent directly to family and friends without government intervention, thus, it seems to be less volatile than other nontrade foreign currency inflows Moreover, remittances also expected to promote consumption and reduce cost of capital in recipients’ country Therefore, remittance inflows can have important implications for economic growth of recipient’s countries While there is vast literature on the effect of remittances on the development prospect in migrant-sending countries, empirical studies on this issue have been done at the worldwide level or for developing countries as a whole with mixed result For example, the study was done by Vargas-Silva, Jha and Sugiyarto (2009) pointed out that a 10 percent higher in remittances as a share of GDP leaded to a 0.9-1.2 percent higher in GDP growth Some other researchers argued that even though the impact of remittances on growth of receiving countries still depended on how this money was spent, and even when the households not use remittances for investment, remittances may have an important multiplier effect Lowell and De La Garza (2000) investigated that each one remittance dollar spent on additional consumption could encourage retail sales and further goods and services demand, and then helped to greatly stimulate growth and employment However, Straibhaar and Wolburg (1999) discovered that strongly dependence on remittance can encourage continuing migration of the working-age population, especially high- skilled Then, the welfare loss due to emigration cannot be compensated by Besides, if remittances provoke goods and services demand higher than the economy’s capacity, especially on non-tradable goods, remittances may cause inflation In Egypt, because of the massive rise in remittances, agriculture land price had increased by 600% between 1980 and 1986 (Adams, 1991) Finally, remittances may create negative impact on growth by existing significant moral hazard problems In particular, with additional income by remittances, people tends to work less and to diminish labor supply (Chami, Fullenkamp & Jashjah, 2003) This study addresses the question whether remittance inflows promote economic growth in developing countries in Asia and the Pacific It differs from previous studies in that it examines whether remittances can effect on the impact of labor and capital on growth in developing Asia and the Pacific countries The study uses a balanced panel data on remittance flows to 25 countries in Asia and the Pacific for the period 2000-2012 The results show that while population growth and remittances is complementary, human capital development and remittances are used as substitutes to promote growth The structure of the paper is as follows Chapter provides an overview of existing theories and previous empirical studies Chapter presents the analytical framework, estimation technique and data descriptive Empirical results are described in chapter The last chapter will conclude and provides some policy recommendations Research objectives  To evaluate the impact of remittances inflows on economic growth in developing Asia and the Pacific countries  To examine how remittances affect the impact of investment, population growth and human capital formation on growth in developing Asia and the Pacific countries CHAPTER II: LITERATURE REVIEW Remittance flows are expected to have potential effect on economic growth due to its considerable increase This chapter provides theoretical framework along with empirical studies of the impact of remittances on growth First, remittance definition and remittance in growth model are discussed Then, section examines the channel through which remittances may effect on economic growth In the last section, empirical studies on the factors effect economic growth including remittances, investment, fiscal balance, trade openness, inflation, population growth and human capital formation are reviewed, respectively Remittance definition Remittances take place when one or more family members live and work abroad send money back to their remaining family in the home country (Chami, Cosimano & Gapen, 2006) IMF defined remittances in the fifth edition of Balance of Payments Manual (BPM5) as the sum of three items including worker’s remittances, compensation of employees and migrants’ transfer  Worker’s remittance is the current transfer by a migrant worker, a resident of another country or a worker who stays or expects to stay abroad for more than one year to their home country According to BPM5, it is recorded under current transfer  Compensation of employees is gross earnings of nonresident workers who live abroad for less than one year like border, seasonal worker or local embassy staff… It is included under income in the current account (BPM5)  Migrants’ transfer represents the capital transfer of financial assets by individuals who have a change of residence from one country to another country According to BPM5, it is documented in the capital account of the balance of payments However, it has been argued that the inclusion of migrants’ transfers on remittance calculation is misspecification because of two underlying reasons Firstly, since remittance refers to change in wealth transfer, migrants’ transfers involve assets remain in the same hands of people who have moved their accumulated assets from one country to another country Secondly, there is no special need for any actual flows because of a change in residence status The certain transaction is reclassification of assets Therefore, in the third annual meeting in July 2005, the UN Advisory Experts Group in National Accounts particularly recommended to remove migrants’ transfers from capital account because of no change on ownership Because of the demand on the accuracy of measuring remittance flows, a working group composed of the World Bank, IMF and other international financial institutions was established in order to clarify remittance definition as well as provide guidance for collecting and estimating remittance statistics This technical group made recommendations to the IMF Committee on Balance of Payments Statistics and the Advisory Experts Group in National Accounts with the following items:  Replacing workers’ remittances by personal transfers which focus on household transfer  A new item, personal remittances, is created They will be measured as the sum of personal transfers and net compensation of employees  Migrants’ transfer is removed from the balance of payments framework The assets’ transactions related to changes in individual’s residence will be recorded under other changes of assets and liabilities  The concept of migrant is eliminated in the balance of payments because personal transfers’ definition is based on the residency rather than migration status Remittance in Growth model Economic growth is defined as an increasing not only in actual output over time but also the capacity of the economy to produce goods and services The economists have devoted special attention on the importance of economic growth centuries ago in the attempt to find the way nations become healthier and how to increase the standard of living There are numerous macroeconomists who have contributed significantly to the development of the study of economic growth both in theoretical and empirical However, this part will concentrate on the model that closely relevant to the study The direct and permanent growth effects of enhancing variables like remittances, reforms or globalizations have been captured by using production function However, empirical studies found that in annual data or even with short panel, these effects could not be estimated by regressing the growth rate of output on these enhance variables These effects will be capture through the impact of remittances on total factor productivity (Rao & Hassan, 2011) The model based on the augmented Solow framework, where per worker output y was defined as a function of stock of technology A and capital per worker k The production function takes Cobb-Douglas form with the constant returns: 𝑦𝑡 = 𝐴𝑡 𝑘𝑡𝛼 where < α < (1) The evolution of technology, as assumption of Solow model, is given by: 𝐴𝑡 = 𝐴0 𝑒 𝑔𝑇 (2) A0 is the initial stock of knowledge, T is time and g is the steady state growth of output per worker Thus, the production function for estimation will be modified as: 𝑙𝑛𝑦𝑖𝑡 = 𝑙𝑛𝐴0 + 𝑔𝑇 + 𝑙𝑛𝑘𝑖𝑡 (3) It is also plausible to assume that: 10 Appendix Obs Mean Std D Min Armenia 13 13.50 6.07 4.46 Azerbaijan 13 2.88 0.90 1.08 Bangladesh 13 8.42 2.84 4.18 Cambodia 13 2.39 0.81 1.25 China 13 0.57 0.19 0.39 Fiji 13 5.21 1.14 2.61 Georgia 13 8.28 1.93 5.63 India 13 3.16 0.44 2.60 Indonesia 13 1.04 0.42 0.63 Kazakhstan 13 0.31 0.28 0.08 Kyrgyz Republic 13 14.90 10.81 0.65 Lao PDR 13 0.31 0.39 0.03 Malaysia 13 0.55 0.16 0.37 Maldives 13 0.25 0.10 0.14 Mongolia 13 4.68 2.49 1.06 Nepal 13 15.45 7.49 2.03 Pakistan 13 4.27 1.37 1.45 Papua New Guinea 13 0.14 0.06 0.04 Philippines 13 11.34 1.30 8.59 Solomon Islands 13 0.93 0.68 0.22 Sri Lanka 13 7.95 0.77 7.14 Tajikistan 11 31.75 16.57 6.43 Thailand 13 1.01 0.26 0.64 Tonga 12 25.94 6.90 12.63 Vanuatu 13 3.96 5.85 1.05 Table 6: List of countries and remittances (share of GDP, 2000-2012) 66 Max 19.23 4.71 12.17 3.33 1.01 6.77 11.17 4.08 1.90 0.83 30.75 1.34 0.84 0.51 10.17 24.96 6.23 0.24 13.32 2.32 10.10 49.29 1.38 36.49 20.43 STATA Results for model (12) Dynamic panel-data estimation, two-step system GMM -Group variable: ctry Number of obs = 239 Time variable : Year Number of groups = 24 Number of instruments = 12 Obs per group: = Wald chi2(8) = 135.58 avg = 9.96 Prob > chi2 = 0.000 max = 13 -| GDP | Corrected Coef Std Err z P>|z| [95% Conf Interval] -+ -GDP1 | -2.495173 7488393 -3.33 0.001 -3.962871 -1.027475 Rem | -.0924422 1090861 -0.85 0.397 -.306247 1213626 LnInv | 3.600887 1.105095 3.26 0.001 1.43494 5.766833 Pop | -1.018858 5890527 -1.73 0.084 -2.17338 1356644 LnOpe | 2.229144 1.05602 2.11 0.035 1593825 4.298905 Inf | 0921851 0456161 2.02 0.043 0027792 181591 Fis | 2479678 1699408 1.46 0.145 -.0851101 5810457 LnHcf | 3.09188 1.482294 2.09 0.037 1866382 5.997123 _cons | -10.92825 7.192854 -1.52 0.129 -25.02598 3.169485 -Instruments for first differences equation Standard D.(GDP1 LnInv Pop LnOpe Inf Fis LnHcf) GMM-type (missing=0, separate instruments for each period unless collapsed) L(5/7).Rem collapsed Instruments for levels equation Standard _cons GDP1 LnInv Pop LnOpe Inf Fis LnHcf GMM-type (missing=0, separate instruments for each period unless collapsed) DL4.Rem collapsed -Arellano-Bond test for AR(1) in first differences: z = -1.78 Pr > z = 0.075 Arellano-Bond test for AR(2) in first differences: z = -1.48 Pr > z = 0.140 -Sargan test of overid restrictions: chi2(3) = 4.55 Prob > chi2 = 0.208 Prob > chi2 = 0.598 (Not robust, but not weakened by many instruments.) Hansen test of overid restrictions: chi2(3) = 1.88 (Robust, but can be weakened by many instruments.) Difference-in-Hansen tests of exogeneity of instrument subsets: GMM instruments for levels Hansen test excluding group: chi2(2) = 1.62 Prob > chi2 = 0.446 Difference (null H = exogenous): chi2(1) = 0.26 Prob > chi2 = 0.609 67 STATA Results for model (13a) Dynamic panel-data estimation, two-step system GMM -Group variable: ctry Number of obs = 239 Time variable : Year Number of groups = 24 Number of instruments = 13 Obs per group: = Wald chi2(9) = 162.51 avg = 9.96 Prob > chi2 = 0.000 max = 13 -| GDP | Corrected Coef Std Err z P>|z| [95% Conf Interval] -+ -GDP1 | -2.437564 6835797 -3.57 0.000 -3.777356 -1.097773 Rem | -.2101962 7432044 -0.28 0.777 -1.66685 1.246458 InRem | 0445107 2476176 0.18 0.857 -.4408108 5298322 LnInv | 3.240301 1.751704 1.85 0.064 -.1929767 6.673578 Pop | -.9746381 5920283 -1.65 0.100 -2.134992 185716 LnOpe | 2.225759 1.056902 2.11 0.035 1542684 4.29725 Inf | 0902976 0489134 1.85 0.065 -.0055709 1861662 Fis | 2589094 2149088 1.20 0.228 -.1623041 6801229 LnHcf | 3.027569 1.489896 2.03 0.042 1074269 5.947711 _cons | -10.02764 8.264904 -1.21 0.225 -26.22655 6.171279 -Instruments for first differences equation Standard D.(GDP1 LnInv Pop LnOpe Inf Fis LnHcf InRem LnInv Pop LnOpe Inf Fis LnHcf) GMM-type (missing=0, separate instruments for each period unless collapsed) L(5/7).Rem collapsed Instruments for levels equation Standard _cons GDP1 LnInv Pop LnOpe Inf Fis LnHcf InRem LnInv Pop LnOpe Inf Fis LnHcf GMM-type (missing=0, separate instruments for each period unless collapsed) DL4.Rem collapsed -Arellano-Bond test for AR(1) in first differences: z = -1.78 Pr > z = 0.074 Arellano-Bond test for AR(2) in first differences: z = -1.45 Pr > z = 0.146 -Sargan test of overid restrictions: chi2(3) = 4.49 Prob > chi2 = 0.213 Prob > chi2 = 0.602 (Not robust, but not weakened by many instruments.) Hansen test of overid restrictions: chi2(3) = 1.86 (Robust, but can be weakened by many instruments.) Difference-in-Hansen tests of exogeneity of instrument subsets: GMM instruments for levels Hansen test excluding group: chi2(2) = 1.84 Prob > chi2 = 0.398 Difference (null H = exogenous): chi2(1) = 0.02 Prob > chi2 = 0.885 68 STATA Results for model (13b) Dynamic panel-data estimation, two-step system GMM -Group variable: ctry Number of obs = 239 Time variable : Year Number of groups = 24 Number of instruments = 15 Obs per group: = Wald chi2(9) = 52.06 avg = 9.96 Prob > chi2 = 0.000 max = 13 -| GDP | Corrected Coef Std Err z P>|z| [95% Conf Interval] -+ -GDP1 | -1.85567 8401205 -2.21 0.027 -3.502276 -.209064 Rem | PopRem | -.2929344 1322667 131682 0699804 -2.21 0.027 -.5521723 -.0336964 1.88 0.060 -.005477 2688411 LnInv | 3.429237 Pop | -1.965177 1.563138 2.19 0.028 3655427 6.492932 1.148232 -1.71 0.087 -4.215672 LnOpe | 2853169 1.63952 1.475841 1.11 0.267 -1.253077 4.532116 Inf | 0922795 0682168 1.35 0.176 -.0414229 2259819 Fis | 2688464 2450873 1.10 0.273 -.2115158 7492086 LnHcf | 1.006927 1.830526 0.55 0.582 -2.580839 4.594693 _cons | -2.175491 8.171591 -0.27 0.790 -18.19152 13.84053 -Instruments for first differences equation Standard D.(GDP1 LnInv Pop LnOpe Inf Fis LnHcf PopRem) GMM-type (missing=0, separate instruments for each period unless collapsed) L(4/8).Rem collapsed Instruments for levels equation Standard _cons GDP1 LnInv Pop LnOpe Inf Fis LnHcf PopRem GMM-type (missing=0, separate instruments for each period unless collapsed) DL3.Rem collapsed -Arellano-Bond test for AR(1) in first differences: z = -1.68 Pr > z = 0.092 Arellano-Bond test for AR(2) in first differences: z = -1.54 Pr > z = 0.124 -Sargan test of overid restrictions: chi2(5) = 15.46 Prob > chi2 = 0.009 Prob > chi2 = 0.248 (Not robust, but not weakened by many instruments.) Hansen test of overid restrictions: chi2(5) = 6.65 (Robust, but can be weakened by many instruments.) Difference-in-Hansen tests of exogeneity of instrument subsets: GMM instruments for levels Hansen test excluding group: chi2(4) = 6.60 Prob > chi2 = 0.158 Difference (null H = exogenous): chi2(1) = 0.05 Prob > chi2 = 0.825 69 STATA Results for model (13c) Dynamic panel-data estimation, two-step system GMM -Group variable: ctry Number of obs = 239 Time variable : Year Number of groups = 24 Number of instruments = 13 Obs per group: = Wald chi2(9) = 159.31 avg = 9.96 Prob > chi2 = 0.000 max = 13 -| Corrected GDP | Coef Std Err z P>|z| [95% Conf Interval] -+ -GDP1 | -2.247111 5439117 -4.13 0.000 -3.313158 -1.181064 Rem | 1.688385 HcfRem | -.401767 7279595 2.32 160354 -2.51 0.020 2616105 3.115159 0.012 -.7160549 -.087479 LnInv | 3.843461 Pop | -.9548579 9050555 4.25 5634321 -1.69 0.000 2.069585 5.617337 0.090 -2.059164 LnOpe | 2.590035 1.000618 1494487 2.59 0.010 6288588 4.551211 Inf | 094745 Fis | 2843329 0414989 2.28 0.022 0134087 1760813 1310904 2.17 0.030 0274005 5412653 LnHcf | 3.810698 _cons | -18.32675 1.685979 2.26 0.024 5062398 7.115157 6.941943 -2.64 0.008 -31.93271 -4.720792 -Instruments for first differences equation Standard D.(GDP1 LnInv Pop LnOpe Inf Fis LnHcf HcfRem) GMM-type (missing=0, separate instruments for each period unless collapsed) L(1/3).Rem collapsed Instruments for levels equation Standard _cons GDP1 LnInv Pop LnOpe Inf Fis LnHcf HcfRem GMM-type (missing=0, separate instruments for each period unless collapsed) D.Rem collapsed -Arellano-Bond test for AR(1) in first differences: z = -1.77 Pr > z = 0.077 Arellano-Bond test for AR(2) in first differences: z = -1.61 Pr > z = 0.107 -Sargan test of overid restrictions: chi2(3) = 0.43 Prob > chi2 = 0.934 Prob > chi2 = 0.985 (Not robust, but not weakened by many instruments.) Hansen test of overid restrictions: chi2(3) = 0.15 (Robust, but can be weakened by many instruments.) Difference-in-Hansen tests of exogeneity of instrument subsets: GMM instruments for levels Hansen test excluding group: chi2(2) = 0.15 Prob > chi2 = 0.926 Difference (null H = exogenous): chi2(1) = 0.00 Prob > chi2 = 0.999 70 Test of endogeneity for model (12) Number of obs = F( 8, 150) = 159 5.50 Prob > F = 0.0000 Centered R2 = 0.1936 7500.871981 Uncentered R2 = 0.6171 2871.942999 Root MSE Total (centered) SS = 3561.62588 Total (uncentered) SS = Residual SS = = 4.25 -GDP | Coef Std Err z P>|z| [95% Conf Interval] -+ -Rem | -.1556016 0645125 -2.41 0.016 -.2820438 -.0291595 GDP1 | -2.467077 9457521 -2.61 0.009 -4.320717 -.6134373 LnInv | 2.625155 1.920347 1.37 0.172 -1.138657 6.388967 Pop | -.1373066 6127865 -0.22 0.823 -1.338346 1.063733 LnOpe | 1.652266 1.005063 1.64 0.100 -.3176211 3.622152 Inf | 2319654 0823442 2.82 0.005 0705737 3933571 Fis | 3057292 1527263 2.00 0.045 0063912 6050672 LnHcf | 3.73075 2.11036 1.77 0.077 -.4054792 7.86698 _cons | -9.759126 6.704677 -1.46 0.146 -22.90005 3.3818 -Underidentification test (Anderson canon corr LM statistic): Chi-sq(1) P-val = 81.927 0.0000 -Weak identification test (Cragg-Donald Wald F statistic): 159.448 Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005) Reproduced by permission -Sargan statistic (overidentification test of all instruments): 0.000 (equation exactly identified) -endog- option: Endogeneity test of endogenous regressors: 4.715 Chi-sq(1) P-val = Regressors tested: 0.0299 Rem -Instrumented: Rem Included instruments: GDP1 LnInv Pop LnOpe Inf Fis LnHcf Excluded instruments: L4.Rem 71 Test of endogeneity for model (13a) Number of obs = F( 9, 159 149) = 2.32 Prob > F = 0.0180 Centered R2 = -0.6925 0.1963 Total (centered) SS = 3561.62588 Total (uncentered) SS = 7500.871981 Uncentered R2 = Residual SS = 6028.091237 Root MSE = 6.157 -GDP | Coef Std Err z P>|z| [95% Conf Interval] -+ -Rem | -4.571949 3.467727 -1.32 0.187 -11.36857 GDP1 | -4.059517 2.217833 -1.83 0.067 -8.40639 2.224671 287356 LnInv | -7.185923 9.212125 -0.78 0.435 -25.24135 10.86951 Pop | 3.402998 2.8117 1.21 0.226 -2.107833 8.913829 LnOpe | 1.604435 1.452775 1.10 0.269 -1.242953 4.451822 Inf | 0514024 1692374 0.30 0.761 -.2802969 3831017 Fis | 3428735 2166127 1.58 0.113 -.0816796 7674267 LnHcf | 8.93677 6.177089 1.45 0.148 -3.170101 21.04364 InRem | 1.457718 1.123068 1.30 0.194 -.7434558 3.658891 _cons | 6.283667 16.69979 0.38 0.707 -26.44732 39.01465 -Underidentification test (Anderson canon corr LM statistic): Chi-sq(1) P-val = 4.190 0.0407 -Weak identification test (Cragg-Donald Wald F statistic): 4.033 Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005) Reproduced by permission -Sargan statistic (overidentification test of all instruments): 0.000 (equation exactly identified) -endog- option: Endogeneity test of endogenous regressors: 5.207 Chi-sq(1) P-val = Regressors tested: 0.0225 Rem -Instrumented: Rem Included instruments: GDP1 LnInv Pop LnOpe Inf Fis LnHcf InRem Excluded instruments: L4.Rem 72 Test of endogeneity for model (13b) Estimates efficient for homoskedasticity only Statistics consistent for homoskedasticity only Number of obs = F( 9, 179 169) = 5.87 Prob > F = 0.0000 Total (centered) SS = 3774.027087 Centered R2 = 0.2243 Total (uncentered) SS = 8249.191715 Uncentered R2 = 0.6451 Residual SS = 2927.39444 Root MSE = 4.044 -GDP | Coef Std Err z P>|z| [95% Conf Interval] -+ -Rem | -.3946207 1079384 -3.66 0.000 -.6061761 -.1830652 GDP1 | -1.514972 7447389 -2.03 0.042 -2.974633 -.0553105 LnInv | 3.873551 1.506274 2.57 0.010 9213082 6.825793 Pop | -2.367091 8297002 -2.85 0.004 -3.993274 -.740909 LnOpe | 1.111179 8939668 1.24 0.214 -.6409639 2.863322 Inf | 1590563 0728244 2.18 0.029 0163232 3017895 6802532 Fis | 4173518 1341358 3.11 0.002 1544503 LnHcf | 9565206 1.589442 0.60 0.547 -2.158729 4.07177 PopRem | 1835744 0537648 3.41 0.001 0781974 2889514 _cons | -2.064102 6.242971 -0.33 0.741 -14.3001 10.1719 -Underidentification test (Anderson canon corr LM statistic): Chi-sq(1) P-val = 80.577 0.0000 -Weak identification test (Cragg-Donald Wald F statistic): 138.356 Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005) Reproduced by permission -Sargan statistic (overidentification test of all instruments): 0.000 (equation exactly identified) -endog- option: Endogeneity test of endogenous regressors: 2.966 Chi-sq(1) P-val = Regressors tested: 0.0850 Rem -Instrumented: Rem Included instruments: GDP1 LnInv Pop LnOpe Inf Fis LnHcf PopRem Excluded instruments: L3.Rem 73 Test of endogeneity for model (13c) Number of obs = F( 9, 159 149) = 3.17 Prob > F = 0.0015 Centered R2 Total (centered) SS = 3561.62588 = -0.2450 Total (uncentered) SS = 7500.871981 Uncentered R2 = 0.4088 Residual SS = 4434.386513 Root MSE = 5.281 -GDP | Coef Std Err z P>|z| [95% Conf Interval] -+ -Rem | 8.063657 4.743793 1.70 0.089 -1.234007 17.36132 GDP1 | 0897349 1.413258 0.06 0.949 -2.680199 2.859669 LnInv | 7.699304 2.777313 2.77 0.006 2.255871 13.14274 Pop | 413688 8065553 0.51 0.608 -1.167131 1.994507 LnOpe | 2.896289 1.508482 1.92 0.055 -.0602808 5.852859 Inf | 1215075 1124826 1.08 0.280 -.0989544 3419693 Fis | 63702 2382766 2.67 0.008 1700065 1.104034 LnHcf | 6.764558 3.855011 1.75 0.079 -.7911245 14.32024 HcfRem | -1.819523 1.062404 -1.71 0.087 -3.901797 2627512 _cons | -62.61081 30.8758 -2.03 0.043 -123.1263 -2.09535 -Underidentification test (Anderson canon corr LM statistic): Chi-sq(1) P-val = 6.806 0.0091 -Weak identification test (Cragg-Donald Wald F statistic): 6.664 Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005) Reproduced by permission -Sargan statistic (overidentification test of all instruments): 0.000 (equation exactly identified) -endog- option: Endogeneity test of endogenous regressors: 4.214 Chi-sq(1) P-val = Regressors tested: 0.0401 Rem -Instrumented: Rem Included instruments: GDP1 LnInv Pop LnOpe Inf Fis LnHcf HcfRem Excluded instruments: L4.Rem 74 Test of omitted variables and heteroskedasticity Source | SS df MS Number of obs = -+ F( 231) = 10.94 Model | 1338.39044 191.198635 Prob > F = 0.0000 Residual | 4038.73259 231 17.4836909 R-squared = 0.2489 -+ -Total | 5377.12303 238 22.5929539 7, 239 Adj R-squared = 0.2261 Root MSE 4.1814 = -GDP | Coef Std Err t P>|t| [95% Conf Interval] -+ -Rem | -.0496972 0314963 -1.58 0.116 -.111754 0123596 LnInv | 5.487497 1.002821 5.47 0.000 3.511652 7.463341 Pop | -.586921 4495826 -1.31 0.193 -1.472728 2988857 LnOpe | 6374217 660247 0.97 0.335 -.6634541 1.938298 Inf | 1792707 061403 2.92 0.004 0582892 3002521 Fis | 2876231 1068384 2.69 0.008 0771209 4981253 LnHcf | -.2184008 9994574 -0.22 0.827 -2.187618 1.750817 _cons | -13.72694 4.855817 -2.83 0.005 -23.29429 -4.159587 -(84 missing values generated) (86 missing values generated) (84 missing values generated) (84 missing values generated) (84 missing values generated) (3 missing values generated) (3 missing values generated) (3 missing values generated) (27 missing values generated) (27 missing values generated) (27 missing values generated) (1 missing value generated) (1 missing value generated) (1 missing value generated) (15 missing values generated) (15 missing values generated) (15 missing values generated) (71 missing values generated) (71 missing values generated) (71 missing values generated) Two versions of Ramsey's RESET test: First, as a test of heteroskedasticity: Source | SS df MS 75 Number of obs = 239 -+ 235) = 1.51 Model | 76.1415336 25.3805112 Prob > F F( = 0.2139 Residual | 3962.59106 235 16.8620896 R-squared = 0.0189 -+ -Total | 4038.73259 238 16.9694647 3, Adj R-squared = 0.0063 Root MSE 4.1063 = -_res | Coef Std Err t P>|t| [95% Conf Interval] -+ -yh2 | -.0694289 0440093 -1.58 0.116 -.1561319 0172742 yh3 | -.0003108 0030614 -0.10 0.919 -.0063421 0057204 yh4 | 0011101 0005497 2.02 0.045 0000272 002193 _cons | 65043 6072562 1.07 0.285 -.5459315 1.846792 -( 1) yh2 = ( 2) yh3 = ( 3) yh4 = F( 3, 235) = Prob > F = 1.51 0.2139 Second, as a test of whether there are any omitted variables, or the functional form is misspecified: Source | SS df MS Number of obs = -+ F( 10, 239 228) = 8.21 Model | 1423.36936 10 142.336936 Prob > F = 0.0000 Residual | 3953.75368 228 17.3410249 R-squared = 0.2647 -+ -Total | 5377.12303 238 22.5929539 Adj R-squared = 0.2325 Root MSE 4.1643 = -GDP | Coef Std Err t P>|t| [95% Conf Interval] -+ -Rem | -.0613207 036424 -1.68 0.094 -.1330913 LnInv | 6.494728 2.34254 2.77 0.006 1.878933 0104499 11.11052 Pop | -.8035161 5148502 -1.56 0.120 -1.817989 2109567 LnOpe | 7281875 7001069 1.04 0.299 -.6513194 2.107694 Inf | 2056382 097772 2.10 0.037 0129859 3982904 Fis | 3891232 1686154 2.31 0.022 0568796 7213669 LnHcf | -.2985797 1.006655 -0.30 0.767 -2.282115 1.684956 yh2 | -.0969962 0658334 -1.47 0.142 -.2267158 0327234 yh3 | -.0035472 0064467 -0.55 0.583 -.01625 0091555 yh4 | 0015715 000934 1.68 0.094 -.0002689 0034119 _cons | -15.09977 7.417817 -2.04 0.043 -29.716 -.4835278 76 ( 1) yh2 = ( 2) yh3 = ( 3) yh4 = F( 3, 228) = Prob > F = Source | 1.63 0.1824 SS df MS Number of obs = -+ F( 28, 239 210) = 5.86 Model | 2358.19392 28 84.2212115 Prob > F = 0.0000 Residual | 3018.92911 210 14.3758529 R-squared = 0.4386 Adj R-squared = 0.3637 Root MSE 3.7916 -+ -Total | 5377.12303 238 22.5929539 = -GDP | Coef Std Err t P>|t| [95% Conf Interval] -+ -Rem | 5801118 3119123 1.86 0.064 -.0347686 1.194992 LnInv | 782.5278 430.5879 1.82 0.071 -66.30079 1631.356 Pop | -2.742311 1.208161 -2.27 0.024 -5.12399 -.3606334 LnOpe | -941.3617 1140.631 -0.83 0.410 -3189.916 1307.193 Inf | -.9782323 430216 -2.27 0.024 -1.826328 -.1301367 Fis | 1313765 2760529 0.48 0.635 -.4128134 6755663 LnHcf | 427.117 1438.815 0.30 0.767 -2409.254 3263.488 zsq1 | -.0665552 0339668 -1.96 0.051 -.1335148 0004043 zcu1 | 0020703 0012439 1.66 0.098 -.0003818 0045224 zqu1 | -.0000184 0000141 -1.30 0.194 -.0000462 9.43e-06 zsq2 | -395.2635 228.7904 -1.73 0.086 -846.2837 55.75676 zcu2 | 87.24544 53.01671 1.65 0.101 -17.26772 191.7586 zqu2 | -7.065623 4.53033 -1.56 0.120 -15.99637 1.865129 zsq3 | 6035395 2.240636 0.27 0.788 -3.813482 5.020561 zcu3 | 1.976992 2.029763 0.97 0.331 -2.024331 5.978315 zqu3 | -.8297459 5095939 -1.63 0.105 -1.834321 1748292 zsq4 | 331.9224 399.8984 0.83 0.407 -456.4073 1120.252 zcu4 | -51.56623 61.85873 -0.83 0.405 -173.5099 70.37742 zqu4 | 2.982198 3.563289 0.84 0.404 -4.042202 10.0066 zsq5 | 2091802 0844368 2.48 0.014 0427279 3756324 zcu5 | -.0125902 0060959 -2.07 0.040 -.0246072 -.0005732 zqu5 | 0002341 000138 1.70 0.091 -.0000379 0005061 zsq6 | -.2209145 0725143 -3.05 0.003 -.3638638 -.0779652 zcu6 | -.0254867 0138162 -1.84 0.066 -.052723 0017496 zqu6 | -.0005953 0005441 -1.09 0.275 -.0016679 0004774 zsq7 | -161.4398 569.1387 -0.28 0.777 -1283.397 960.5175 zcu7 | 26.71508 99.289 0.27 0.788 -169.0158 222.4459 zqu7 | -1.636598 6.449494 -0.25 0.800 -14.35065 11.07745 _cons | 6.887828 1723.61 0.00 0.997 -3390.907 3404.683 77 -( 1) zsq1 = ( 2) zcu1 = ( 3) zqu1 = ( 4) zsq2 = ( 5) zcu2 = ( 6) zqu2 = ( 7) zsq3 = ( 8) zcu3 = ( 9) zqu3 = (10) zsq4 = (11) zcu4 = (12) zqu4 = (13) zsq5 = (14) zcu5 = (15) zqu5 = (16) zsq6 = (17) zcu6 = (18) zqu6 = (19) zsq7 = (20) zcu7 = (21) zqu7 = F( 21, 210) = Prob > F = 3.38 0.0000 78 t-test for two group: high and low remittances Two-sample t test with unequal variances -Group | Obs Mean Std Err Std Dev [95% Conf Interval] -+ -0 | 167 4.83976 4300707 5.557738 3.990647 5.688874 | 156 3.981741 3200135 3.996968 3.349591 4.613892 -+ -combined | 323 4.425361 2714396 4.878366 3.891342 4.95938 -+ -diff | 8580192 5360685 -.196888 1.912926 -diff = mean(0) - mean(1) Ho: diff = t = 1.6006 Satterthwaite's degrees of freedom = 301.668 Ha: diff < Ha: diff != Ha: diff > Pr(T < t) = 0.9447 Pr(|T| > |t|) = 0.1105 Pr(T > t) = 0.0553 79 Collinearity Diagnostics SQRT Variable VIF VIF RTolerance Squared -Rem 1.47 1.21 0.6814 0.3186 GDP1 2.67 1.63 0.3742 0.6258 LnInv 1.62 1.27 0.6164 0.3836 Pop 1.57 1.25 0.6352 0.3648 LnOpe 1.74 1.32 0.5747 0.4253 Inf 1.14 1.07 0.8780 0.1220 Fis 1.13 1.06 0.8821 0.1179 LnHcf 2.94 1.72 0.3400 0.6600 -Mean VIF 1.79 Cond Eigenval Index 2.4348 1.0000 1.4746 1.2850 1.3466 1.3446 0.9093 1.6364 0.7012 1.8635 0.5929 2.0265 0.3570 2.6115 0.1836 3.6412 Condition Number 3.6412 Eigenvalues & Cond Index computed from deviation sscp (no intercept) Det(correlation matrix) 0.1198 80